Re: GnuTLS updates proposal

[Ted Zlatanov]

On Sun, 26 May 2013 17:34:10 -0400 Ted Zlatanov <address@hidden> wrote: 

TZ> On Sat, 25 May 2013 03:10:36 -0400 Stefan Monnier <address@hidden> wrote: 
>>> Correction: they think that the current arrangement, whereby users are
>>> pointed to specific distributions of GnuTLS and other 3rd party
>>> packages, are adequate enough, and any additional effort is something
>>> the Emacs developers are unlikely to be able to sustain.

SM> I sure won't spend any time on this, indeed, and I hope you don't
SM> either.  But I guess if someone wants to provide some Elisp code that
SM> eases up the installation (taking care of downloading the right DLL
SM> (which I guess might depend on the Emacs version and even the specific
SM> build) and installing it at the right place), that would be OK and could
SM> very well fit in GNU ELPA (I don't think I'd want the DLL itself to be
SM> there).  Hopefully such code would handle not just gnutls but also the
SM> other libraries, like libxml.

TZ> Sounds good, I'll put it on my TODO list after ELPA package signing.
TZ> Thank you for looking at my proposal.

Update: I talked to the GnuTLS developers.  They now have W32 binaries
but do *not* guarantee anything about them except their internal tests
have passed, and don't consider W32 an important platform.  So the Emacs
W32 integration could fail and they will still release those DLLs.  But
at least the W32 build is considered somewhat important by them.

It seems that it's better to keep relying on Eli and myself as the
build+QA team for the Emacs side, and to keep using Eli's binaries, with
me as the fallback if he can't produce them.  So basically this does not
change the plans we discussed earlier, but we now have more assurance of
a working W32 baseline from the GnuTLS developers.

Ted