vm team mailing list archive

["Stephen J. Turnbull" <stephen@xxxxxxxxxx>]

Uday S Reddy writes:

 > You mean the VM package in XEmacs doesn't ship vcard.el?  Yes, it
 > appears so. But the public releases of VM 8.0.x always did.  You
 > can check here:
 > 
 >    http://www.very-clever.com/download/nongnu/viewmail/
 > 
 > It is possible that the upstream and downstream versions of VM
 > diverged over time.  It would be good to merge them back.

Well, if you're interested in doing (some of) the work, patches would
be welcomed.  I'm not sure how the merge would work administratively,
and to do the work in the XEmacs package we would need to find out
Robert Fenk's intentions.  If he plans to continue as maintainer, we'd
have to work with him at his pace unless somebody of known reliablity
wants to fork our version (the only candidate at present is Aidan
Kehoe, and I don't think he wants that).

Regarding the improvements needed:

The same person (Robert Fenk) was responsible for both versions; I
don't think it's very likely that they have diverged very much.
However, after the burst of activity that led to him taking over from
Kyle, VM development seems to have been almost totally absent,
upstream or downstream, except for a few contributions from Aidan.

In fact, it's almost trivial to keep XEmacs packages in sync unless
(a) a major refactoring is done, which might require revisions to the
install process, or (b) features not currently available in XEmacs or
its packages are used.  But the major refactoring was already done in
the run-up to v8, and the problems faced by VM developers seem to be
more features or APIs not available in GNU rather than vice-versa,
based on a rather thorough audit of vm-mime.el I just completed.

Unfortunately, I also discovered that much of vm-mime.el is junk.  Bad
style, weird incorrect algorithms, and RFC nonconformance are rampant.

Also, as Mike Kupfer points out, it is a very bad idea to distribute
multiple versions of a given library.  It's also really unfortunate
that so much elisp code is duplicative at the function level; I know
of at least 5 implementations of MIME for Emacs (tm, SEMI, WEMI,
gnu/emacs-mime, and vm-mime.el), and none of them qualify as excellent
code, although emacs-mime is very RFC-conformant (which is why I use
message-mode in preference to VM's native sendmail).

In the (very) long run I would like to see much of Gnus, VM, MH-E,
Wanderlust, MEW, etc gutted, and the best implementations of RFC 822,
the MIME RFCs, RFC 1036, etc moved into an mua-lib.  RMS has been
thinking along the same lines (cf. the long "RMail sucks/no it
doesn't" thread on emacs-devel a few months back).