Sender header?

Sender header?

[Kai Großjohann]

What does Gnus use these days to find out if a Sender header should be
added?

Proposal: by default, don't add a Sender header if the message does
not contain a From header.  Don't add a Sender header if the message
contains a From header which is equal to the value computed
automatically (for folks with message-generate-headers-first set to
t).  Only add a Sender header if the From header is different from the
normal value.

What do you think?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> Only add a Sender header if the From header is different from the
> normal value.

Are you talking about mail or news?  Sender means something different
for each.


paul

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Wed, 23 May 2001
| Proposal: by default, don't add a Sender header if the message does
| not contain a From header.

This case is fundamentally flawed.  There are times when a Sender header
MUST be generated by the MUA before the message is handed to the MTA.  This
cannot be done correctly unless a From header exists.

Or do you mean that Gnus should automatically generate a From header at the
time of sumission instead of generating a Sender header?  I think this is
dangerous because if Gnus already generates bad Sender headers, it is going
to generate equivalently bad From headers.

| Don't add a Sender header if the message contains a From header which is
| equal to the value computed automatically (for folks with
| message-generate-headers-first set to t).

Correct.  RFC 2822 qualifies this as a "SHOULD NOT" case.  It is allowed
but discouraged to have duplicate From and Sender field mailboxes.  Note
that Sender MUST be a single mailbox whereas From may be many.

| Only add a Sender header if the From header is different from the normal
| value.

This is what Gnus does now, yes?
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball may stick to certain types
Minion of Nathan - Nathan says Hi! \ of skin.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Kai Großjohann]

On Wed, 23 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Wed, 23 May
>   2001
> | Proposal: by default, don't add a Sender header if the message does
> | not contain a From header.
> 
> This case is fundamentally flawed.  There are times when a Sender
> header MUST be generated by the MUA before the message is handed to
> the MTA.  This cannot be done correctly unless a From header exists.
> 
> Or do you mean that Gnus should automatically generate a From header
> at the time of sumission instead of generating a Sender header?  I
> think this is dangerous because if Gnus already generates bad Sender
> headers, it is going to generate equivalently bad From headers.

By default, Gnus creates a From header for all outgoing messages.
This is because From is mentioned in message-required-mail-headers and
message-required-news-headers.

> | Don't add a Sender header if the message contains a From header
> | which is equal to the value computed automatically (for folks with
> | message-generate-headers-first set to t).
> 
> Correct.  RFC 2822 qualifies this as a "SHOULD NOT" case.  It is
> allowed but discouraged to have duplicate From and Sender field
> mailboxes.  Note that Sender MUST be a single mailbox whereas From
> may be many.
> 
> | Only add a Sender header if the From header is different from the
> | normal value.
> 
> This is what Gnus does now, yes?

No.  As you can see in this message, Gnus automatically adds a Sender
header.  I have changed the variable user-mail-address.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 23 May 2001, Paul Jarc wrote:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>> Only add a Sender header if the From header is different from the
>> normal value.
> 
> Are you talking about mail or news?  Sender means something
> different for each.

I guess I'm primarily talking about mail.  But just now I looked at
RFC 1036, and the example given there does not match what Gnus is
doing.  Gnus generates a Sender header as user@host.domain.com, but
look at the example given there:

              From: smith@ucbvax.Berkeley.EDU (John Smith)
              Sender: jones@cca.COM (Sarah Jones)

Surely, `cca.com' is not the host name of the machine?  Surely the
host name would be something like `foo.cca.com'?

But I'm not sure what son of RFC 1036 and grandson of RFC 1036 say
about this.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
| By default, Gnus creates a From header for all outgoing messages.
| This is because From is mentioned in message-required-mail-headers and
| message-required-news-headers.

And nobody complains about From being wrong.  Therefore it is safe to
assume that in cases where a correct From field is (or would be) generated,
a correct Sender field would also be generated when the From field is not
canonical for the sending host.

| > | Only add a Sender header if the From header is different from the
| > | normal value.
| > This is what Gnus does now, yes?
| No.  As you can see in this message, Gnus automatically adds a Sender
| header.  I have changed the variable user-mail-address.

Then what I said is correct, and what Gnus does is also correct as far as
RFC 2822 is concerned.

I have several mailboxes: ratinox@rei.nerv.gweep.net, ratinox@newsguy.com,
SamuraiRat@hotmail.com, ratinox@peorth.gweep.net, and others.  They are all
me, but that does not mean ratinox@newsguy.com == ratinox@peorth.gweep.net.
They are different mailboxes.  If I were to send a message from my local
machine (peorth.gweep.net) and address it from my Hotmail address, the MUA
is required to generate a Sender header with ratinox@peorth.gweep.net as
its contents.  You have done exactly the same thing.  The fact that your
mailboxes are within the same domain is irrelevant, and lucy should have an
MX record that points it at the cs.uni-dortmund.de mail servers, which in
fact it does, which makes Kai.Grossjohann@lucy.CS.Uni-Dortmund.DE a valid
mailbox.

Put susinctly, Gnus is doing exactly what it should, when it should,
according to RFC 2822, and I do not understand why you want to break it :).
-- 
Rat <ratinox@peorth.gweep.net>    \ When not in use, Happy Fun Ball should be
Minion of Nathan - Nathan says Hi! \ returned to its special container and
PGP Key: at a key server near you!  \ kept under refrigeration.

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Thu, 24 May
>   2001
> | By default, Gnus creates a From header for all outgoing messages.
> | This is because From is mentioned in message-required-mail-headers and
> | message-required-news-headers.
> 
> And nobody complains about From being wrong.  Therefore it is safe
> to assume that in cases where a correct From field is (or would be)
> generated, a correct Sender field would also be generated when the
> From field is not canonical for the sending host.

I dare say that people don't complain because people can frob their
>From header via user-mail-address.  They cannot, however, frob their
Sender header via user-mail-address.

Right now, if user-mail-address is set, Gnus behaves as follows:

  - If no From header is specified, generate one using
    user-mail-address.

  - Generate a Sender header using user-login-name, followed by "@",
    followed by system-name.

  - If the Sender header is redundant, delete it.

I wish to change it as follows:

  - If no From header is specified, generate one using
    user-mail-address.

  - Generate a Sender header using user-mail-address.

  - If the Sender header is redundant, delete it.

It seems that my suggestion is what you want, since you assume that
the algorithms for generating From and Sender headers are the same.
However, in the current Gnus they are not the same.  I want to make
them the same.

> | > | Only add a Sender header if the From header is different from
> | > | the normal value.
> | > This is what Gnus does now, yes?
> | No.  As you can see in this message, Gnus automatically adds a
> | Sender header.  I have changed the variable user-mail-address.
> 
> Then what I said is correct, and what Gnus does is also correct as
> far as RFC 2822 is concerned.

I don't know what does RFC 2822 say.  I only know about RFC 822.  The
local RFC server doesn't seem to know about this RFC.  Can you help
out?

> I have several mailboxes: ratinox@rei.nerv.gweep.net,
> ratinox@newsguy.com, SamuraiRat@hotmail.com,
> ratinox@peorth.gweep.net, and others.  They are all me, but that
> does not mean ratinox@newsguy.com == ratinox@peorth.gweep.net.  They
> are different mailboxes.  If I were to send a message from my local
> machine (peorth.gweep.net) and address it from my Hotmail address,
> the MUA is required to generate a Sender header with
> ratinox@peorth.gweep.net as its contents.

I'm with you so far.  I presume that you have set user-mail-address to
"ratinox@peorth.gweep.net".  Then my proposal does what you want.

> You have done exactly the same thing.  The fact that your mailboxes
> are within the same domain is irrelevant, and lucy should have an MX
> record that points it at the cs.uni-dortmund.de mail servers, which
> in fact it does, which makes Kai.Grossjohann@lucy.CS.Uni-Dortmund.DE
> a valid mailbox.

I am lucky that user-login-name at system-name is a valid mail address
and that mail sent there reaches me.  However, not everybody might be
as lucky.

As I see it, Gnus has no way of automatically finding out what is the
email address of the user running Emacs.  Therefore, it should be
possible to tell Gnus.  But just this is not possible.

> Put susinctly, Gnus is doing exactly what it should, when it should,
> according to RFC 2822, and I do not understand why you want to break
> it :).

I do not understand why you think I'm breaking anything.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Stainless Steel Rat]

Before this gets too long, here is a hint: system-name.  Think about it for
a moment.

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
| I dare say that people don't complain because people can frob their
| From header via user-mail-address.  They cannot, however, frob their
| Sender header via user-mail-address.

And they should not.  Frobbing it directly would be a violation of the
requirements of RFCs 2822 and 1034.

[...]
|   - Generate a Sender header using user-login-name, followed by "@",
|     followed by system-name.

Which is as canonical as a program can get.  Sender is supposed to be
canonical.  By "canonical" I mean "an attempt has been made to ensure that
the mailbox is valid".  For a correctly configured system, and that
includes the mail hubs and what-not, not just the local machine, `login at
FQDN' is canonical.

I apparantly misunderstood something, because using user-mail-address for
generating Sender would break that.  Therefore it should not be done.

[...]
| I don't know what does RFC 2822 say.  I only know about RFC 822.  The
| local RFC server doesn't seem to know about this RFC.  Can you help
| out?

RFCs 2821 and 2822 obsoleted RFCs 821 and 822 about a month ago.  They
clarify a lot of things, not the least of which is Sender, both by standard
and defacto use.

[...]
| I'm with you so far.  I presume that you have set user-mail-address to
| "ratinox@peorth.gweep.net".  Then my proposal does what you want.

No, I don't.  I have my system configured correctly.  It knows itself as
peorth.gweep.net, so Emacs knows it as peorth.gweep.net, so Gnus knows it
as peorth.gweep.net, and nothing needs to be kludged.

You seem to believe that because only one person is involved in originating
a message and submitting it that the one person has only one identity.
This is not true.  If I were to send a message from rei.nerv.gweep.net (one
of peorth's MX hosts), I am still ratinox@peorth, but I am also
ratinox@rei.nerv.  As the originator of a message I am ratinox@peorth and I
put that in the From header -- how I accomplish that is not relevant.  As
the sender I am ratinox@rei.nerv, and that goes in the Sender header.  It
is up to rei's admins to ensure that ratinox@rei.nerv is deliverable.  And
everything Just Works, no fuss, no muss.

Handling mail incorrectly seems easy, until you discover that everything is
an exception.  Doing it right is hard to set up, but there are few special
cases and everything else Just Works.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball may stick to certain types
Minion of Nathan - Nathan says Hi! \ of skin.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> You seem to believe that because only one person is involved in originating
> a message and submitting it that the one person has only one identity.
> This is not true.

A person may have multiple mailboxes, but where does RFC 2822 say that
a mailbox is an identity?

The RFC doesn't seem to be aware of the multiple-mailbox case, though.
3.6.2:
# The "Sender:" field specifies the mailbox of the agent responsible
# for the actual transmission of the message.
Note "the mailbox", not "a mailbox". :(

Also, I see no indication here that Sender should be tied to the
system where the message originated.  All the examples of uses of
Sender involve one person sending a message on behalf of someone else;
none of them involve one person sending their own message with a From
field that doesn't indicate the system they're sending from.  An
"agent" is apparently a person, and your mail address on one system
still identifies you even when you're sending mail from another.

By your interpretation, the RFC is requiring that every machine that
sends mail must be usable in a recipient address - i.e., the MX (or A)
for that name must be configured to accept mail addressed to that
individual host.  (Otherwise, user-login-name@system-name wouldn't be
a mailbox of the originator of the message.)  Do you think the authors
really intended to require this?


paul

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
| A person may have multiple mailboxes, but where does RFC 2822 say that
| a mailbox is an identity?

A mailbox is a mailbox.  A mailbox can be considered to identify its
owner.  That is mine.

[...]
| By your interpretation, the RFC is requiring that every machine that
| sends mail must be usable in a recipient address - i.e., the MX (or A)
| for that name must be configured to accept mail addressed to that
| individual host.  (Otherwise, user-login-name@system-name wouldn't be
| a mailbox of the originator of the message.)  Do you think the authors
| really intended to require this?

Yes, that is exactly what I think.  A records are required for all machines
on the Internet that have anything to do with mail, and MX records should
exist for all such hosts.  Proper mail handling depends on that being the
case.  If your host has neither A nor MX records then it is not configured
correctly.  Trying to make Gnus work around that does not fix the problem.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball may stick to certain types
Minion of Nathan - Nathan says Hi! \ of skin.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
>| By your interpretation, the RFC is requiring that every machine that
>| sends mail must be usable in a recipient address - i.e., the MX (or A)
>| for that name must be configured to accept mail addressed to that
>| individual host.  (Otherwise, user-login-name@system-name wouldn't be
>| a mailbox of the originator of the message.)  Do you think the authors
>| really intended to require this?
> 
> Yes, that is exactly what I think.  A records are required for all machines
> on the Internet that have anything to do with mail, and MX records should
> exist for all such hosts.

But your interpretation requires more than that.  It's not enough just
to give random-sending-host.domain.com an MX record pointing to
mail.domain.com - mail.domain.com must also accept mail explicitly
addressed to random-sending-host.domain.com, even if it would
otherwise only accept mail addressed to domain.com itself, because you
would put random-sending-host.domain.com in Sender, and Sender must be
a mailbox of the person who sends the message.

Where does 2822 say that Sender must identify the host used to send
the mail?

> Proper mail handling depends on that being the case.

What breaks when Sender does not identify the host the mail was sent
from?

> If your host has neither A nor MX records then it is not configured
> correctly.

Right, but the A and MX records aren't the problem.


paul

Re: Sender header?

[Christoph Conrad]

Hello Kai,

    > No. As you can see in this message, Gnus automatically adds a
    > Sender header. I have changed the variable user-mail-address.

I C-u g this article, no sender header...

Best regards,
cu, -cc-
-- 
=> GNU Emacs Webring @ <http://www.gnusoftware.com/WebRing/> <=
Look Ma, this man can twist his fingers as if they were made of rubber,
isn't that amazing? -- Not really, he's been using emacs for years...!

Re: Sender header?

[Paul Jarc]

Christoph Conrad <christoph.conrad@gmx.de> writes:
>    > No. As you can see in this message, Gnus automatically adds a
>    > Sender header. I have changed the variable user-mail-address.
> 
> I C-u g this article, no sender header...

I see:
Sender: owner-ding@hpc.uh.edu
But that's obviously not Gnus's doing.

Sending myself a message, I see:
Sender: prj@multivac.cwru.edu
To: prj@multivac.cwru.edu
Subject: test
From: prj@po.cwru.edu (Paul Jarc)


paul

Re: Sender header?

[Graham Murray]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:

> Yes, that is exactly what I think.  A records are required for all machines
> on the Internet that have anything to do with mail, and MX records should
> exist for all such hosts.  Proper mail handling depends on that being the
> case.  If your host has neither A nor MX records then it is not configured
> correctly.  Trying to make Gnus work around that does not fix the problem.

I think that there are at least 2 situations in which that does not
apply.

1) Dial-up hosts using dynamic IP. In this situation, the user is
   often given a mailbox of user@isp.com. There are MX records for
   isp.com which points at the ISP's mail server(s). However there is
   no (fixed) A record for the customer's system, and its name is
   dynamically assigned when the system connects to the
   ISP. Therefore, it is not possible to configure the system with its
   FQDN.

2) Hosts on an internal LAN (probably using RFC 1918 addresses) behind
   a firewall. All users on the LAN are allocated mailboxes in the
   form of user@x.com. Internally the hosts on the LAN are configured
   as a.x.com, b.x.com etc. The internal host names should never be
   visible to the Internet "at large". Currently if user@a.x.com
   sends mail with a from address of user1@x.com (the externally
   visible mailbox), gnus will set the Sender to user@a.x.com - which
   address should be visible outside of x.com's internal network.

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
| But your interpretation requires more than that.  It's not enough just
| to give random-sending-host.domain.com an MX record pointing to
| mail.domain.com - mail.domain.com must also accept mail explicitly
| addressed to random-sending-host.domain.com, even if it would
| otherwise only accept mail addressed to domain.com itself, because you
| would put random-sending-host.domain.com in Sender, and Sender must be
| a mailbox of the person who sends the message.

Yes, it should.  In fact, it must.  That is the nature of DNS.

| Where does 2822 say that Sender must identify the host used to send
| the mail?

The canonical mailbox is required by RFC 2822.  login @ FQDN is the most
canonical name for any given user.

| > Proper mail handling depends on that being the case.
| What breaks when Sender does not identify the host the mail was sent
| from?

The correct question is, what breaks the Sender field does not canonically
identify the sender?  The answer is, if the Sender field does not contain
the canonical mailbox of the sender and there is a problem, perhaps with
that machine, I may be unable to contact the sender.

You see, Sender is for human consumption only.  It exists at least
partially so that humans, like me, can track down problems at their
source.  If the Sender field does not include the FQDN of the sending host
it makes finding and solving problems that much more difficult.

| > If your host has neither A nor MX records then it is not configured
| > correctly.
| Right, but the A and MX records aren't the problem.

Correct, they are the solution.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
>| But your interpretation requires more than that.  It's not enough just
>| to give random-sending-host.domain.com an MX record pointing to
>| mail.domain.com - mail.domain.com must also accept mail explicitly
>| addressed to random-sending-host.domain.com, even if it would
>| otherwise only accept mail addressed to domain.com itself, because you
>| would put random-sending-host.domain.com in Sender, and Sender must be
>| a mailbox of the person who sends the message.
> 
> Yes, it should.  In fact, it must.  That is the nature of DNS.

It is not the nature of DNS (or anything else) that the mail exchanger
for domain.com must also accept mail for any other domain, such as
random-sending-host.domain.com.

>| Where does 2822 say that Sender must identify the host used to send
>| the mail?
> 
> The canonical mailbox is required by RFC 2822.

I didn't ask you to restate the requirement.  I asked where it was.
RFC 2822 does not use the word "canonical" in connection with Sender.
I haven't been able to find the requirement.

> login @ FQDN is the most canonical name for any given user.

It isn't a canonical address if it isn't an address at all.  It also
isn't known to be canonical at all - it's merely the only possible
address that can be determined automatically.  user-mail-address is
extremely more likely to be the user's canonical address.  A user's
canonical address does not depend on what machine ey happens to be
using at the moment.

>| > Proper mail handling depends on that being the case.
>| What breaks when Sender does not identify the host the mail was sent
>| from?
> 
> The correct question is, what breaks the Sender field does not canonically
> identify the sender?

No, it isn't.  You're demanding that Sender identify not just the
person who sent the message, but also the host it was sent from.  So I
ask, what will break if your additional requirement isn't met?

> You see, Sender is for human consumption only.  It exists at least
> partially so that humans, like me, can track down problems at their
> source.

I can't find any support for that statement in RFC 2822.

> If the Sender field does not include the FQDN of the sending host
> it makes finding and solving problems that much more difficult.

You still have Received fields.  Sender is unreliable anyway, since
it's under the control of a possible malicious person.

>| Right, but the A and MX records aren't the problem.
> 
> Correct, they are the solution.

They are necessary, but not sufficient, to make your requirement
work.


paul

Re: Sender header?

[Stainless Steel Rat]

* Graham Murray <graham@barnowl.demon.co.uk>  on Thu, 24 May 2001
| 1) Dial-up hosts using dynamic IP. In this situation, the user is often
|    given a mailbox of user@isp.com. [snip]

There must be an A record (and FQDN) for the IP address.  A properly
configured dialup server will provide the FQDN associated with the IP
address, and a properly configured dialup host will set its name to that
FQDN for the duration of its connection.  Gnus generates Sender fields
using that FQDN (or it should).  The MX record associated with the A record
will make it a deliverable mailbox.

Correct configuration elminates the perceived need to coerce Gnus into
doing things that are less than Kosher.

| 2) Hosts on an internal LAN (probably using RFC 1918 addresses) behind a
|    firewall. [snip]

Internally those fields should be as canonical as possible.  It is the
responsiblity of the mail gateway to properly masqerade header fields.  A
mail gateway that fails to do this is not configured correctly.  Gnus
should not be encouraged to work around such misbehaviour.
-- 
Rat <ratinox@peorth.gweep.net>    \ If Happy Fun Ball begins to smoke, get
Minion of Nathan - Nathan says Hi! \ away immediately. Seek shelter and cover
PGP Key: at a key server near you!  \ head.

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
| It is not the nature of DNS (or anything else) that the mail exchanger
| for domain.com must also accept mail for any other domain, such as
| random-sending-host.domain.com.

Actually, it should.  If it does not then steps should be taken to prevent
"random-sending-host.domain.com" from sending mail or to masquerade it.

| I didn't ask you to restate the requirement.  I asked where it was.
| RFC 2822 does not use the word "canonical" in connection with Sender.
| I haven't been able to find the requirement.

Sender is of type mailbox.  It must be possible to deliver mail to an
addr-spec for it to be a mailbox.  Canonical means that at least some
effort has been made to ensure the accuracy of the thing; it does not need
to be defined by RFC 2822 any more than the word "obsolete".

| It isn't a canonical address if it isn't an address at all.

But if login @ fqdn is not an address at all then something in the system
is not configured correctly.

[...]
| No, it isn't.  You're demanding that Sender identify not just the
| person who sent the message, but also the host it was sent from.  So I
| ask, what will break if your additional requirement isn't met?

And I answered.

[...]
| You still have Received fields.  Sender is unreliable anyway, since
| it's under the control of a possible malicious person.

So is From, To, and just about everything else.  Your point?

| They are necessary, but not sufficient, to make your requirement
| work.

Actually, proper DNS records are all that is required to make it work.
-- 
Rat <ratinox@peorth.gweep.net>    \ Do not taunt Happy Fun Ball.
Minion of Nathan - Nathan says Hi! \ 
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Graham Murray <graham@barnowl.demon.co.uk>  on Thu, 24 May 2001
>| 1) Dial-up hosts using dynamic IP. In this situation, the user is often
>|    given a mailbox of user@isp.com. [snip]
> 
> There must be an A record (and FQDN) for the IP address.  A properly
> configured dialup server will provide the FQDN associated with the IP
> address, and a properly configured dialup host will set its name to that
> FQDN for the duration of its connection.  Gnus generates Sender
> fields using that FQDN (or it should).

But I though you wanted Sender to have the user's canonical address.
In this case, the user's cacnonical address very clearly does not
include the system's current FQDN.

> The MX record associated with the A record will make it a
> deliverable mailbox.

You don't seem to understand DNS.  The dialup's name has an A record,
and may (but need not) have an MX record.  The A record certainly does
not have an MX record associated with it.

>| 2) Hosts on an internal LAN (probably using RFC 1918 addresses) behind a
>|    firewall. [snip]
> 
> Internally those fields should be as canonical as possible.  It is the
> responsiblity of the mail gateway to properly masqerade header fields.

Where does RFC 2822 make that requirement?


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
>| It is not the nature of DNS (or anything else) that the mail exchanger
>| for domain.com must also accept mail for any other domain, such as
>| random-sending-host.domain.com.
> 
> Actually, it should.  If it does not then steps should be taken to prevent
> "random-sending-host.domain.com" from sending mail or to masquerade it.

So you say.  I can't find any such statement in the RFCs.

> Canonical means that at least some effort has been made to ensure
> the accuracy of the thing;

That's completely wrong.  Go find a dictionary.

>| It isn't a canonical address if it isn't an address at all.
> 
> But if login @ fqdn is not an address at all then something in the system
> is not configured correctly.

I repeat, yet again: where is that requirement in the RFC?

>| You still have Received fields.  Sender is unreliable anyway, since
>| it's under the control of a possible malicious person.
> 
> So is From, To, and just about everything else.  Your point?

That those fields should not be used to track down problems, because
you have something better: Received.

> Actually, proper DNS records are all that is required to make it work.

False.  If domain.com's mail servers accept mail for domain.com but
not for random-sending-host.domain.com, then Sender (as you would have
it) would not be an actual mailbox, even with the DNS records in
place.


paul

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> Before this gets too long, here is a hint: system-name.  Think about
> it for a moment.

system-name helps to find an IP address, but it does not
(particularly) help with sending mail.

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Thu, 24 May
>   2001
>| I dare say that people don't complain because people can frob their
>| From header via user-mail-address.  They cannot, however, frob
>| their Sender header via user-mail-address.
> 
> And they should not.  Frobbing it directly would be a violation of
> the requirements of RFCs 2822 and 1034.

Please quote chapter and verse.

> [...]
>|   - Generate a Sender header using user-login-name, followed by
>|     "@", followed by system-name.
> 
> Which is as canonical as a program can get.  Sender is supposed to
> be canonical.  By "canonical" I mean "an attempt has been made to
> ensure that the mailbox is valid".  For a correctly configured
> system, and that includes the mail hubs and what-not, not just the
> local machine, `login at FQDN' is canonical.
> 
> I apparantly misunderstood something, because using
> user-mail-address for generating Sender would break that.  Therefore
> it should not be done.

RFC 2822 has the following to say about the From and Sender headers.

/----
| The originator fields indicate the mailbox(es) of the source of the
| message.  The "From:" field specifies the author(s) of the message,
| that is, the mailbox(es) of the person(s) or system(s) responsible for
| the writing of the message.  The "Sender:" field specifies the mailbox
| of the agent responsible for the actual transmission of the message.
\----

Note that this talks about both headers containing mailbox
specifications.  It does not at all suggest that the mailbox
specifications in the Sender header should in any way differ from the
mailbox specifications in the From header.  It does not say that the
part after `@' should be a FQDN for the From header mailbox specs, nor
does it say this for the Sender header mailbox specs.

I don't see why you require the part after `@' to be a FQDN for the
Sender header, but not for the From header.

And: I don't see why the part before the `@' should be the login name
of the user.  I don't see this for the From header, and I don't see
this for the Sender header.

Can you point out where the RFC makes such a distinction between the
>From and the Sender headers?

> [...]
>| I don't know what does RFC 2822 say.  I only know about RFC 822.
>| The local RFC server doesn't seem to know about this RFC.  Can you
>| help out?
> 
> RFCs 2821 and 2822 obsoleted RFCs 821 and 822 about a month ago.
> They clarify a lot of things, not the least of which is Sender, both
> by standard and defacto use.

Sadly, the examples are not quite clear.  However, the Message-ID
headers mention local.machine.example as an `after-@' part.  Since the
`after-@' part in a Message-ID should be a FQDN, I conclude that
local.machine.example is intended to be a FQDN.

The examples for the Sender header use machine.example, which is NOT a
FQDN.

How can you require FQDN if even the examples in the RFC don't have
it?

> [...]
>| I'm with you so far.  I presume that you have set user-mail-address
>| to "ratinox@peorth.gweep.net".  Then my proposal does what you
>| want.
> 
> No, I don't.  I have my system configured correctly.  It knows
> itself as peorth.gweep.net, so Emacs knows it as peorth.gweep.net,
> so Gnus knows it as peorth.gweep.net, and nothing needs to be
> kludged.

Well, if you have configured system-name to be correct for your From
address, then you have the special case where the existing Gnus
algorithm and my proposed algorithm return the same result.

But this special case is not true for all people.

> You seem to believe that because only one person is involved in
> originating a message and submitting it that the one person has only
> one identity.

I don't think that.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> Which is as canonical as a program can get.  Sender is supposed to
> be canonical.  By "canonical" I mean "an attempt has been made to
> ensure that the mailbox is valid".

The word canonical occurs exactly once in RFC 2822.  This occurrence
is not in the context you are talking about.

The RFC does not appear to distinguish between mailbox specifications
in the Sender header and in the From header, w.r.t. validity.  Hence,
if you require FQDN in the Sender header because of validity, then you
also require FQDN in the From header because of validity.  It is very
easy to show that there are thousands of people where the FQDN does
not lead to a very valid mailbox spec.  Hence, the FQDN requirement is
not useful in practice.

And it is not present in the RFC 2822, as far as I can see!

Hm.  Now I looked in RFC 2821, and there I can't see the FQDN
requirement, either.  So where do you get it from?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> The correct question is, what breaks the Sender field does not
> canonically identify the sender?  The answer is, if the Sender field
> does not contain the canonical mailbox of the sender and there is a
> problem, perhaps with that machine, I may be unable to contact the
> sender.

If the Sender field contains the FQDN of the host, and the machine is
turned off, you're screwed.  So maybe it is better to give a domain
part in the address which identifies a host which is sure to be on?
Maybe this is the case in my situation?  How can you know?  How can
Gnus know?

Why do you want to forbid me from putting there an address which I
know to work better than user@FQDN?

> You see, Sender is for human consumption only.  It exists at least
> partially so that humans, like me, can track down problems at their
> source.  If the Sender field does not include the FQDN of the
> sending host it makes finding and solving problems that much more
> difficult.

Reall?  What is the FQDN needed for?  The address should specify a
mailbox, and knowing about host names is not necessary for specifying
mailboxes.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> But if login @ fqdn is not an address at all then something in the
> system is not configured correctly.

Really?  Why should a login have anything to do with a mail address?
A login is only something you type at a login prompt.  Maybe for
security reasons you don't want to disclose login names to the outside
world?  I don't see any reason whatsoever for a login name to have
something to do with a mail address.

Why should a FQDN be (the after-@ part of) a mail address?  Requiring
for the host to be able to receive mail is akin to requiring every
host on the Internet to provide a Web server.  Sure, I have this host,
and Netscape is running on it.  But of course the host does not have
to be reachable from other Netscape programs on other hosts (ie, no
httpd required).  In the same way, I have this host and it can send
mail.  But of course the host is not required to be able to receive
mail.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On Thu, 24 May 2001, Stainless Steel Rat wrote:

> * Graham Murray <graham@barnowl.demon.co.uk>  on Thu, 24 May 2001
>| 1) Dial-up hosts using dynamic IP. In this situation, the user is
>|    often given a mailbox of user@isp.com. [snip]
> 
> There must be an A record (and FQDN) for the IP address.  A properly
> configured dialup server will provide the FQDN associated with the
> IP address,

Some people have to work with a dialup server which is not properly
configured.  I want to make it easy for Gnus users to use such dialup
servers.  You want to make it hard for Gnus users to use them.

These guys' lives are so miserable already, why make it harder?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Barry Fishman]

You posted to Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> But if login @ fqdn is not an address at all then something in the system
> is not configured correctly.

My login@fqdn is "barry@ecube.local".  The closest thing to a local
mail address is my pop server address "barry_fishman@att.net".
I connect to the internet using a PPP connection PAP authentication
with a variety of addresses.  At the moment I'm at "12.78.17.131", but
this varies each call and is probably bogus anyway.  I send mail by
direct smtp connection to mailhost.att.net.

So what would be my properly configured sender address?  How could I
fix my configuration to have one?  Please be specific.

My gut feeling is that it is up to the smtp server on mailhost.att.net
to provide it.  In this case I need to prevent gnus from creating one
even though my "From:"  address does not match my login@fqdn address.
Presently I am just doing a:

        ;; Fix message send address (I don't have a real hostname
        (defun message-make-sender ()
          "Return own mail address as sender"
          (message-make-address))

Barry Fishman

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
| But I though you wanted Sender to have the user's canonical address.
| In this case, the user's cacnonical address very clearly does not
| include the system's current FQDN.

You are assuming that sender == originator.  Don't do that.

| You don't seem to understand DNS.  The dialup's name has an A record,
| and may (but need not) have an MX record.  The A record certainly does
| not have an MX record associated with it.

It should have an MX record associated with it.  All A records should have
associated MX records.

| > Internally those fields should be as canonical as possible.  It is the
| > responsiblity of the mail gateway to properly masqerade header fields.
| Where does RFC 2822 make that requirement?

Definition of "mailbox".
-- 
Rat <ratinox@peorth.gweep.net>    \ Caution: Happy Fun Ball may suddenly
Minion of Nathan - Nathan says Hi! \ accelerate to dangerous speeds.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
| These guys' lives are so miserable already, why make it harder?

Then set "system-name" to something useful and be done with it.  That is
about as easy as it gets.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
| False.  If domain.com's mail servers accept mail for domain.com but
| not for random-sending-host.domain.com, then Sender (as you would have
| it) would not be an actual mailbox, even with the DNS records in
| place.

If domain.com's mail servers do not accept mail for the domain.com domain,
they are so horribly misconfigured that anyone stuck with them should jump
ship for AOL.
-- 
Rat <ratinox@peorth.gweep.net>    \ Warning: pregnant women, the elderly, and
Minion of Nathan - Nathan says Hi! \ children under 10 should avoid prolonged
PGP Key: at a key server near you!  \ exposure to Happy Fun Ball.

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
| Really?  Why should a login have anything to do with a mail address?

Because a mailbox is defined as type "addr-spec", which is"

3.4.1. Addr-spec specification

   An addr-spec is a specific Internet identifier that contains a
   locally interpreted string followed by the at-sign character ("@",
   ASCII value 64) followed by an Internet domain.  The locally
   interpreted string is either a quoted-string or a dot-atom.  If the
   string can be represented as a dot-atom (that is, it contains no
   characters other than atext characters or "." surrounded by atext

The defacto standard for the locally interpreted string is one's login
name, has been that way for more than 30 years.


| Why should a FQDN be (the after-@ part of) a mail address?

Because a domain-literal is exactly that.

Together, "login @ fqdn" absolutely identifies the sender given a
reasonably configured system.
-- 
Rat <ratinox@peorth.gweep.net>    \ Ingredients of Happy Fun Ball include an
Minion of Nathan - Nathan says Hi! \ unknown glowing substance which fell to
PGP Key: at a key server near you!  \ Earth, presumably from outer space.

Re: Sender header?

[Stainless Steel Rat]

* Barry Fishman <barry_fishman@acm.org>  on Thu, 24 May 2001
| My login@fqdn is "barry@ecube.local".

*snrk*.

That is not a legal fqdn, or did someone create the "local" TLD when I
blinked? :)

| The closest thing to a local mail address is my pop server address
| "barry_fishman@att.net".  I connect to the internet using a PPP
| connection PAP authentication with a variety of addresses.  At the moment
| I'm at "12.78.17.131", but this varies each call and is probably bogus
| anyway.  I send mail by direct smtp connection to mailhost.att.net.

The IP is not bogus.  The fact that AT&T has no A record for it is bogus.
That is probably a real pain when you hit an SSH server that refuses your
connection because it cannot do a reverse IP lookup.

| So what would be my properly configured sender address?  How could I
| fix my configuration to have one?  Please be specific.

There should be an A record in att.net's DNS associated with that IP, and
it should include that name in the DHCP packet it gives you when you make
your dialup connection.  Your host should set its name to the name
specified in that packet for the duration of the session.  What you do when
you are not connected is up to you.

| My gut feeling is that it is up to the smtp server on mailhost.att.net
| to provide it.  In this case I need to prevent gnus from creating one
| even though my "From:"  address does not match my login@fqdn address.

Given that att.net won't let you do things right, dumping it on them is
probably not a bad thing to do.  They've screwed it up so let them deal
with it.
-- 
Rat <ratinox@peorth.gweep.net>    \ If Happy Fun Ball begins to smoke, get
Minion of Nathan - Nathan says Hi! \ away immediately. Seek shelter and cover
PGP Key: at a key server near you!  \ head.

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
| If the Sender field contains the FQDN of the host, and the machine is
| turned off, you're screwed.

*sigh*
Not if your MX records are set up correctly.

| So maybe it is better to give a domain part in the address which
| identifies a host which is sure to be on?  Maybe this is the case in my
| situation?  How can you know?  How can Gnus know?

| Why do you want to forbid me from putting there an address which I
| know to work better than user@FQDN?

Because right now, for example, ratinox@newsguy.com could be who I am when
I write this message, but it is not who I am when I send it.  It would be
incorrect for me to put ratinox@newsguy.com in the Sender header no matter
how much "better" (read convenient) it is.


| Reall?  What is the FQDN needed for?  The address should specify a
| mailbox, and knowing about host names is not necessary for specifying
| mailboxes.

The domain part of an RFC 2822 mailbox must be a fully qualified domain
name.  If you are not doing RFC 2822 mail then you can do whatever you
want, but don't expect it to work correctly anywhere else.
-- 
Rat <ratinox@peorth.gweep.net>    \ If Happy Fun Ball begins to smoke, get
Minion of Nathan - Nathan says Hi! \ away immediately. Seek shelter and cover
PGP Key: at a key server near you!  \ head.

Re: Sender header?

[Karl Kleinpaste]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> What do you think?

(add-to-list 'message-syntax-checks '(sender . disabled))

Re: Sender header?

[Bjørn Mork]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:

> But if login @ fqdn is not an address at all then something in the system
> is not configured correctly.

Where do you find this requirement?

Blindly creating sender fields based on this assumption violates
RFC2822. If present, the sender field must contain a single
mailbox. "login @ fqdn" does not qualify in many cases, and there is
nothing requiring it should (even if there were: Would it justify
violating RFC2822?)


   If the originator of the message can be indicated
   by a single mailbox and the author and transmitter are identical, the
   "Sender:" field SHOULD NOT be used.

I do see that this text can be interpreted in different ways, but
let's interpret it in a way that makes sense. When are the author and
transmitter identical? Most of us write mail wearing different hats and
using different machines to transmit it, maybe even different logins
on the same machine. But the important difference is always which hat
we are wearing. In my opionion, the author and transmitter are always
identical when they represent the same person. If I write a message
with the bmork@dod.no hat on, then that's the hat I am using when
transmitting it too. Where I am transmitting it from doesn't change
this. 

With this interpretation, including a sender field would be illegal in
all cases where a single author transmits his own mail. A sender field
should only be added when transmitting mail on behalf of others, or
when the mail has multiple authors. 



Bjørn

Re: Sender header?

[Harry Putnam]

Barry Fishman <barry_fishman@acm.org> writes:

> So what would be my properly configured sender address?  How could I
> fix my configuration to have one?  Please be specific.
> 
> My gut feeling is that it is up to the smtp server on mailhost.att.net
> to provide it.  In this case I need to prevent gnus from creating one
> even though my "From:"  address does not match my login@fqdn address.
> Presently I am just doing a:
> 
>         ;; Fix message send address (I don't have a real hostname
>         (defun message-make-sender ()
>           "Return own mail address as sender"
>           (message-make-address))


I got hollered at on comp.editors for having a made up rhs of message
id which is somewhat related to this discussion.

Describing another situation where it isn't really clear what I
should have for a From or sender address.

My situation isn't quite as complex as Barrys but is probably fairly
typical of home users in US.   I have a dsl connection on static ip.
Right now my router (not a machine with an OS running on it) has the
IP address assigned to me.  My computer has an internal class c 192
address as do the others on my home network.  All are capable of
sending mail, all are capable of recieving mail, but not at the static
ip address,  But at my IP's mail machine address.

The router is NAT enabled and knows who sent what.

>From what I'm seeing on this thread, the router which has my static IP
address is what the sender would be derived from, however it cannot
recieve mail for that address, only for either my IP domain or newsguy  

To complicate things a little further.  I don't actually use my IPs
address but have a newsguy account I do most of my mail with.  That
from address uses newguy.com domain but is not my machine but the
address of a machine several hundred miles from me.

I've been thoroughly confused about these matters for quite some time.

I'd be interested to hear Rats or others views of what I should be doing
regarding From and sender.  (Message-ID would be a bonus).

Currently I've told gnus to stick my IP smtp server address in
Message-ID.  And disabled sender generation.

(defun message-make-fqdn ()
  "My  hacked message-id."
  "ptw.com")

Re: Sender header?

[Bjørn Mork]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:

> It should have an MX record associated with it.  All A records should have
> associated MX records.

No. All mail names should have MX records. You are assuming that all
names with an A record are mail hosts. That's not a requirement, and
certainly not true.

RFC 1912 recommends adding MX records even for hosts which are not
supposed to send or receive mail, but I don't think that justifies
ignoring those who don't.

> | > Internally those fields should be as canonical as possible.  It is the
> | > responsiblity of the mail gateway to properly masqerade header fields.
> | Where does RFC 2822 make that requirement?
> 
> Definition of "mailbox".

Yeah right. You're reading standards like Microsoft.


Bjørn

Re: Sender header?

[Bjørn Mork]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
> | Really?  Why should a login have anything to do with a mail address?
> 
> Because a mailbox is defined as type "addr-spec", which is"
> 
> 3.4.1. Addr-spec specification
> 
>    An addr-spec is a specific Internet identifier that contains a
>    locally interpreted string followed by the at-sign character ("@",
>    ASCII value 64) followed by an Internet domain.  The locally
>    interpreted string is either a quoted-string or a dot-atom.  If the
>    string can be represented as a dot-atom (that is, it contains no
>    characters other than atext characters or "." surrounded by atext
> 
> The defacto standard for the locally interpreted string is one's login
> name, has been that way for more than 30 years.

Sure. But by requiring it to always be true is adding unnecessary
restrictions. The RFC doesn't, so why should Gnus?

> | Why should a FQDN be (the after-@ part of) a mail address?
> 
> Because a domain-literal is exactly that.
> 
> Together, "login @ fqdn" absolutely identifies the sender given a
> reasonably configured system.

For some weird definition of "reasonably configured". 

RFC2822 requires a valid mailbox in the sender field. Can you please
point to the specification that guarantees that "login @ fqdn" is a
valid mailbox? 


Bjørn
-- 
I mean, How can you be so primitive?  

Re: Sender header?

[Russ Allbery]

Barry Fishman <barry_fishman@acm.org> writes:

> Presently I am just doing a:

>         ;; Fix message send address (I don't have a real hostname
>         (defun message-make-sender ()
>           "Return own mail address as sender"
>           (message-make-address))

(setq message-syntax-checks '((sender . disabled)))

has worked great for years.  Hopefully it hasn't been broken in later
versions of Gnus.  It should be the default.

I've heard all of the same arguments that Rat's making many, many times.
I don't think they're supported by the RFCs (and the discussion I've read
on DRUMS backs that up), and whether they are or not, the Sender header is
broken beyond recovery and shouldn't be encouraged.  Like automatically
converting eight spaces to a tab or the Lines header on Usenet, it's one
of those ideas that had some marginal utility originally but never
actually worked very well in practice and is now better forgotten.

Sender is useful if you are using in the From field an address that has
*nothing* to do with the entity that actually sent the mail, in which case
it's sometimes useful to put some valid mailbox in there.  Anything that
automatically generates the header is just going to cause far more
problems than it solves.  Tracing is what Received headers are for.

The only real widespread use of Sender these days is in mailing lists
(particularly ones based on Majordomo); hopefully List-ID and relatives
will replace that last use with a better-designed system.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Russ Allbery]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:

> It should have an MX record associated with it.  All A records should
> have associated MX records.

There's nothing in any RFC that says that.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Stainless Steel Rat]

* "Bjørn Mork" <bmork@dod.no>  on Thu, 24 May 2001
| RFC 1912 recommends adding MX records even for hosts which are not
| supposed to send or receive mail, but I don't think that justifies
| ignoring those who don't.

In what way is this different from "all A records should have associated MX
records"?
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Stainless Steel Rat]

* Russ Allbery <rra@stanford.edu>  on Thu, 24 May 2001
| > It should have an MX record associated with it.  All A records should
| > have associated MX records.

| There's nothing in any RFC that says that.

RFC 1912 does.
-- 
Rat <ratinox@peorth.gweep.net>    \ Caution: Happy Fun Ball may suddenly
Minion of Nathan - Nathan says Hi! \ accelerate to dangerous speeds.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* "Bjørn Mork" <bmork@dod.no>  on Thu, 24 May 2001
| RFC2822 requires a valid mailbox in the sender field. Can you please
| point to the specification that guarantees that "login @ fqdn" is a
| valid mailbox?

Already did.  Four or five times, now.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Stainless Steel Rat]

* Harry Putnam <reader@newsguy.com>  on Thu, 24 May 2001
| From what I'm seeing on this thread, the router which has my static IP
| address is what the sender would be derived from, however it cannot
| recieve mail for that address, only for either my IP domain or newsguy

Somewhere in this you should have your MTA (or plural) masqerade outgoing
headers so that anyone on the outside of your firewall sees nothing that
points to anything inside your firewall.  This is easiest to manage if you
have one mail gate and everything behind the firewall uses it as the smart
host.

| I'd be interested to hear Rats or others views of what I should be doing
| regarding From and sender.  (Message-ID would be a bonus).

>From should be reader@newsguy.com.  Sender should start out being something
appropriate for the sending machine inside your firewall, and be masqeraded
(and possibly rewritten entirely) by your mail gate with your One True FQDN
associated with your single IP address.

Message-ID strings can be simpler or harder, depending on how militant one
is.  The one absolute requirement is that any given message's Message-ID be
unique.  Usually the easiest way to accomplish this is to hash/mash login,
FQDN and time stamp.  You might be better off telling Gnus not to generate
Message-ID strings either, and let your mail gate handle it.
-- 
Rat <ratinox@peorth.gweep.net>    \ If Happy Fun Ball begins to smoke, get
Minion of Nathan - Nathan says Hi! \ away immediately. Seek shelter and cover
PGP Key: at a key server near you!  \ head.

Re: Sender header?

[Stainless Steel Rat]

* Russ Allbery <rra@stanford.edu>  on Thu, 24 May 2001
| (setq message-syntax-checks '((sender . disabled)))

| has worked great for years.  Hopefully it hasn't been broken in later
| versions of Gnus.  It should be the default.

FWIW, I'm in agreement with Russ on this, and to <mumble> with what the
GNKSA has to say about it.
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Harry Putnam]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:

> Somewhere in this you should have your MTA (or plural) masqerade outgoing
> headers so that anyone on the outside of your firewall sees nothing that
> points to anything inside your firewall.  This is easiest to manage if you
> have one mail gate and everything behind the firewall uses it as the smart
> host.

Ahh, that tallies with what I've been doing..

> | I'd be interested to hear Rats or others views of what I should be doing
> | regarding From and sender.  (Message-ID would be a bonus).

[...]

> Message-ID strings can be simpler or harder, depending on how militant one
> is.  The one absolute requirement is that any given message's Message-ID be
> unique.  Usually the easiest way to accomplish this is to hash/mash login,
> FQDN and time stamp.  You might be better off telling Gnus not to generate
> Message-ID strings either, and let your mail gate handle it.

Possibly a bit OT but on the Message-ID part, how big of a concern is
it that the rhs be a real FQDN  (or is it mailbox?)  I thought I
recalled this being discussed heavily here at one point and some
consensus being that it wasn't to important.  And in keeping with your
comments, the real item is uniqueness.

I get a little confused about the uniqness issue.  Not really seeing
how a mail machine name is more unique than some homeboy thing I
hacked in there (rhs).

Re: Sender header?

[Russ Allbery]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Russ Allbery <rra@stanford.edu>  on Thu, 24 May 2001

> | > It should have an MX record associated with it.  All A records should
> | > have associated MX records.

> | There's nothing in any RFC that says that.

> RFC 1912 does.

That depends on what you mean by should.  RFC 1912 definitely doesn't say
SHOULD.  It says that it's a good idea (and goes on to explain that it
thinks it's a good idea for incredibly minor and unimportant performance
reasons, so personally I feel free to disagree with it -- the added
minor complexity is more of a drawback to me than the incredibly minor
performance difference is a win).

In practice, there's no important reason for a host that receives e-mail
for itself and that doesn't have a backup mail server to have an MX
record.  An A record is entirely sufficient.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Russ Allbery]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Russ Allbery <rra@stanford.edu>  on Thu, 24 May 2001
> | (setq message-syntax-checks '((sender . disabled)))

> | has worked great for years.  Hopefully it hasn't been broken in later
> | versions of Gnus.  It should be the default.

> FWIW, I'm in agreement with Russ on this, and to <mumble> with what the
> GNKSA has to say about it.

*blink blink*

For some reason, I thought we were vehemently disagreeing.  Cool!  I love
it when that happens.  :)

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Christoph Conrad]

Hello Paul,

you wrote:

    > I see: Sender: owner-ding@hpc.uh.edu But that's obviously not
    > Gnus's doing.

Em, i am surprised, really. I had a look in the original downloaded
article on the disk,

~/News/agent/nntp/quimby.gnus.org/gnus/ding/32560 

I looked in it with Emacs and with less, no "Sender:" in header. I am
using gnus agent. It could be only Gnus who filtered it out when
downloading.

Oort Gnus v0.04
GNU Emacs 21.0.103.1 (i586-pc-linux-gnu, X toolkit, Xaw3d scroll bars)
 of 2001-05-16 on mutzel

Mhmmm, what's going on here?

Best regards,
cu, -cc-
-- 
=> GNU Emacs Webring @ <http://www.gnusoftware.com/WebRing/> <=
Look Ma, this man can twist his fingers as if they were made of rubber,
isn't that amazing? -- Not really, he's been using emacs for years...!

Re: Sender header?

[Kai Großjohann]

On 24 May 2001, Karl Kleinpaste wrote:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>> What do you think?
> 
> (add-to-list 'message-syntax-checks '(sender . disabled))

This is what I currently use.  However, it does not automatically add
a Sender header if I have manually modified the From header.  If
somebody else uses my Gnus for sending a message, and manually frobs
the From header, I want Gnus to insert a Sender header with
user-mail-address in it.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Stainless Steel Rat wrote:

> * Russ Allbery <rra@stanford.edu>  on Thu, 24 May 2001
>| (setq message-syntax-checks '((sender . disabled)))
> 
>| has worked great for years.  Hopefully it hasn't been broken in
>| later versions of Gnus.  It should be the default.
> 
> FWIW, I'm in agreement with Russ on this, and to <mumble> with what
> the GNKSA has to say about it.

Either you have changed your mind, or I fundamentally don't grok what
you're saying.  I thought you were saying that Gnus should create a
Sender header (with whatever contents), but the above prevents Gnus
from ever creating one.  What's the story?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 24 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Thu, 24 May
>   2001
>| Really?  Why should a login have anything to do with a mail
>| address?
> 
> Because a mailbox is defined as type "addr-spec", which is"
> 
> 3.4.1. Addr-spec specification
> 
>    An addr-spec is a specific Internet identifier that contains a
>    locally interpreted string followed by the at-sign character
>    ("@", ASCII value 64) followed by an Internet domain.  The
>    locally interpreted string is either a quoted-string or a
>    dot-atom.  If the string can be represented as a dot-atom (that
>    is, it contains no characters other than atext characters or "."
>    surrounded by atext
> 
> The defacto standard for the locally interpreted string is one's
> login name, has been that way for more than 30 years.

The defacto standard for the locally interpreted string has changed.
These days, people want to hide the login names for security purposes.

>| Why should a FQDN be (the after-@ part of) a mail address?
> 
> Because a domain-literal is exactly that.

A domain-literal need not be a FQDN, it can also be a domain name.
For example, uni-dortmund.de is a possible domain-literal, even though
there is no such host.

> Together, "login @ fqdn" absolutely identifies the sender given a
> reasonably configured system.

The way I see it, your requirements for `reasonably configured' are
pretty steep, and what you want makes life really hard for people
working on systems which don't match your idea of `reasonably
configured'.  This is pointless.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Bjørn Mork wrote:

> RFC2822 requires a valid mailbox in the sender field. Can you please
> point to the specification that guarantees that "login @ fqdn" is a
> valid mailbox? 

Statements from Rat I remember are `on a reasonably configured system'
and `it's been this way for 30 years'.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 24 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Thu, 24 May
>   2001
>| These guys' lives are so miserable already, why make it harder?
> 
> Then set "system-name" to something useful and be done with it.
> That is about as easy as it gets.

It's not so easy, for system-name is used for two different things:
(a) for the mail address, and (b) for the Message-ID header.  The
requirements for these two are different.  It's not always possible to
have one value fulfill both requirements.

Now you can claim that a new variable should be introduced for the
Message-ID header, and I claim that a different variable should be
used for the mail address.  The end result is the same.  Except that
the variable for the mail address already exists -- mail-host-address
could be used.

In Germany, there are a number of dialup providers which provide
call-by-call accounting.  You call them, they give you an IP address,
you can do IP.  These providers do not provide a mailbox for the
people dialing them.  However, there are other freemail providers
which one can use.  Your requirements would prevent me from using
their convenient and cheap service.  I think it is not the job of a
standard to prevent me from using a cheap service.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Stainless Steel Rat wrote:

> * "Bjørn Mork" <bmork@dod.no>  on Thu, 24 May 2001
>| RFC 1912 recommends adding MX records even for hosts which are not
>| supposed to send or receive mail, but I don't think that justifies
>| ignoring those who don't.
> 
> In what way is this different from "all A records should have
> associated MX records"?

It's the difference between a recommendation and a requirement.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Per Abrahamsen]

Russ Allbery <rra@stanford.edu> writes:

> (setq message-syntax-checks '((sender . disabled)))
> 
> has worked great for years.  Hopefully it hasn't been broken in later
> versions of Gnus.  It should be the default.

I agree, except that it breaks cancels when using
'gnus-posting-styles'.  Nonetheless, I'd actually prefer disabling
sender, and let someone who cares about 'gnus-posting-styles' fix
that.

Re: Sender header?

[Karl Kleinpaste]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> This is what I currently use.  However, it does not automatically add
> a Sender header if I have manually modified the From header.

That was merely my (too subtle?) expression of an attitude about the
actual, real-life, practical utility of Sender.

I crawled through my archives for a little while last night, looking
for when I last generated Sender with any regularity.  It seems to
have been early 1995 or so.  And evidently I've generated on the order
of 25,000 messages with Gnus since that time.  Never once in 6 full
years of busy net.activity have I had actual trouble befall me from
having left Sender out.  Thus, I conclude that, RFCs be damned
regardless of their conflicting instructions, the practical utility of
Sender is on a par with the practical utility of Resent-From, i.e.,
none at all.

To and Cc have practical utility.  Subject and Xref have practical
utility.  Keywords has enough practical utility to me personally that
I auto-generate it where it doesn't exist.  But Sender doesn't do
anything for anybody, as a practical matter.  We have argued over the
semantics of Sender for years, and evidently even now have never
gotten it quite right (else the argument wouldn't have resumed now),
and yet I've found that just *not generating it* is the simplest
practical solution with no evident negative effects.

You folks go ahead and argue once again over Sender.  I'll continue to
keep Sender disabled.  And I won't have any problems, regardless of
the conclusions reached by, as well as the code modifications that
result from, the resumption of the Sender argument.

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> Since the `after-@' part in a Message-ID should be a FQDN, I
> conclude that local.machine.example is intended to be a FQDN.
> 
> The examples for the Sender header use machine.example, which is NOT a
> FQDN.

>From a strict DNS point of view, both of those names are not FQ, but
both would be if you added a "." at the end.  ("gnus.org." is a FQDN;
this has nothing to do with whether it has an address.)  More relevant
to the current discussion, though, is that there is no suggestion that
Sender should include the *local* FQDN.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Thu, 24 May 2001
>| If the Sender field contains the FQDN of the host, and the machine is
>| turned off, you're screwed.
> 
> *sigh*
> Not if your MX records are set up correctly.

Where do the RFCs support your notion of correctness WRT MX records?

>| Why do you want to forbid me from putting there an address which I
>| know to work better than user@FQDN?
> 
> Because right now, for example, ratinox@newsguy.com could be who I am when
> I write this message, but it is not who I am when I send it.

You are always the same person, and all your mailboxes identify you
equally well all the time.  The fact that you're using only one of
them at any given moment changes nothing.

> The domain part of an RFC 2822 mailbox must be a fully qualified domain
> name.

I see no such requirement.  Using an unqualified name like
"prj@multivac" will indeed break things, of course.  But you're also
inventing the requirement for a *particular* FQDN - the local one.
This requirement is not in RFC 2822.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
>| False.  If domain.com's mail servers accept mail for domain.com but
>| not for random-sending-host.domain.com, then Sender (as you would have
>| it) would not be an actual mailbox, even with the DNS records in
>| place.
> 
> If domain.com's mail servers do not accept mail for the domain.com domain,
> they are so horribly misconfigured that anyone stuck with them should jump
> ship for AOL.

I agree.  But that has nothing to do with what I said.


paul

Re: Sender header?

[Paul Jarc]

Karl Kleinpaste <karl@charcoal.com> writes:
> We have argued over the semantics of Sender for years, and evidently
> even now have never gotten it quite right (else the argument
> wouldn't have resumed now),

Gnus's current behavior is correct for news.  But for mail, Sender is
a completely different entity which just happens to have the same name.


paul

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> If somebody else uses my Gnus for sending a message, and manually
> frobs the From header, I want Gnus to insert a Sender header with
> user-mail-address in it.

Gnus can't distinguish between this case and the case where you
manually enter a different From address that still refers to you; the
case could be made that Gnus should never add Sender automatically for
mail messages (except when there are multiple From addresses - in that
case, Gnus should definitely add Sender: user-mail-address).  But it
wouldn't do any harm to add Sender in these ambiguous cases as well.
In all cases, if Sender is added automatically, its contents should be
user-mail-address, but this applies only for mail, not news.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Thu, 24 May 2001
>| Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
>|> A properly configured dialup server will provide the FQDN
>|> associated with the IP address, and a properly configured dialup
>|> host will set its name to that FQDN for the duration of its
>|> connection.  Gnus generates Sender fields using that FQDN (or it
>|> should).
>| 
>| But I though you wanted Sender to have the user's canonical address.
>| In this case, the user's cacnonical address very clearly does not
>| include the system's current FQDN.
> 
> You are assuming that sender == originator.  Don't do that.

I am not assuming that sender==author, if that's what you mean.  A
user identified by login@dialup.isp.com does not have
login@dialup.isp.com as their primary email address, nor is that
necessarily a valid address at all, thus it should not be used for
Sender.

>| > It is the responsiblity of the mail gateway to properly masqerade
>| > header fields.
>| Where does RFC 2822 make that requirement?
> 
> Definition of "mailbox".

I don't see it.  Chapter and verse, please.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Barry Fishman <barry_fishman@acm.org>  on Thu, 24 May 2001
>| My login@fqdn is "barry@ecube.local".
> 
> *snrk*.
> 
> That is not a legal fqdn, or did someone create the "local" TLD when I
> blinked? :)

Creating a TLD is easy.  Getting ICANN's root servers to serve it is
the hard part. :)

>| The closest thing to a local mail address is my pop server address
>| "barry_fishman@att.net".  I connect to the internet using a PPP
>| connection PAP authentication with a variety of addresses.  At the moment
>| I'm at "12.78.17.131", but this varies each call and is probably bogus
>| anyway.  I send mail by direct smtp connection to mailhost.att.net.
> 
> The IP is not bogus.  The fact that AT&T has no A record for it is bogus.

We don't know whether there's an A record with that address, because
we don't know what name to look up to find out; we only know that
there's no PTR record for the name 131.17.78.12.in-addr.arpa.  Please
stop pretending to know about DNS.

> That is probably a real pain when you hit an SSH server that refuses your
> connection because it cannot do a reverse IP lookup.

Maybe his provider gives him some way around that.  Since we don't
know, we can't say that this setup is bogus.

> There should be an A record in att.net's DNS associated with that IP, and
> it should include that name in the DHCP packet it gives you when you make
> your dialup connection.

FAYK, this is already the case.

> Your host should set its name to the name specified in that packet
> for the duration of the session.

I don't agree.  The host should use the hostname provided by DHCP for
dealings with that same network, but not necessarily globally, nor on
other networks the host may be on.


paul

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Paul Jarc wrote:

> In all cases, if Sender is added automatically, its contents should
> be user-mail-address,

That's what I've been saying all along...

> but this applies only for mail, not news.

... except that I didn't know about that distinction.  What's the
story for news?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> Possibly a bit OT but on the Message-ID part, how big of a concern is
> it that the rhs be a real FQDN  (or is it mailbox?)

It should uniquely identify the host that generated the Message-ID -
but it need not be the host's FQDN.  The point is that there should be
no chance of another host using that same identifier by accident.  If
you control all of domain.com, and you want the host foo.domain.com to
use msgid.foo.domain.com in its Message-IDs, feel free to do that -
just make sure other hosts in domain.com respect that convention.

> I get a little confused about the uniqness issue.  Not really seeing
> how a mail machine name is more unique than some homeboy thing I
> hacked in there (rhs).

If you don't control your DNS domain, you can't ensure that your
convention will be respected.


paul

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> I'd be interested to hear Rats or others views of what I should be doing
> regarding From and sender.

As long as you're sending just your own messages, you don't need
Sender at all.  (You can read RFC 2822, 3.6.2 for yourself to confirm
that the require Rat claims is there, isn't.)  It's all a lot simpler
than this thread makes it out to be.

> Currently I've told gnus to stick my IP smtp server address in
> Message-ID.

Don't do that unless you control the SMTP server, and know that it
won't generate Message-IDs using that same RHS.  See my other message
for more details.


paul

Re: Sender header?

[Paul Jarc]

Russ Allbery <rra@stanford.edu> writes:
> (setq message-syntax-checks '((sender . disabled)))
> 
> has worked great for years.

I want to disable Sender for mail, but not for news.  Can that be
easily done?


paul

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> What's the story for news?

According to RFC 1036, Sender is for news more or less what Rat says
it should be for mail - it's used for tracking down where a message
entered the network.  So user-login-name@system-name is the right
thing in that case.  Sender for mail is something completely
different, and just happens to have the same name.


paul

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Paul Jarc wrote:

> According to RFC 1036, Sender is for news more or less what Rat says
> it should be for mail - it's used for tracking down where a message
> entered the network.  So user-login-name@system-name is the right
> thing in that case.  Sender for mail is something completely
> different, and just happens to have the same name.

RFC 1036 has this example:

/----
|     For example, if John Smith is visiting CCA and wishes to post a
|     message to the network, using friend Sarah Jones' account, the
|     message might read:
| 
|               From: smith@ucbvax.Berkeley.EDU (John Smith)
|               Sender: jones@cca.COM (Sarah Jones)
\----

cca.COM doesn't look like the system name of any host.  So the
requirement for the rhs to be the system name would contradict this
example.

Also, I'm having difficulty finding support in RFC 1036 for Sender
addresses to be any different than other mail addresses (in
particular, the mail addresses in the From header).  There is a word
`verified' in there, but the RFC does not say what it means.  Also,
the intent of the whole thing seems to be that it is possible to send
mail to the address in the Sender header -- and if `user@host.dom.ain'
is not the right place to send email for mails, then it's not the
right place to send emails for news, either.

Or am I mistaken?  Maybe son or grandson have better information, but
I keep forgetting where to find them.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> On 25 May 2001, Paul Jarc wrote:
>> According to RFC 1036, Sender is for news more or less what Rat says
>> it should be for mail - it's used for tracking down where a message
>> entered the network.
> 
> RFC 1036 has this example:

Yes, but it also has this one:
# If a gateway program enters a mail message into the network at host
# unix.SRI.COM, the lines might read:
#
#           From: John.Doe@A.CS.CMU.EDU
#           Sender: network@unix.SRI.COM

I don't think it's suggesting that network@unix.sri.com issupposed to
be a valid email address.

> cca.COM doesn't look like the system name of any host.

It looks like it probably isn't, but it could be.  A host could even
have just "z." as its FQDN, given the appropriate DNS records.

> So the requirement for the rhs to be the system name would
> contradict this example.

I think user-login-name@system-name would work best for news, but
nothing specific is required.  I think user-mail-address would work
too, though not as well, based on the explanatory text in 2.2.2.  It's
all pretty vague.

> Also, the intent of the whole thing seems to be that it is possible
> to send mail to the address in the Sender header

I don't see that.

> Maybe son or grandson have better information, but I keep forgetting
> where to find them.

I don't think I've ever seen them, so my information may be well out
of date.


paul

Re: Sender header?

[Harry Putnam]

prj@po.cwru.edu (Paul Jarc) writes:

> 
> > Currently I've told gnus to stick my IP smtp server address in
> > Message-ID.
> 
> Don't do that unless you control the SMTP server, and know that it
> won't generate Message-IDs using that same RHS.  See my other message
> for more details.

Now I'm really confused..
Wait... shouldn't that smtp server AlWAYS use the same rhs?

Pauls' other post in part:

> It should uniquely identify the host that generated the Message-ID -
> but it need not be the host's FQDN.  The point is that there should be
> no chance of another host using that same identifier by accident.  If
> you control all of domain.com, and you want the host foo.domain.com to
> use msgid.foo.domain.com in its Message-IDs, feel free to do that -
> just make sure other hosts in domain.com respect that convention.

The machine has a name like (/etc/hosts entry):
reader.local.lan        192.168.xx.xxx   reader
There must be several (100/1000 ?) that could possibly have that FQDN.
How is this uniqueness obtained?

If I allowed gnus to generate the Message-ID (rhs) by its own devices,
then it would be:

Message-ID: <m1hey9z673.fsf@reader.local.lan>

Instead of:
Message-ID: <m1hey9z673.fsf@ptw.com>

Hard to see how either of these can be guarranteed to be unique any
more than:
Message-ID: <m1hey9z673.fsf@super.clever.insertion>

Re: Sender header?

[Bjørn Mork]

prj@po.cwru.edu (Paul Jarc) writes:
> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> > On 25 May 2001, Paul Jarc wrote:
> >
> >> According to RFC 1036, Sender is for news more or less what Rat says
> >> it should be for mail - it's used for tracking down where a message
> >> entered the network.
> > 
> > RFC 1036 has this example:
> 
> Yes, but it also has this one:
> # If a gateway program enters a mail message into the network at host
> # unix.SRI.COM, the lines might read:
> #
> #           From: John.Doe@A.CS.CMU.EDU
> #           Sender: network@unix.SRI.COM
> 
> I don't think it's suggesting that network@unix.sri.com issupposed to
> be a valid email address.

RFC 1036 is confusing and unclear at best. It's better to look forward
and see what's proposed in
http://www.ietf.org/internet-drafts/draft-ietf-usefor-article-04.txt :

   A "sender" is the person or software (usually, but not always, the
   same as the poster) responsible for the operation of the posting
   agent or, which amounts to the same thing, for passing the article to
   the injecting agent. The sender is analogous to [MESSFOR]'s sender.

 [..]

 6.2.  Sender

   The Sender header specifies the mailbox of the entity which actually
   sent this article, if that entity is different from that given in the
   From header or if more than one address appears in the From header.
   This header SHOULD NOT appear in an article unless the sender is
   different from the author. This header is appropriate for use by
   automatic article posters. The content syntax makes use of syntax
   defined in [MESSFOR].

      Sender-content      = mailbox

([MESSFOR] was P. Resnick, "Internet Message Format Standard", 
draft-ietf-drums-msg-fmt-07.txt, March 1998, which is now published as
RFC 2822)

I.e. they propose changing the definition of the sender field in news
to mean the same as in mail. Would make things easier IMHO.


Bjørn

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Paul Jarc wrote:

> I think user-login-name@system-name would work best for news, but
> nothing specific is required.  I think user-mail-address would work
> too, though not as well, based on the explanatory text in 2.2.2.
> It's all pretty vague.

Right.  Hm.  Son has this:

               NOTE: The intent is that the Sender header have  a
               fairly  high probability of identifying the person
               who really posted the  article.   The  ability  to
               specify  a  From  header naming someone other than
               the poster is useful but can be abused.

It also talks about verifying the address, but given that
login-name@system-name may not be the right address.  Son of RFC 1036
_very_ clearly says, however, that the contents of the Sender header
should be a valid mailing address.

So, I'm left with the feeling that Son of RFC 1036 and RFC 2822
specify pretty much the same thing about the Sender header.  However,
IANAL.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> prj@po.cwru.edu (Paul Jarc) writes:
>> > Currently I've told gnus to stick my IP smtp server address in
>> > Message-ID.
>> 
>> Don't do that unless you control the SMTP server, and know that it
>> won't generate Message-IDs using that same RHS.  See my other message
>> for more details.
> 
> Now I'm really confused..
> Wait... shouldn't that smtp server AlWAYS use the same rhs?

Probably.  But if you control it, you could configure it to do
otherwise.  The main principle is that two different hosts should use
two different domain-like things for their Message-IDs.  So if you're
using your SMTP server's hostname to generate Message-IDs on a
different host, then you should make sure that the SMTP server doesn't
use the same thing.  If you don't control the SMTP server, then you
can't ensure this, and so you should use something else on your other
host.

> The machine has a name like (/etc/hosts entry):
> reader.local.lan        192.168.xx.xxx   reader
> There must be several (100/1000 ?) that could possibly have that FQDN.
> How is this uniqueness obtained?

Normally - i.e., "normal" at the time Message-ID was designed - you'd
have a static FQDN on the Internet, and you'd use that.

If possible, you should set things up so that Message-IDs are
generated by hosts with static FQDNs on the Internet, and those hosts
should use (probably) their own FQDNs for the RHS.  Hosts that don't
have static FQDNs should pass along the message without a Message-ID,
and the the host that has the static FQDN generate one.  This may
require special configuration on both sides.

If you can't do that, then try to use the FQDN of a host with a static
FQDN on the Internet and which you control, so you can tweak each of
their generation algorithms, if necessary, to ensure that the two
won't ever generate the same LHS by accident.

If you can't do that, then you're stuck, and all you can do is make a
best effort.  Throw some entropy into the RHS (maybe something like
o4igh2r3.username.this-does-not-exist.your-isp.com, where the first
part is preferably different for each message), and explain to anyone
who objects that Message-ID was not designed with your situation in
mind, and this is the best you can do.

> Hard to see how either of these can be guarranteed to be unique any
> more than:
> Message-ID: <m1hey9z673.fsf@super.clever.insertion>

They can't, except that by convention, the RHS is at least based on
the FQDN of the Internet host where the Message-ID was generated.  If
you stray too far from that convention, you run the risk of choosing
the same clever insertion that someone else did.  Stay as close to it
as your situation allows - if you can use a host's FQDN, use it;
otherwise, use your organization's domain name along with something
which is unique in that context, etc.


paul

Re: Sender header?

[Paul Jarc]

"Bjørn Mork" <bmork@dod.no> writes:
>  6.2.  Sender
> 
>    The Sender header specifies the mailbox of the entity which actually
>    sent this article, if that entity is different from that given in the
>    From header or if more than one address appears in the From header.
>    This header SHOULD NOT appear in an article unless the sender is
>    different from the author. This header is appropriate for use by
>    automatic article posters. The content syntax makes use of syntax
>    defined in [MESSFOR].
...
> I.e. they propose changing the definition of the sender field in news
> to mean the same as in mail. Would make things easier IMHO.

Yeah, that makes sense.  One wonders about an automatic poster, though -
an email address used to designate it would probably just forward to
the person ultimately responsible for its postings anyway.


paul

Re: Sender header?

[Harry Putnam]

prj@po.cwru.edu (Paul Jarc) writes:

> . . . . . . . . . . . . . . . . . . . . . . . . Hosts that don't
> have static FQDNs should pass along the message without a Message-ID,

Of the several choices presented, the one above looks like the only
operational one for me.  Can someone remind me how to turn off local
message-id generation?

PS-- I hope Eli doesn't see this, but I couldn't find it with an `i'
search on `message-id'..... he he.

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
| It's the difference between a recommendation and a requirement.

You should look up the definitions of "should" and "must" sometime.
-- 
Rat <ratinox@peorth.gweep.net>    \ Warning: pregnant women, the elderly, and
Minion of Nathan - Nathan says Hi! \ children under 10 should avoid prolonged
PGP Key: at a key server near you!  \ exposure to Happy Fun Ball.

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
| Either you have changed your mind, or I fundamentally don't grok what
| you're saying.  I thought you were saying that Gnus should create a
| Sender header (with whatever contents), but the above prevents Gnus
| from ever creating one.  What's the story?

Gnus should not automatically generate a Sender header.  That is my
personal opinion, which along with $5 will get you a small coffee at
Starbucks.

I have not been arguing that Gnus should generate Sender headers, I have
been arguing how Gnus should generate the fields given the premise that
Gnus do so.
-- 
Rat <ratinox@peorth.gweep.net>    \ When not in use, Happy Fun Ball should be
Minion of Nathan - Nathan says Hi! \ returned to its special container and
PGP Key: at a key server near you!  \ kept under refrigeration.

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
| The defacto standard for the locally interpreted string has changed.
| These days, people want to hide the login names for security purposes.

That is what things like sendmail's rewrite rules and tables are for.

| A domain-literal need not be a FQDN, it can also be a domain name.
| For example, uni-dortmund.de is a possible domain-literal, even though
| there is no such host.

A domain name is by its nature fully qualified.

| The way I see it, your requirements for `reasonably configured' are
| pretty steep, and what you want makes life really hard for people
| working on systems which don't match your idea of `reasonably
| configured'.  This is pointless.

While I tend to be very strict when it comes to networking, doing it right
means that things are easier for everyone, not harder.  It is when things
are done wrong that life is made difficult for some people.
-- 
Rat <ratinox@peorth.gweep.net>    \ Do not use Happy Fun Ball on concrete.
Minion of Nathan - Nathan says Hi! \ 
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
| Where do the RFCs support your notion of correctness WRT MX records?

MX records make it possible to deliver mail that would otherwise bounce off
of a host that is not configured to accept mail.  The rest should be as
obvious as adding 2 and 2.

[...]
| You are always the same person, and all your mailboxes identify you
| equally well all the time.  The fact that you're using only one of
| them at any given moment changes nothing.

You are incorrect.

| I see no such requirement.  Using an unqualified name like
| "prj@multivac" will indeed break things, of course.  But you're also
| inventing the requirement for a *particular* FQDN - the local one.
| This requirement is not in RFC 2822.

You are again incorrect.  Read the definitions again.
-- 
Rat <ratinox@peorth.gweep.net>    \ Caution: Happy Fun Ball may suddenly
Minion of Nathan - Nathan says Hi! \ accelerate to dangerous speeds.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
>| The defacto standard for the locally interpreted string has changed.
>| These days, people want to hide the login names for security purposes.
> 
> That is what things like sendmail's rewrite rules and tables are for.

The fact that it can be done outside of Gnus doesn't alone make a good
case that it shouldn't also be possible to do it within Gnus.  Not all
MTAs may have such features.

>| A domain-literal need not be a FQDN, it can also be a domain name.
>| For example, uni-dortmund.de is a possible domain-literal, even though
>| there is no such host.
> 
> A domain name is by its nature fully qualified.

RFC 1034 doesn't use the term "fully qualified", but it does use
"absolute" and "relative".  A domain name is a sequence of labels
which, in text representation, are separated by dots.  An absolute
domain name is one whose last label is the empty string, which
designates the root domain - thus, its last character is a dot.
Anything else is relative, and is subject to absolutification
according to local configuration.  (Also, there is no notion of a
"hostname" which is different from "domain name"; a hostname is simply
a domain name with an address record, and the term "hostname" is not
used in the DNS RFCs.)

In practice, people think of a domain name containing a dot *anywhere*
as being absolute ("fully qualified"), and only those with *no* dots
as being relative.  The only reason people can get away with this is
because their own local configurations make it work.  The name
"www.gnus.org." is absolute, but the name "www.gnus.org" is not.  The
most common way of making it absolute is to first try it in the root
domain (i.e., just add a dot at the end), and then if that fails, try
it in the local domain(s).  OTOH, the name "www" is usually looked for
first in the local domain(s) and then in the root.  But the presence
of an internal dot only affects the local absolutification
configuration, not whether the name is absolute to begin with.

I've set user-mail-address to end with "." (I think this violates
2822, but I won't care until it actually breaks something), and I set
message-syntax-checks to include (from . disabled), but my From fields
still have no trailing ".". :(  Anyone know how to do this?  I don't
want to depend on others' local configurations.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
>| Where do the RFCs support your notion of correctness WRT MX records?
> 
> MX records make it possible to deliver mail that would otherwise bounce off
> of a host that is not configured to accept mail.

Right (although it also requires help from the server named in the MX
record - even if that server accepts mail, it doesn't necessarily
accept mail addressed to arbitrary domains).  But who says I want to
receive such mail anyway?  If all the addresses I use have
"@domain.com", then I can expect that no one will send mail addressed
to blah.domain.com, so I have no need to bother with extra MX records.

>| You are always the same person, and all your mailboxes identify you
>| equally well all the time.  The fact that you're using only one of
>| them at any given moment changes nothing.
> 
> You are incorrect.

Gee, thanks for explaining.

I see nothing in 2822 to suggest that Sender should indicate the host
where a message originated.  If you do, please point it out.

>| I see no such requirement.  Using an unqualified name like
>| "prj@multivac" will indeed break things, of course.  But you're also
>| inventing the requirement for a *particular* FQDN - the local one.
>| This requirement is not in RFC 2822.
> 
> You are again incorrect.  Read the definitions again.

The string "prj@multivac" is generated by the grammar for "mailbox".
If the requirement for a FQDN is somewhere else, I missed it.  And
actually, the grammar *prohibits* an absolute domain name as defined
by RFC 1034.  Or did you mean something other than "absolute" by:
> The domain part of an RFC 2822 mailbox must be a fully qualified domain
> name.


paul

Re: Sender header?

[Christoph Conrad]

    > instructions, the practical utility of Sender is on a par with

I always bother in my company when official statements from the CEO
are mailed to all from the secretary with their "From:". It would be
nice if the mails would come from the CEO, and the sender would be the
secretary (in case of technical problems, e.g. unreadable parts of the
message), as the standard explains this. I think that's quite useful.

    > the practical utility of Resent-From, i.e., none at all.

I also think this is quite useful. Sometimes i re-send mail to other
people, and they can see in the header that it is not originally from
me. Often i re-send mail from the company to home or vice versa. I
like the re-sent header.

Best regards,
cu, -cc-
-- 
=> GNU Emacs Webring @ <http://www.gnusoftware.com/WebRing/> <=
Look Ma, this man can twist his fingers as if they were made of rubber,
isn't that amazing? -- Not really, he's been using emacs for years...!

Re: Sender header?

[Kai Großjohann]

On Fri, 25 May 2001, Stainless Steel Rat wrote:

> While I tend to be very strict when it comes to networking, doing it
> right means that things are easier for everyone, not harder.  It is
> when things are done wrong that life is made difficult for some
> people.

I'm thinking about the people who are not in a position to do it
right, but have to live with what they have: a customer at a dialup
isp has no control over the ISP's MTA config, for example.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On Fri, 25 May 2001, Stainless Steel Rat wrote:

> Gnus should not automatically generate a Sender header.  That is my
> personal opinion, which along with $5 will get you a small coffee at
> Starbucks.

I'd need a few hundred bucks to get to Starbucks, first :-)

I see.  So you think Gnus should not generate a Sender header.
Amazing how long we talked without me realizing this.  How about
having Gnus check the message before sending, and if the user has
manually entered a From header, have Gnus tell the user to add a
Sender header, too?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Harry Putnam wrote:

> If I allowed gnus to generate the Message-ID (rhs) by its own
> devices, then it would be:
> 
> Message-ID: <m1hey9z673.fsf@reader.local.lan>
> 
> Instead of:
> Message-ID: <m1hey9z673.fsf@ptw.com>
> 
> Hard to see how either of these can be guarranteed to be unique any
> more than:
> Message-ID: <m1hey9z673.fsf@super.clever.insertion>

I think you post via Newsguy, right?  Hm.  Here's how to find a good
rhs for the message-id: talk to the news server admin at newsguy.
They might have reserved a set of rhs entries for this case.  For
example, they could say that the rhs jrl.customers.newsguy.com, where
jrl is the login name or account name or whatever you have that
identifies you, can be used.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Jesper Harder]

abraham@dina.kvl.dk (Per Abrahamsen) writes:

> Russ Allbery <rra@stanford.edu> writes:
> 
> > (setq message-syntax-checks '((sender . disabled)))
> > 
> > has worked great for years.  Hopefully it hasn't been broken in later
> > versions of Gnus.  It should be the default.
> 
> I agree, 

Me to.

> except that it breaks cancels when using 'gnus-posting-styles'.
> Nonetheless, I'd actually prefer disabling sender, and let someone who
> cares about 'gnus-posting-styles' fix that.

An easy way to fix it would be to use cancel locks instead of the
heuristics used in message-cancel-news to determine if you're allowed to
cancel.

All you need is to replace the checking code in message-cancel-news
with:

   (if (canlock-verify) (error "This article is not yours"))

and add this to .gnus

(add-hook 'message-header-hook 'canlock-insert-header)
(setq canlock-password "hemmeligt")
(require 'canlock)

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Harry Putnam wrote:

> Of the several choices presented, the one above looks like the only
> operational one for me.  Can someone remind me how to turn off local
> message-id generation?

The variable message-syntax-checks ought to do the trick.  Or you say
this: (setq message-required-news-headers (delete 'Message-ID
message-required-news-headers))

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Bjørn Mork]

prj@po.cwru.edu (Paul Jarc) writes:

> I've set user-mail-address to end with "." (I think this violates
> 2822, but I won't care until it actually breaks something), and I set
> message-syntax-checks to include (from . disabled), but my From fields
> still have no trailing ".". :(  Anyone know how to do this?  I don't
> want to depend on others' local configurations.

Umm, this reminds me of the time I defined a wildcard MX record for a
domain name that was in my search list. A few trailing dots might have
saved it, but who uses that? At least I learnt one reason why wildcard
MX records are discouraged...


Bjørn
-- 
Why, it's a wonderful day!  

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Paul Jarc wrote:

> I see nothing in 2822 to suggest that Sender should indicate the
> host where a message originated.  If you do, please point it out.

My idea of Rat's reasoning: Sender should be canonical, canonical
means login@fully.qualified.host.name.  Qed.

(Maybe there's an additional step `Sender should be verified, verified
means canonical' in the beginning.)

Right, Rat? ;-)

I doubt that he will quote chapter and verse to show that verified
means canonical, or that Sender should be canonical.  That just
follows from common sense.  Or 30 years of practice.

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> On 25 May 2001, Paul Jarc wrote:
>> I see nothing in 2822 to suggest that Sender should indicate the
>> host where a message originated.  If you do, please point it out.
> 
> My idea of Rat's reasoning: Sender should be canonical, canonical
> means login@fully.qualified.host.name.  Qed.

None of that is in the RFC, though, and violating it won't cause any
problems in practice.  Also note the difference between an address
verified by software and one constructed by software.  (I find this
use of "canonical" rather odd, but we can ignore that.)

> I doubt that he will quote chapter and verse

Me, too.  Which suggests that either there is indeed no such
requirement, or that he doesn't care enough about Gnus getting it
right to show us where it is.


paul

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Paul Jarc wrote:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> 
>> I doubt that he will quote chapter and verse
> 
> Me, too.

If, however, the whole thing comes from 30 years of practice, it's not
possible to quote chapter and verse.  Hm.  Maybe we just don't
understand.  (I think I do understand, but I'm biased...)

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> If, however, the whole thing comes from 30 years of practice, it's not
> possible to quote chapter and verse.

But in that case, you'd expect him to say that, instead of things
like:
> Frobbing [Sender] directly would be a violation of the requirements of
> RFCs 2822 and 1034.

and:
> The canonical mailbox is required by RFC 2822.

Kai again:
> Hm.  Maybe we just don't understand.  (I think I do understand, but
> I'm biased...)

Aren't we all?


paul

Re: Sender header?

[Harry Putnam]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> I think you post via Newsguy, right?  Hm.  Here's how to find a good
> rhs for the message-id: talk to the news server admin at newsguy.

Post yes, but mail is sent out thru my IP smtp server.  

> They might have reserved a set of rhs entries for this case.  For

Possible, I'll check it out

So is your thinking that I should use two rhs for message ID?
One for usenet and a different one for mail?

My outgoing smtp server is not newsguy, it could be but its just not
as simple to setup and its slower that way,

This seems pretty similar to the case about sender header, in that the
question is which machine should put its `John Hancock'*  on it.
In fact the rhs of message-ed parrallels `Sender' in many ways.

[* `John Hancock' is used this way in american idiom to mean signature.
John Hancock was one of the original signers of our constitution] 

PS- I've removed any Message-id from this message on my end, so will
see what smtp server does.

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
| The fact that it can be done outside of Gnus doesn't alone make a good
| case that it shouldn't also be possible to do it within Gnus.  Not all
| MTAs may have such features.

Any MTA that cannot do header rewriting has no business playing mail
gateway for a firewalled site.

The other side of that is if your MTA is configured to do all the
rewriting, you can use *any* MUA you want and everything, not just Gnus,
will just work, right out of the box.  Making all this transparent to the
end users is a Good Thing(tm).

[...]
| RFC 1034 doesn't use the term "fully qualified", but it does use
| "absolute" and "relative". [...]

The point of bringing up RFC 1034's less than precise language is...?

Within the context of mail and news, the root-level dot is understood to be
there even if it is not actually there.  I really don't see the point of
this digression.
-- 
Rat <ratinox@peorth.gweep.net>    \ Do not taunt Happy Fun Ball.
Minion of Nathan - Nathan says Hi! \ 
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
| Right (although it also requires help from the server named in the MX
| record - even if that server accepts mail, it doesn't necessarily
| accept mail addressed to arbitrary domains).

But the server mail.foo.com does not accept mail for arbitrary domains, it
accepts mail for foo.com and those domains it controls.

| But who says I want to receive such mail anyway?  If all the addresses I
| use have "@domain.com", then I can expect that no one will send mail
| addressed to blah.domain.com, so I have no need to bother with extra MX
| records.

Because that is not how the real world works, except in dotcomville where
the namespace is flat.  Case in point, I have a spam trap mailbox on
peorth's primary MX that does not exist on her secondary MX.  If you try to
send mail to ratspahn@gweep.net, it will bounce, but mail to
ratspahn@rei.nerv.gweep.net will be delivered.

| Gee, thanks for explaining.

| I see nothing in 2822 to suggest that Sender should indicate the host
| where a message originated.  If you do, please point it out.

It doesn't, because a message's sender has absolutely nothing to do with
where the message originated.

| The string "prj@multivac" is generated by the grammar for "mailbox".

It is not an RFC 2822 mailbox.  It may be a local mailbox to you, but we
are not talking about local mail, we are talking about Internet mail.

| If the requirement for a FQDN is somewhere else, I missed it.

You missed it.  From RFC 2822 section 3.4.1:

  "In the domain-literal form, the domain is interpreted as the literal
  Internet address of the particular host."

A "literal Internet address" of a host is its local host name, a dot, and
the local domain name.  A mailbox is type addr-spec.  addr-spec is
`local-part "@" domain'.  domain is `dot-atom / domain-literal /
obs-domain'.

Colloquially, a mailbox is local-part@fqdn (less the root dot).
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball may stick to certain types
Minion of Nathan - Nathan says Hi! \ of skin.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
| I'm thinking about the people who are not in a position to do it
| right, but have to live with what they have: a customer at a dialup
| isp has no control over the ISP's MTA config, for example.

My take on that is, if I am paying for the service then I have a reasonable
expectation that the service not be broken.
-- 
Rat <ratinox@peorth.gweep.net>    \ Caution: Happy Fun Ball may suddenly
Minion of Nathan - Nathan says Hi! \ accelerate to dangerous speeds.
PGP Key: at a key server near you!  \ 

Re: Sender header?

[Stainless Steel Rat]

* Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann)  on Fri, 25 May 2001
| I see.  So you think Gnus should not generate a Sender header.  Amazing
| how long we talked without me realizing this.

I think any MUA should be as conservative as it possibly can while being as
accurate as possible.

| How about having Gnus check the message before sending, and if the user
| has manually entered a From header, have Gnus tell the user to add a
| Sender header, too?

Of course, if we have a hundred people asking how to turn off Sender now,
Gnus goes and does that and we will have three hundred asking how to turn
off this new behaviour.  It is difficult to call that an improvement. :)
-- 
Rat <ratinox@peorth.gweep.net>    \ Happy Fun Ball contains a liquid core,
Minion of Nathan - Nathan says Hi! \ which, if exposed due to rupture, should
PGP Key: at a key server near you!  \ not be touched, inhaled, or looked at.

Re: Sender header?

[Russ Allbery]

Paul Jarc <prj@po.cwru.edu> writes:

> I want to disable Sender for mail, but not for news.  Can that be easily
> done?

OOC, why do you want to keep it for news?  The only earthly purpose that I
can think of it serving is to provide spammers with another possibly valid
e-mail address to use.  It's not like anyone on Usenet *cares* about the
actual account that you were using when you posted.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Harry Putnam]

Richard Hoskins <rmh@bandersnatch> writes:

> Harry Putnam <reader@newsguy.com> writes:
 
> [...]
> 
> > [* `John Hancock' is used this way in american idiom to mean
> > signature.  John Hancock was one of the original signers of our
> > constitution]
>   ^^^^^^^^^^^^
> 
> You misspelled "Declaration of Independance."
> 
> AFAIK, Mr. Hancock wasn't a member of the Constitutional Convention.

Well at least I didn't spell it `Magna Carta'

There goes my career as an authority on historical american idiom.

My attempt was mainly for Kai, who likes to know about stuff like that.
Sorry Kai, I'll research before post in future... 

Re: Sender header?

[Barry Fishman]

Thanks for all the help.

Since I don't know how to respond to multiple useful messages at the
same time I'll try.  I think this is important since there must be a
many users that are comming in via ppp phone connections that consider
only windows machines.  I think the gnus manual should deal with it.
I didn't find anything.

Stainless Steel Rat <ratinox@peorth.gweep.net> on Thu, 26, May 2001
SSR> * Barry Fishman <barry_fishman@acm.org>  on Thu, 26 May 2001
SSR> | My login@fqdn is "barry@ecube.local".
SSR>
SSR> * snrk*.
SSR>
SSR> That is not a legal fqdn, or did someone create the "local" TLD when I
SSR> blinked? :)

I did!  Internet domain setup on unix requires a domain name.  Like most
people I don't want to buy one.  Rather than supply something that
looks correct, but isn't, I use local, which I think is clear in
meaning.  I don't spend all my time connected through att.net.  Even
when I am connected, I am not sure what my A record would be.  It does
not show up in my PPP/PAP dialog.

SSR> Given that att.net won't let you do things right, dumping it on them is
SSR> probably not a bad thing to do.  They've screwed it up so let them deal
SSR> with it.

I've tried that. The just use the @ecube.local which they got from
my SMTP HELO, add their dommain and then the regular hashing.  I finally
decided to use a unique identifier (although not really a legal
domain name, my e-mail address.  It was just as unique and a lot
shorter. Stealing the idea from:

Harry Putnam <reader@newsguy.com> writes:
HP> I got hollered at on comp.editors for having a made up rhs of message
HP> id which is somewhat related to this discussion.
HP> 
HP> Describing another situation where it isn't really clear what I
HP> should have for a From or sender address.
HP> 
HP> Currently I've told gnus to stick my IP smtp server address in
HP> Message-ID.  And disabled sender generation.
HP> 
HP> (defun message-make-fqdn ()
HP>   "My  hacked message-id."
HP>   "ptw.com")

Here you are posing as another machine,  A sane person would
expect the Message-ID's to still be unique with all the hashing added.
However, I'm a programmer and would want to at least make my fqdn
unique and identifiable.

Russ Allbery <rra@stanford.edu> writes:
RA> (setq message-syntax-checks '((sender . disabled)))
RA>
RA> has worked great for years.  Hopefully it hasn't been broken in later
RA> versions of Gnus.  It should be the default.

This seems better.  Personally I think that "Sender:" is like "From:"
and "ReplyTo:",  chosen by person/agent sending the mail and not
automaticlly produced.  Defaults are nice, but if gnus wants too
construct a "Sender:", it probably should be set to the person who wrote
the elisp that constructs a sender. ;)  Mail transport agents have their
own fields to fill in if they don't like the from field.

On 25 May 2001, Harry Putnam wrote:
HP> Of the several choices presented, the one above looks like the only
HP> operational one for me.  Can someone remind me how to turn off local
HP> message-id generation?
 
Kai Grossjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:
KG> The variable message-syntax-checks ought to do the trick.  Or you say
KG> this: (setq message-required-news-headers (delete 'Message-ID
KG> message-required-news-headers))                                                
Until I can confirm a A record, I'll use the elisp:

;; I don't really have valid FQDN information so clean up header
(setq message-syntax-checks '((sender . disabled)))

(defun smtpmail-fqdn ()
   "Supply a domain which is meaningful but wrong (my email address)"
   "barry_fishman.att.net")

(defun message-make-fqdn ()
  "Supply a unique root for message id"
  (smtpmail-fqdn))                                                             

I have Kai's removal of Message-ID commented out since I think gnus's
Message-ID is much shorter.

This seems to leave behind the most meaningful mail headers.  Probably
too meaningful if you which to avoid spam.

Barry

Re: Sender header?

[Harry Putnam]

Barry Fishman <barry_fishman@acm.org> writes:

> (defun message-make-fqdn ()
>   "Supply a unique root for message id"
>   (smtpmail-fqdn))                                                             

What do you get with that?

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Harry Putnam wrote:

> So is your thinking that I should use two rhs for message ID?
> One for usenet and a different one for mail?

Well, afaik Message-ID in both news and mail is only required to be
unique.  Now if you know that noone is using the newsguy rhs for news
articles, it's also reasonable to assume that noone is using the
newsguy rhs for mail, either.

kai

PS: The German for `John Hancock' is `Bill'.  Interesting, isn't it?
    (Actually, it's `Willi' which is the German short form of Wilhelm,
    which is German for William.)  Maybe there was a German by that
    name.  Hm.  Yes, we had a monarch.  Emperor(?) William the
    first(?) I think...  I have no idea if it's him, though.  Any
    Germans here to elucidate?
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 25 May 2001, Russ Allbery wrote:

> OOC

?  What does this expand do?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 26 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Fri, 25 May
>   2001
> 
>| How about having Gnus check the message before sending, and if the
>| user has manually entered a From header, have Gnus tell the user to
>| add a Sender header, too?
> 
> Of course, if we have a hundred people asking how to turn off Sender
> now, Gnus goes and does that and we will have three hundred asking
> how to turn off this new behaviour.  It is difficult to call that an
> improvement. :)

He he.  However, this depends on how you define `manually entered'.
One possibility would be to say that a From header which comes from
user-mail-address is _not_ manually entered.  Only if the header is
_really_ manually entered, should we ask.

How many people do M-<, then type `From: SPC bla RET'?

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Kai Großjohann]

On 26 May 2001, Stainless Steel Rat wrote:

> * Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) on Fri, 25 May
>   2001
>| I'm thinking about the people who are not in a position to do it
>| right, but have to live with what they have: a customer at a dialup
>| isp has no control over the ISP's MTA config, for example.
> 
> My take on that is, if I am paying for the service then I have a
> reasonable expectation that the service not be broken.

Well, they don't _say_ that SMTP is guaranteed to work.  Maybe they
only sell HTTP...  (And Realaudio and this Internet radio stuff,
maybe.)

kai
-- 
~/.signature: No such file or directory

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> John Hancock was one of the original signers of our constitution] 

He may (or may not) have signed the Constitution; he's famous for his
signature on the Declaration of Independence.

> PS- I've removed any Message-id from this message on my end, so will
> see what smtp server does.

I see: Message-ID: <m17kz5xaln.fsf@ptw.com>
But if this were generated by an SMTP receiver, then I would expect it
to be mixed in with the Received fields, which it wasn't.  Are you
sure you disabled msgid generation entirely, both in Gnus and local
MTA?


paul

Re: Sender header?

[Paul Jarc]

Russ Allbery <rra@stanford.edu> writes:
> Paul Jarc <prj@po.cwru.edu> writes:
>> I want to disable Sender for mail, but not for news.  Can that be easily
>> done?
> 
> OOC, why do you want to keep it for news?

I was going by my interpretation of 1036.  Newer specs have convinced
me that I don't need Sender for news either.

> It's not like anyone on Usenet *cares* about the actual account that
> you were using when you posted.

Well, there isn't quite any equivalent for Received (Path isn't quite
as informative), so it would indeed be removing information.  When all
is well, this wouldn't matter; when things break, extra information
might help.


paul

Re: Sender header?

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> On 25 May 2001, Russ Allbery wrote:
>> OOC
> 
> ?  What does this expand do?

"Out of curiosity", I'm guessing.


paul

Re: Sender header?

[Harry Putnam]

prj@po.cwru.edu (Paul Jarc) writes:

> I see: Message-ID: <m17kz5xaln.fsf@ptw.com>
> But if this were generated by an SMTP receiver, then I would expect it
> to be mixed in with the Received fields, which it wasn't.  Are you
> sure you disabled msgid generation entirely, both in Gnus and local
> MTA?

Yeah, so I noticed.  I just removed message-id field when mailing.
I have a variable set that puts all outgoing (gnus generated) headers
in the compose buffer.  I thought just removing it there would cause
gnus to omit it.  Apparently not.  Apparently gnus regenerated it on
send because that is a gnus/emacs Message-id.

Apparently that doesn't work.  
The method Kai posted does work, but still sendmail or something
generates a local message-id.
I've decided for the moment to just generate
the right thing with gnus by hacking.  Doesn't seem possible to stop
message id generation locally.  Gnus is easy, but the mta is another
story.   What a big pain in the ass the whole thing is.

I'd hoped to disable message id generation locally and let my outgoing
smtp IP machine handle it.  But Just stopping gnus doens'nt produce
that effect either.  I tried in other messges else were after
disabling gnus altogether and my Machine name still gets stuck in there by
sendmail I guess. 
Can one stop sendmail from generating a message-id?  If I can do that
then mutt pine or whatever will all stop sticking in my local machine
name I hope.

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
>| Right (although it also requires help from the server named in the MX
>| record - even if that server accepts mail, it doesn't necessarily
>| accept mail addressed to arbitrary domains).
> 
> But the server mail.foo.com does not accept mail for arbitrary domains, it
> accepts mail for foo.com and those domains it controls.

It may.  Or it may not, depending on what the admin wants.  Nothing is
necessarily broken with a server that accepts mail addressed to
foo.com but rejects mail addressed to bar.foo.com.

>| But who says I want to receive such mail anyway?  If all the addresses I
>| use have "@domain.com", then I can expect that no one will send mail
>| addressed to blah.domain.com, so I have no need to bother with extra MX
>| records.
> 
> Because that is not how the real world works, except in dotcomville where
> the namespace is flat.

And RFC 2822 applies to dotcomville just as everywhere else.  I'm not
saying that accepting mail for subdomains is a bad or uncommon
practice.  I'm just saying it's not required by any RFC, and it isn't
necessary in all situations.

>| I see nothing in 2822 to suggest that Sender should indicate the host
>| where a message originated.  If you do, please point it out.
> 
> It doesn't, because a message's sender has absolutely nothing to do with
> where the message originated.

But you would require Gnus to use the hostname where the message
originated, right?  You've said this is a consequence of *some*
requirement of 2822; show it to us.

>| The string "prj@multivac" is generated by the grammar for "mailbox".
> 
> It is not an RFC 2822 mailbox.

I would agree that RFC probably doesn't intend to include such strings
as mailboxes, but I'm not certain it accomplishes that goal.  The
grammar alone certainly isn't enough to do it.

>| If the requirement for a FQDN is somewhere else, I missed it.
> 
> You missed it.  From RFC 2822 section 3.4.1:
> 
>   "In the domain-literal form, the domain is interpreted as the literal
>   Internet address of the particular host."
> 
> A "literal Internet address" of a host is its local host name, a dot, and
> the local domain name.

Wrong.  You're trying to apply too much intelligence - the "address"
is just that; you don't have to figure out what they really meant (in
this case).  E.g., "prj@[129.22.4.2]".  Anyway, this is just one
possible form of address; requirements on one form don't necessarily
apply to other forms of addresses.

> A mailbox is type addr-spec.  addr-spec is `local-part "@" domain'.
> domain is `dot-atom / domain-literal / obs-domain'.

And dot-atom matches "foo".  In order to require an internal dot via
the grammar, this rule:
  dot-atom-text   =       1*atext *("." 1*atext)
would have to be replaced with:
  dot-atom-text   =       1*atext 1*("." 1*atext)

> Colloquially, a mailbox is local-part@fqdn (less the root dot).

Colloquially, yes.


paul

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> The method Kai posted does work, but still sendmail or something
> generates a local message-id.  I've decided for the moment to just
> generate the right thing with gnus by hacking.  Doesn't seem
> possible to stop message id generation locally.  Gnus is easy, but
> the mta is another story.

qmail makes it easy to use a particular hostname for Message-IDs.
Disabling Message-ID generation would take a little hacking, though.
But have you checked whether your ISP's SMTP server will generate
Message-IDs?  I wouldn't expect it to.


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Fri, 25 May 2001
>| The fact that it can be done outside of Gnus doesn't alone make a good
>| case that it shouldn't also be possible to do it within Gnus.  Not all
>| MTAs may have such features.
> 
> Any MTA that cannot do header rewriting has no business playing mail
> gateway for a firewalled site.

I'm not sure that's true in all cases.  I think it's possible to set
things up such that no rewriting is needed on the gateway.

> The other side of that is if your MTA is configured to do all the
> rewriting, you can use *any* MUA you want and everything, not just Gnus,
> will just work, right out of the box.  Making all this transparent to the
> end users is a Good Thing(tm).

Yes.  Another way to make things work is to know that 2822 doesn't
require any particular address for Sender, or any software effort to
generate one, or any software effort to verify a user-supplied one.

>| RFC 1034 doesn't use the term "fully qualified", but it does use
>| "absolute" and "relative". [...]
> 
> The point of bringing up RFC 1034's less than precise language is...?

This part of 1034 is perfectly precise.  But I though some folks might
be interested; people often toss around "fqdn" without knowing the
whole story.  I also wanted to clarify that there is no necessary
difference between a "hostname" and a "domain name" in terms of, e.g.,
the number of dots.


paul

Re: Sender header?

[Russ Allbery]

Paul Jarc <prj@po.cwru.edu> writes:

> Well, there isn't quite any equivalent for Received (Path isn't quite as
> informative), so it would indeed be removing information.  When all is
> well, this wouldn't matter; when things break, extra information might
> help.

I can't think of a form of breakage for a personal post (as opposed to a
gateway or the like) where the information in Sender would be particularly
useful.  NNTP-Posting-Host, X-Trace, and the like, yes, but that's
different.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Sun, 27 May 2001
| It may.  Or it may not, depending on what the admin wants.  Nothing is
| necessarily broken with a server that accepts mail addressed to
| foo.com but rejects mail addressed to bar.foo.com.

Then foo.com's mail gateway should masqerade mail headers so that nobody
outside ever sees bar.foo.com.  If it does not, and there is no server that
accepts mail for bar.foo.com, then the whole site is configured badly.

[...]
| But you would require Gnus to use the hostname where the message
| originated, right?  You've said this is a consequence of *some*
| requirement of 2822; show it to us.

If Gnus is generating headers automatically, then yes, that is what it
should use.  It is up to the site's mail gateway to masqerade what needs to
be masqeraded.

[...]
| Wrong.  You're trying to apply too much intelligence - the "address"
| is just that; you don't have to figure out what they really meant (in
| this case).  E.g., "prj@[129.22.4.2]".  Anyway, this is just one
| possible form of address; requirements on one form don't necessarily
| apply to other forms of addresses.

Now you are deliberately being a pain in the ass.
-- 
Rat <ratinox@peorth.gweep.net>    \ If Happy Fun Ball begins to smoke, get
Minion of Nathan - Nathan says Hi! \ away immediately. Seek shelter and cover
PGP Key: at a key server near you!  \ head.

Re: Sender header?

[Stainless Steel Rat]

* prj@po.cwru.edu (Paul Jarc)  on Sun, 27 May 2001
| I'm not sure that's true in all cases.  I think it's possible to set
| things up such that no rewriting is needed on the gateway.

What do you mean, you are not sure?  Have you never set up a masqerading
SMTP gateway?  Have you never had to deal with users who don't know what
the hell that means?  Sounds like you never have.

I do, every working day.  Everything you have said probably looks good to
you on paper, but it is worthless in the real world.  Because the users
cannot tweak Sender and Message-ID fields with MUAs like Outlook.  You'll
have to get in line behind me when it comes to saying that Outlook is
broken, but the fact is that the majority of people at my site use it.  I
needed something that works for them.  That something is masqerading, and
it automagically works for me using Gnus as well.

The last thing I will say to you, Paul, is to shut up until you have some
real world experience in dealing with mail handling, because right now you
have no idea what you are talking about, and you are just making things
difficult for everyone by spewing crap for answers.
-- 
Rat <ratinox@peorth.gweep.net>    \ Ingredients of Happy Fun Ball include an
Minion of Nathan - Nathan says Hi! \ unknown glowing substance which fell to
PGP Key: at a key server near you!  \ Earth, presumably from outer space.

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Sun, 27 May 2001
>| It may.  Or it may not, depending on what the admin wants.  Nothing is
>| necessarily broken with a server that accepts mail addressed to
>| foo.com but rejects mail addressed to bar.foo.com.
> 
> Then foo.com's mail gateway should masqerade mail headers so that nobody
> outside ever sees bar.foo.com.

Or the internal clients could be configured to use addresses that work
from the outside.  You may have a preference for one strategy, but
that doesn't mean the other doesn't work, or is prohibited by any RFC.

>| But you would require Gnus to use the hostname where the message
>| originated, right?  You've said this is a consequence of *some*
>| requirement of 2822; show it to us.
> 
> If Gnus is generating headers automatically, then yes, that is what it
> should use.

Are you still claiming that an RFC requires this?  If so, which and
where?  If not, tell us exactly why user-mail-address is worse.  It
isn't for informative value for humans tracking down problems, because
Received still carries all the information you need.

>| Wrong.  You're trying to apply too much intelligence - the "address"
>| is just that; you don't have to figure out what they really meant (in
>| this case).  E.g., "prj@[129.22.4.2]".  Anyway, this is just one
>| possible form of address; requirements on one form don't necessarily
>| apply to other forms of addresses.
> 
> Now you are deliberately being a pain in the ass.

No, everything in that paragraph is true.  Try sending some test
messages addressed to [domain.com] and to [1.2.3.4]; see which ones
reach the SMTP server.  (Which ones are accepted, once received, is
another matter.)


paul

Re: Sender header?

[Paul Jarc]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> * prj@po.cwru.edu (Paul Jarc)  on Sun, 27 May 2001
>| I'm not sure that's true in all cases.  I think it's possible to set
>| things up such that no rewriting is needed on the gateway.
> 
> What do you mean, you are not sure?  Have you never set up a masqerading
> SMTP gateway?  Have you never had to deal with users who don't know what
> the hell that means?  Sounds like you never have.

That's right.  But who cares?  I'm not saying that it's feasible in
all situations - only for some.  Possibly this would require a
controlled environment, with no clueless users to deal with, etc.  You
may say that this is not "the real world", or that you don't care
about dealing with such circumstances, but that doesn't mean that they
don't exist, can't be created, or are prohibited by any RFC.

> Everything you have said probably looks good to you on paper, but it
> is worthless in the real world.

I never claimed it was useful in what you think of as "the real
world".  I only claim that it could be useful in some situations, and
it is permitted by the RFCs; in particular, I do not claim anything
about how common such situations might be.

> The last thing I will say to you, Paul, is to shut up until you have some
> real world experience in dealing with mail handling, because right now you
> have no idea what you are talking about, and you are just making things
> difficult for everyone by spewing crap for answers.

No, *you* don't know what I'm talking about.  You think I'm trying to
be ambitious and make (logically) strong claims which are false.  And
I agree that those claims are false, but those aren't the claims I'm
making.  I'm making (logically) weaker claims which are true.


paul

Re: Sender header?

[Paul Jarc]

Harry Putnam <reader@newsguy.com> writes:
> Barry Fishman <barry_fishman@acm.org> writes:
>> (defun message-make-fqdn ()
>>   "Supply a unique root for message id"
>>   (smtpmail-fqdn))
> 
> What do you get with that?

Just above that defun, he wrote:
(defun smtpmail-fqdn ()
   "Supply a domain which is meaningful but wrong (my email address)"
   "barry_fishman.att.net")


paul

Re: Sender header?

[Paul Jarc]

Barry Fishman <barry_fishman@acm.org> writes:
> I don't spend all my time connected through att.net.  Even when I am
> connected, I am not sure what my A record would be.  It does not
> show up in my PPP/PAP dialog.

The A (address) record would be a mapping from a domain name (like
"somehost.att.net") to an IP address.  It would be added to att.net's
DNS servers when you connect.  It wouldn't appear as such in your
dialup dialog, but the same information it contains - your Internet
hostname and address - probably does appear.


paul

Re: Sender header?

[Amos Gouaux]

Frightening thread.  

Personally, what I'm most concerned about is the envelope sender.
If I set an identity in PINE, the "Sender:" will be
me@mymachine.domain, but the "From:" and the envelope sender
(eventually "Return-Path:") will be the identity address.  If the
envelope sender doesn't match the "From:" (before posting to a
list), then it's really hell to deal with.

Currently, to deal with this in gnus, I use the following (this was
done a while ago, so the comments might be way out of date now):

;;; Using smtpmail because it allows me to easily keep the "From:" address
;;; and envelope sender the same, just by setting `user-mail-address' via
;;; some posting styles (see below).  Why is this important?  Well, some
;;; MLMs are really anal about using the envelope sender rather than the
;;; "From:" address.  So this just makes sure I don't have too many headaches.
(load "smtpmail" nil t) ;;;; loading my hacked version.....
(add-hook
 'message-load-hook
 (function
  (lambda ()
    ;;; Envelope sender still a problem... try this...
    (add-to-list 'message-syntax-checks '(sender . disabled))
    (require 'smtpmail)
    (setq send-mail-function 'smtpmail-send-it)
    (setq message-send-mail-function 'smtpmail-send-it)
    (setq smtpmail-smtp-server "smtp.utdallas.edu"))))

I guess I also have to add that while the new RFCs do seem to
clarify some points, there still seems to be some room for
interpretation.  For example, take this portion from 2821:

B. Generating SMTP Commands from RFC 822 Headers

   2. The return address in the MAIL command SHOULD, if possible, be
      derived from the system's identity for the submitting (local)
      user, and the "From:" header field otherwise.  If there is a
      system identity available, it SHOULD also be copied to the Sender
      header field if it is different from the address in the From
      header field.  (Any Sender field that was already there SHOULD be
      removed.)

which seems to confirm what Rat has been saying.  HOWEVER, the very
next sentence reads:

      Systems may provide a way for submitters to override the
      envelope return address, but may want to restrict its use to 
      privileged users.  This will not prevent mail forgery, but may
      lessen its incidence; see section 7.1.

Well, this seems to weaken the SHOULDs above.  Section 7.1 then goes
on to say:

   Efforts to make it more difficult for users to set envelope return
   path and header "From" fields to point to valid addresses other than
   their own are largely misguided: they frustrate legitimate
   applications in which mail is sent by one user on behalf of another
   or in which error (or normal) replies should be directed to a special
   address.

Oh well, so much for that.  Not meaning to be a pain in the ass, but
I sure rely on such identities (with Cyrus shared folders) bloody
heavily........ 

-- 
Amos