message-user-mail-address and other addresses

message-user-mail-address and other addresses

[Hallvard B Furuseth]

(message-user-mail-address) in message.el seems to think that
user-mail-address can contain more than just the address, such as
"name <address>".  This is wrong, and if we teach users to put "name
<address>" in user-mail-address, packages that expect a plain address
will fail.  Examples: sendmail.el, add-log.el, rmailsum.el.

But if that's the idea, (message-user-mail-address) does not work
correctly anyway:
	(let ((user-mail-address "Hallvard Furuz <hbf@usit.uio.no>"))
	 	 (message-user-mail-address))
	==> "<hbf@usit.uio.no>"
oops, should remove the <>'s.  This works, however:
	(let ((user-mail-address "hbf@usit.uio.no (Hallvard Furuz)"))
	 	 (message-user-mail-address))
	==> "hbf@usit.uio.no"

Also, it does not allow for "foo!bar!baz" type addresses.

BTW, I haven't checked but I hope all your mail-address code allows for
addreses like these (real-life) addresses:
  @bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su
  /S=ZHAOQING.LTD/O=SERVICE/PRMD=JMPTT/ADMD=CHINAMAIL/C=CN/@x400.bta.net.cn
  sevcik@[147.231.40.1]
  Egon_Hagen_at_ROGALAND__A-SENTER@cclink.rf.no
  idgst1+@pitt.edu		(note: end of localpart is not end of word)
Some real-life idiot addresses:
  "gry oftedal"@nlhstud.nlh.no	(stupid, but some people do that)
  ingrid.haug.@nlhstud.nlh.no
  Joern..Nicolaisen@hinesna.no

Huh. Space in localpart is actually "sensible" in one case:
  "/I=hb/S=furuseth/OU=usit/O=uio/PRMD=uninett/ADMD= /c=NO/"@mons.uio.no
i.e. mail sent to an X.400 -> rfc822 gateway.  X.400 addresses, in
particular their ADMD parts, sometimes contain space.

Oh, and here is one more.  Is this address nuts or am I missing a
message format?  It's the envelope (`From ') address of a message to
directory@uninett.no:
   hsba1@dagobert#u#br.ccmail.compuserve.com


Regards,

Hallvard

Re: message-user-mail-address and other addresses

[Lars Magne Ingebrigtsen]

Hallvard B Furuseth <h.b.furuseth@usit.uio.no> writes:

> (message-user-mail-address) in message.el seems to think that
> user-mail-address can contain more than just the address, such as
> "name <address>".  This is wrong, and if we teach users to put "name
> <address>" in user-mail-address, packages that expect a plain address
> will fail.  Examples: sendmail.el, add-log.el, rmailsum.el.

That's true.  However, many (many) people, especially people mostly
using XEmacs, it seems, have a `user-mail-address' that contains more
than just the address.  I don't normally think one should cater to
brokenness, but in this instance there were just too many people that
experienced problems.

> But if that's the idea, (message-user-mail-address) does not work
> correctly anyway:
> 	(let ((user-mail-address "Hallvard Furuz <hbf@usit.uio.no>"))
> 	 	 (message-user-mail-address))
> 	==> "<hbf@usit.uio.no>"

Yup.  0.74 will use `mail-extract-address-components' to extract the
address parts.  I think it does a good job.  I tried it on all the
examples you gave, and it failed only on one:

>   @bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su

(nth 1 (mail-extract-address-components
"@bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su"))
 =>
"aleck@optoel.urgu.e-burg.su"

Perhaps one should send a bug report to the mail-extr people?

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

Re: message-user-mail-address and other addresses

[Hallvard B Furuseth]

> Yup.  0.74 will use `mail-extract-address-components' to extract the
> address parts.  I think it does a good job.

Great!  And sgnus itself no longer contains any regexps that will fail
if there are, say, spaces in an address?

> I tried it on all the examples you gave, and it failed only on one:
> 
>>   @bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su
> 
> (nth 1 (mail-extract-address-components
> "@bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su"))
>  =>
> "aleck@optoel.urgu.e-burg.su"

No, I think that's correct.  The @foo: stuff appears to be routing
information on the `From ' line.


BTW, about (user-mail-address): Erik Naggum thought the best way was
	(defun user-mail-address (...)
	  (or (and user-mail-address-function
		   (eq user-mail-address user-mail-address-default)
		   (apply user-mail-address-function ...))
	      user-mail-address))
where user-mail-address-function is *not* encouraged to set
user-mail-address, but instead does it own bookkeeping.


> However, many (many) people, especially people mostly
> using XEmacs, it seems, have a `user-mail-address' that contains more
> than just the address.  I don't normally think one should cater to
> brokenness, but in this instance there were just too many people that
> experienced problems.

Hum.  Well, if we are going to un-break these as far as possible --
How do they write those addresses?  Is there some "semi-standard" for
XEmacs on this?  If it's
	"one@address, another@address",
then
	From: Joe Luser <one@address, another@address>
won't work too well...
Maybe we should fix add-log & such things instead...


Regards,

Hallvard

Re: message-user-mail-address and other addresses

[Lars Magne Ingebrigtsen]

Hallvard B Furuseth <h.b.furuseth@usit.uio.no> writes:

> And sgnus itself no longer contains any regexps that will fail if
> there are, say, spaces in an address?

It seemed to respond correctly to addresses with spaces.  I took a
peek at `mail-extr', and it said they had over 100K test cases, so it
would seem that it's quite thoroughly tested.  (It's written by JWZ,
by the way.)

The lesson I've learned from this:  "Don't *ever* try to come up with
my own regexps for mail header recognition.  Someone else will have
done the job properly for me."

-- 
  "Yes.  The journey through the human heart 
     would have to wait until some other time."

SORRY (Was: Re: message-user-mail-address and other addresses)

[Hallvard B Furuseth]

Sorry, skip the section below.  I'm going to bed:-(

I wrote:
>> However, many (many) people, especially people mostly
>> using XEmacs, it seems, have a `user-mail-address' that contains more
>> than just the address.  I don't normally think one should cater to
>> brokenness, but in this instance there were just too many people that
>> experienced problems.
>
>Hum.  Well, if we are going to un-break these as far as possible --
>How do they write those addresses?  Is there some "semi-standard" for
>XEmacs on this?  If it's
>	"one@address, another@address",
>then
>	From: Joe Luser <one@address, another@address>
>won't work too well...
>Maybe we should fix add-log & such things instead...


Regards,

Hallvard

Re: message-user-mail-address and other addresses

[Jason L Tibbitts III]

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

LMI> Yup.  0.74 will use `mail-extract-address-components' to extract the
LMI> address parts.  I think it does a good job.  I tried it on all the
LMI> examples you gave, and it failed only on one:

>> @bsd.urgu.e-burg.su,@optoel.uucp:aleck@optoel.urgu.e-burg.su

RFC822 only permits source-routing inside angle brackets.  I don't think
this address is legal.  Also, I don't think the path components are
separated by commas.  The `@' symbol is the separator.

from RFC822:

route       =  1#("@" domain) ":"           ; path-relative
route-addr  =  "<" [route] addr-spec ">"

 - J<

Re: message-user-mail-address and other addresses

[Hallvard B Furuseth]

> RFC822 only permits source-routing inside angle brackets.  I don't think
> this address is legal.

Well, it's legal inside angle brackets then:-)

> Also, I don't think the path components are
> separated by commas.  The `@' symbol is the separator.
> 
> from RFC822:
> 
> route       =  1#("@" domain) ":"           ; path-relative
> route-addr  =  "<" [route] addr-spec ">"

Huh.  RFC821 uses routes with comma:

                  TO:<@HOSTA.ARPA,@HOSTB.ARPA:USERC@HOSTD.ARPA>

Umm - Does anybody remember which RFC lists recommendations for various
services, including errors in previous RFCs?


Regards,

Hallvard

Re: message-user-mail-address and other addresses

[Jason L Tibbitts III]

>>>>> "JLT" == Jason L Tibbitts <tibbs@uh.edu> writes:

JLT> Also, I don't think the path components are separated by commas.  The
JLT> `@' symbol is the separator.

Doh.  That's patently wrong.  The rest is, I believe, correct.

 - J<

Re: message-user-mail-address and other addresses

[Steven L Baur]

>>>>> "lmi" == Lars Magne Ingebrigtsen <larsi@ifi.uio.no> writes:

lmi> Perhaps one should send a bug report to the mail-extr people?

;;; mail-extr.el --- extract full name and address from RFC 822 mail header.

;; Copyright (C) 1991, 1992, 1993, 1994 Free Software Foundation, Inc.

;; Author: Joe Wells <jbw@cs.bu.edu>
;; Maintainer: Jamie Zawinski <jwz@lucid.com>
               ^^^^^ ^^^^^^^^      ^^^^^^^^^
;; Version: 1.8
;; Keywords: mail

;; This file is part of XEmacs.

Has somebody who can work on it taken over?
-- 
steve@miranova.com baur
Unsolicited commercial e-mail will be proofread for $250/hour.
Andrea Seastrand: For your vote on the Telecom bill, I will vote for anyone
except you in November.