Re: Fixing Gmane's mangled MIDs

[Sven Joachim <svenjoac@gmx.de>]

On 2008-12-08 20:17 +0100, Reiner Steib wrote:

> On Sun, Dec 07 2008, Sven Joachim wrote:
>
>> On 2008-12-07 06:50 +0100, Lars Magne Ingebrigtsen wrote:
>>> When the same message comes from two mailing lists, and there's not
>>> enough information in them to identify them as such (to crosspost the
>>> message between the list), the Message-ID has to be renamed.  Remove the
>>> bit after __ to find the original ID.
>>
>> Here is some Emacs lisp code that does this automatically when replying
>> in Gnus:
>>
>> (defun sj-unmangle-gmane-message-id ()
>>   "Undo the Message-ID mangling that Gmane does on some postings."
>>   (save-excursion
>>     (save-restriction
>>       (message-narrow-to-headers)
>>       (goto-char (point-min))
>>       (while
>> 	  (re-search-forward "\\(<.*\\)" (point-max) t)

Aaargh, did I really write that?!  Emacs must have swallowed characters
in copy-and-paste, I swear that the line in my file looks

	  (re-search-forward "\\(<.*\\)" (point-max) t)

which makes a bit more sense and may sometimes even work (it did for me).

>> 	(replace-match "\\1")))))
>
> I would have expected a search for an expression like
> "__[0-9.]+\\$gmane\\$org@" or similar.  What am I missing?

At least the '.' and '$' characters that also occur in the mangled
Message-IDs, although this is fixed in your patch below.

> I'd guess the unmangling should be done when `message-id' (fixing
> References should not be necessary, I think) is put into
> `message-reply-headers' (in `message-cite-original-1',
> `message-reply', `message-followup'?).
>
> Here's a very rough, untested patch:
>
> Index: message.el
> ===================================================================
> RCS file: /usr/local/cvsroot/gnus/lisp/message.el,v
> retrieving revision 7.277
> diff -u -u -r7.277 message.el
> --- message.el	6 Dec 2008 17:30:03 -0000	7.277
> +++ message.el	8 Dec 2008 19:12:15 -0000
> @@ -3665,6 +3667,23 @@
>  	  (push (buffer-name buffer) buffers))))
>      (nreverse buffers)))
>  
> +(defcustom message-gmane-mangled-mid-regexp
> +  ;; FIXME: Don't use gnus-...
> +  (concat "\\(<" gnus-button-valid-localpart-regexp "\\)"
> +	  "__[0-9a-z.$]+\\$gmane\\$org" ;; bogus part
> +	  ;; message-valid-fqdn-regexp probably would be too strict here
> +	  "\\(@[a-z0-9][-.a-z0-9]+>\\)")
> +  "FIXME"
> +  :version "23.1" ;; No Gnus
> +  :group 'message-headers
> +  :type '(choice (const :tag "Don't unmangle MIDs" nil)
> +		 regexp))
> +
> +(defun message-fix-gmane-mids (mids)
> +  ;; FIXME: Don't use gnus-...
> +  (gnus-replace-in-string mids message-gmane-mangled-mid-regexp
> +			  "\\1\\2"))
> +
>  (defun message-cite-original-1 (strip-signature)
>    "Cite an original message.
>  If STRIP-SIGNATURE is non-nil, strips off the signature from the
> @@ -3694,7 +3713,11 @@
>  		      (or (message-fetch-field "subject") "none")
>  		      (or (message-fetch-field "from") "nobody")
>  		      (message-fetch-field "date")
> -		      (message-fetch-field "message-id" t)
> +		      (funcall
> +		       (if (stringp message-gmane-mangled-mid-regexp)
> +			   'message-fix-gmane-mids
> +			 'identity)
> +		       (message-fetch-field "message-id" t))
>  		      (message-fetch-field "references")
>  		      0 0 ""))))
>        (mml-quote-region start end)
> @@ -6622,7 +6658,11 @@
>  	  (save-excursion
>  	    (setq follow-to
>  		  (funcall message-wide-reply-to-function)))))
> -      (setq message-id (message-fetch-field "message-id" t)
> +      (setq message-id (funcall
> +			(if (stringp message-gmane-mangled-mid-regexp)
> +			    'message-fix-gmane-mids
> +			  'identity)
> +			(message-fetch-field "message-id" t))
>  	    references (message-fetch-field "references")
>  	    date (message-fetch-field "date")
>  	    from (or (message-fetch-field "from") "nobody")
> @@ -6683,7 +6723,11 @@
>  	    date (message-fetch-field "date")
>  	    subject (or (message-fetch-field "subject") "none")
>  	    references (message-fetch-field "references")
> -	    message-id (message-fetch-field "message-id" t)
> +	    message-id (funcall
> +			(if (stringp message-gmane-mangled-mid-regexp)
> +			    'message-fix-gmane-mids
> +			  'identity)
> +			(message-fetch-field "message-id" t))
>  	    followup-to (message-fetch-field "followup-to")
>  	    newsgroups (message-fetch-field "newsgroups")
>  	    posted-to (message-fetch-field "posted-to")
>
> Bye, Reiner.

I'll try this in the next days.

> [ Omitted gmane.dicuss from Mail-Followup-To as this is only relevant
>   for Gnus. ]

Cc'ed it anyway as people may read the thread there and even try the
nonsense I posted; at least I wanted to correct the regexp.

Sven