Re: master a37eba8: Speed up initialization by preferring /dev/urandom to GnuTLS

Re: master a37eba8: Speed up initialization by preferring /dev/urandom to GnuTLS

[Lars Ingebrigtsen]

dancol@dancol.org (Daniel Colascione) writes:

> +  /* If that didn't work, try using GnuTLS, which is secure, but on
> +     some systems, can be somewhat slow.  */

Did anybody make an attempt at determining why this is slow on the
reported system?  If calling that function is slow on that system, then
presumably that means that opening a TLS connection will now be
inexplicably slow the first time, right?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: master a37eba8: Speed up initialization by preferring /dev/urandom to GnuTLS

[Daniel Colascione]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> dancol@dancol.org (Daniel Colascione) writes:
>
>> +  /* If that didn't work, try using GnuTLS, which is secure, but on
>> +     some systems, can be somewhat slow.  */
>
> Did anybody make an attempt at determining why this is slow on the
> reported system?  If calling that function is slow on that system, then
> presumably that means that opening a TLS connection will now be
> inexplicably slow the first time, right?

The system isn't managed by me and has some custom security certificate
crud that I didn't feel like digging into. You're right about TLS
connection delays, but I'm more willing to tolerate the delay in
connecting over a network (which is slow anyway) than I am in
process startup.

Re: master a37eba8: Speed up initialization by preferring /dev/urandom to GnuTLS

[Eli Zaretskii]

> From: Daniel Colascione <dancol@dancol.org>
> Date: Sat, 29 Oct 2016 23:51:21 -0700
> Cc: emacs-devel@gnu.org
> 
> Lars Ingebrigtsen <larsi@gnus.org> writes:
> 
> > dancol@dancol.org (Daniel Colascione) writes:
> >
> >> +  /* If that didn't work, try using GnuTLS, which is secure, but on
> >> +     some systems, can be somewhat slow.  */
> >
> > Did anybody make an attempt at determining why this is slow on the
> > reported system?  If calling that function is slow on that system, then
> > presumably that means that opening a TLS connection will now be
> > inexplicably slow the first time, right?
> 
> The system isn't managed by me and has some custom security certificate
> crud that I didn't feel like digging into. You're right about TLS
> connection delays, but I'm more willing to tolerate the delay in
> connecting over a network (which is slow anyway) than I am in
> process startup.

I think the decision was to install this on emacs-25, so please
cherry-pick from master.

Thanks.