emacs-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] GnuTLS support on Woe32


From: Lars Magne Ingebrigtsen
Subject: Re: [PATCH] GnuTLS support on Woe32
Date: Tue, 08 Mar 2011 10:14:04 +0100
User-agent: Gnus/5.110014 (No Gnus v0.14) Emacs/24.0.50 (gnu/linux)

Eli Zaretskii <address@hidden> writes:

> It should probably simply be a defcustom.  I don't see how ``higher
> levels of code'' could ever DTRT in this respect.  If they can know
> something about that, so can gnutls.el.

Well, there are several issues here.

1) gnutls is still not stable, so it should not be used by people who
are developers for now.  The way to turn on gnutls now is to say
(require 'gnutls), which I think is fine.

2) Virtually all network protocols that exist now support STARTTLS.
Littering every connect function with this code (which can be rather
subtle in when to use it or not (you may require that the server support
it, or you want to bail out, or you may want to use opportunistically if
the server does support it, but only if it's "free" to do so (i.e.,
gnutls is compiled in), or you may want to always have it be used
opportunistically, even though it's significantly slower using
starttls.el (i.e. using the external gnutls-cli program)).

(That was a long sentence.)

This functionality is provided by `open-protocol-stream', and the only
option the user has to consider is really
`proto-stream-always-use-starttls'.

If the basic starttls.el library also starts doing various decisions
here, it gets even more confusing.

So I'd prefer that starttls.el does one thing, and one thing only: Do
the grufty stuff to talk to gnutls-cli.  For now, at least.

This can be revisited later when the built-in gnutls stuff actually
works reliably.

-- 
(domestic pets only, the antidote for overdose, milk.)
  address@hidden * Lars Magne Ingebrigtsen




reply via email to

[Prev in Thread] Current Thread [Next in Thread]