Re: Win32 GnuTLS DLL installer?

[Eli Zaretskii]

> From: Ted Zlatanov <address@hidden>
> Date: Thu, 21 Sep 2017 11:38:01 -0400
> 
> EZ> How do you define "outdated"?  GnuTLS developers maintain 3 branches
> EZ> in parallel, and release versions from all the 3 branches.  Which one
> EZ> should we follow, and does any N+1 release from that branch mean that
> EZ> the N release is considered "outdated" and should be replaced?
> 
> That's usually the platform's responsibility, but in this case it seems
> up to us. I would keep up with the current branch (whatever was
> installed already) at least, on the principle of least surprise.

So whenever the current branch sees a new release, someone would have
to port it to Windows and provide the binaries?

> EZ> Moreover, accommodating a new version of GnuTLS might mean changes to
> EZ> Emacs C and/or Lisp sources -- are we going to release patches to
> EZ> the core sources through ELPA or something?
> 
> Theoretically the current branch will not require that.

Theoretically, yes.  In practice, this can and did happen.

> I would just post a non-intrusive message to the user for now.

Not sure how a message could help with source-level changes.  People
who track the Git repository normally don't keep local patches.
People who use official releases don't even have the sources in most
cases.

> EZ> IOW, this sounds like a major undertaking on our part, and I'm sure
> EZ> more and more issues will pop up as we consider the implications.  I'm
> EZ> not sure we want to become a de-facto "distro" for MS-Windows users,
> EZ> as I don't think we have the resources, even if we have the desire.
> 
> Agreed, I wasn't suggesting all of that. It does seem like a general
> Emacs update process may be more appropriate, like Cygwin does, but
> simple messaging is a lot easier and less risky.

Again, I'm not sure I see how a message would help.  And Cygwin does
produce packages as GNU/Linux distribution do, something I don't think
we should take upon ourselves.