message.el: mail-host-address ignored

message.el: mail-host-address ignored

[Raymond Scholz]

Hi!

There seems to be an inconsistency in the behaviour of message.el
compared with it's documentation, coming with PGnus 0.98 .

My computer's name is "mde1". Some Linux distributions force the user
to specify a domain part too [1]. I used "home", so the complete name
is "mde1.home", no FQDN as you see. `system-name' gives me that string
too. So I set `mail-host-address' to a FQDN after consulting the
Message info pages and finding:

| `Message-ID'
|     This required header will be generated by Message.  A unique ID
|     will be created based on the date, time, user name and system
|     name.  Message will use `mail-host-address' as the fully qualified
|     domain name (FQDN) of the machine if that variable is defined.  If
|     not, it will use `system-name', which doesn't report a FQDN on
|     some machines - notably Suns.

But Message still uses `system-name' for Message-ID generation. My
tiny knowledge of Lisp tells me, that Message takes `system-name',
looks for a `.' in it and calls it fully qualified after that. If
there is no `.' in it, it uses other alternatives (including
`mail-host-address') to build a unique Message-ID.

My current workaround is to set system-name to "mde1" by now...

Cheers,
  Ray

[1] Yes, I know how to work around this, but others possibly won't.

-- 
Raymond Scholz -*- rscholz@tzi.de -*- http://www.tzi.de/~rscholz/
FB 3 Informatik, Universitaet Bremen -*- PGP key available via WWW

Re: message.el: mail-host-address ignored

[Rene H. Larsen]

[-- Attachment #1: Type: text/plain, Size: 957 bytes --]

Raymond Scholz <rscholz@tzi.de> writes:

> | `Message-ID'
> |     This required header will be generated by Message.  A unique
> |     ID will be created based on the date, time, user name and
> |     system name.  Message will use `mail-host-address' as the
> |     fully qualified domain name (FQDN) of the machine if that
> |     variable is defined.  If not, it will use `system-name', which
> |     doesn't report a FQDN on some machines - notably Suns.
> 
> But Message still uses `system-name' for Message-ID generation. My
> tiny knowledge of Lisp tells me, that Message takes `system-name',
> looks for a `.' in it and calls it fully qualified after that. If
> there is no `.' in it, it uses other alternatives (including
> `mail-host-address') to build a unique Message-ID.

This came up on the list about 2 months ago, but nothing was done
about it.  This patch against current CVS will fix the message-id
generation to match the documentation:


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Fix for message-id generation --]
[-- Type: text/x-patch, Size: 755 bytes --]

--- message.el.orig	Sat Nov  6 04:15:15 1999
+++ message.el	Thu Nov 11 02:25:28 1999
@@ -2939,14 +2939,15 @@
   (let ((system-name (system-name))
 	(user-mail (message-user-mail-address)))
     (cond
-     ((string-match "[^.]\\.[^.]" system-name)
-      ;; `system-name' returned the right result.
-      system-name)
      ;; Try `mail-host-address'.
      ((and (boundp 'mail-host-address)
 	   (stringp mail-host-address)
 	   (string-match "\\." mail-host-address))
       mail-host-address)
+     ;; Else try `system-name'
+     ((string-match "[^.]\\.[^.]" system-name)
+      ;; `system-name' returned the right result.
+      system-name)
      ;; We try `user-mail-address' as a backup.
      ((and user-mail
 	   (string-match "\\." user-mail)

[-- Attachment #3: Type: text/plain, Size: 166 bytes --]

-- 
Echelon bait:
AK-47 Janet Reno DES class struggle jihad NSA CIA Treasury
counter-intelligence David John Oates North Korea domestic disruption
Ortega Serbian PLO

Re: message.el: mail-host-address ignored

[Toby Speight]

Rene> Rene H. Larsen <URL:mailto:rhl@traceroute.dk>

0> In article <87n1slizjk.fsf@worldonline.dk>, Rene wrote:

Rene> This came up on the list about 2 months ago, but nothing was done
Rene> about it.  This patch against current CVS will fix the message-id
Rene> generation to match the documentation:
Rene>
Rene> --- message.el.orig	Sat Nov  6 04:15:15 1999
Rene> +++ message.el	Thu Nov 11 02:25:28 1999
Rene> @@ -2939,14 +2939,15 @@
Rene>    (let ((system-name (system-name))
Rene>  	(user-mail (message-user-mail-address)))
Rene>      (cond
Rene> -     ((string-match "[^.]\\.[^.]" system-name)
Rene> -      ;; `system-name' returned the right result.
Rene> -      system-name)
Rene>       ;; Try `mail-host-address'.
Rene>       ((and (boundp 'mail-host-address)
Rene>  	   (stringp mail-host-address)
Rene>  	   (string-match "\\." mail-host-address))
Rene>        mail-host-address)
Rene> +     ;; Else try `system-name'
Rene> +     ((string-match "[^.]\\.[^.]" system-name)
Rene> +      ;; `system-name' returned the right result.
Rene> +      system-name)
Rene>       ;; We try `user-mail-address' as a backup.
Rene>       ((and user-mail
Rene>  	   (string-match "\\." user-mail)

But doesn't this mean that on a network of machines that share a
mail-host-address (quite common IME), that there's a much greater
chance of ID collision?  Don't use mail-host-address for what it's not
intended for - it's okay by me as a fallback, but as you have it,
there's no way to use it for its intended purpose without getting the
undesirable message-id behaviour.  Create a new variable instead, if
you must (message-id-host).

What would be even better is if someone defined a "Message-ID by UDP
or TCP" service one could connect to for unique IDs.

BTW, is there a way to follow-up a message keeping the original MIME
structure?  The labelling of the part containing a patch (above) has
been lost, and I would have liked to keep it.

Re: message.el: mail-host-address ignored

[Lars Magne Ingebrigtsen]

"Rene H. Larsen" <rhl@traceroute.dk> writes:

> This came up on the list about 2 months ago, but nothing was done
> about it.  This patch against current CVS will fix the message-id
> generation to match the documentation:

I've fixed the documentation to match the code instead for the reasons 
Toby gave.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: message.el: mail-host-address ignored

[Rene H. Larsen]

[-- Attachment #1: Type: text/plain, Size: 1326 bytes --]

Toby Speight <Toby.Speight@streapadair.freeserve.co.uk> writes:

> Rene> Rene H. Larsen <URL:mailto:rhl@traceroute.dk>
> 
> 0> In article <87n1slizjk.fsf@worldonline.dk>, Rene wrote:
> 
> Rene> This came up on the list about 2 months ago, but nothing was done
> Rene> about it.  This patch against current CVS will fix the message-id
> Rene> generation to match the documentation:
> Rene>
> Rene> --- message.el.orig	Sat Nov  6 04:15:15 1999
> Rene> +++ message.el	Thu Nov 11 02:25:28 1999

[...]

> But doesn't this mean that on a network of machines that share a
> mail-host-address (quite common IME), that there's a much greater
> chance of ID collision?

Oh, I hadn't thought of this.  I was writing this from the perspective
of a user on a stand-alone machine.

> Don't use mail-host-address for what it's not intended for - it's
> okay by me as a fallback, but as you have it, there's no way to use
> it for its intended purpose without getting the undesirable
> message-id behaviour.

What exactly is the intended purpose? The doc-string simply says:

  *Name of this machine, for purposes of naming users.

Oh, I see from the code in message.el that this is used to create the
email address if user-mail-address is not set.

> Create a new variable instead, if you must (message-id-host).

Ok, (small) patch follows:

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Better fix for message-make-fqdn --]
[-- Type: text/x-patch, Size: 547 bytes --]

--- message.el.orig	Sat Nov  6 04:15:15 1999
+++ message.el	Thu Nov 11 19:39:26 1999
@@ -2939,6 +2939,12 @@
   (let ((system-name (system-name))
 	(user-mail (message-user-mail-address)))
     (cond
+     ;; First check if `message-id-host' is set and valid.
+     ((and (boundp 'message-id-host)
+	   (stringp message-id-host)
+	   (string-match "\\." message-id-host))
+      message-id-host)
+     ;; Then try `system-name'.
      ((string-match "[^.]\\.[^.]" system-name)
       ;; `system-name' returned the right result.
       system-name)

[-- Attachment #3: Type: text/plain, Size: 267 bytes --]


Note that this does not add documentation nor customize support.  I
don't really grok the elisp side of customize yet.
-- 
Echelon bait:
cracking COSCO genetic Marxist clones NORAD Peking Area 51 CIA David
John Oates Kennedy NWO smuggle Serbian counter-intelligence

Re: message.el: mail-host-address ignored

[Raymond Scholz]

Toby Speight <Toby.Speight@streapadair.freeserve.co.uk> writes:

> What would be even better is if someone defined a "Message-ID by UDP
> or TCP" service one could connect to for unique IDs.

I use `mail-host-address' for kinda such thing. If I dial up at my
ISP, I get a dynamic IP address. After that I use a tools from DynDNS
(http://www.dyndns.org) to point rscholz.dyndns.org on my dynamic
address. So the message IDs are RFC-conform (or whatever). The DynDNS
service is free, BTW.

Cheers,
  Ray
-- 
Raymond Scholz -*- rscholz@tzi.de -*- http://www.tzi.de/~rscholz/
FB 3 Informatik, Universitaet Bremen -*- PGP key available via WWW

Re: message.el: mail-host-address ignored

[Toby Speight]

Raymond> Raymond Scholz <URL:mailto:rscholz@tzi.de>

0> In article <m3hfisu5jf.fsf@rscholz.dyndns.org>, Raymond wrote:

Raymond> Toby Speight <Toby.Speight@streapadair.freeserve.co.uk> writes:

>> What would be even better is if someone defined a "Message-ID by
>> UDP or TCP" service one could connect to for unique IDs.

Raymond> I use `mail-host-address' for kinda such thing.  If I dial
Raymond> up at my ISP, I get a dynamic IP address.  After that I
Raymond> use a tools from DynDNS (http://www.dyndns.org) to point
Raymond> rscholz.dyndns.org on my dynamic address. So the message
Raymond> IDs are RFC-conform (or whatever).

I don't think we're talking about the same thing.  In my Ideal
Universe(TM), I'd connect to the message-id port of some host
(perhaps localhost), and it would reply with a valid message-id.
A bit like how the chargen service works.  Then Gnus and all the
other systems out there using message-ids wouldn't have to each
implement their own generators, and there'd be less chance of
collisions.  But I think that now is too late to put the genie
back into the bottle.