savannah-hackers-public
[Top][All Lists]
Advanced

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

Re: [Savannah-hackers-public] Having trouble pushing


From: arnold
Subject: Re: [Savannah-hackers-public] Having trouble pushing
Date: Thu, 09 Jan 2025 01:13:15 -0700
User-agent: Heirloom mailx 12.5 7/5/10

Bob,

I can't really speak to what you've said. I only note that https://
from Savannah is quite slow, whereas if I https:// clone a Github
repo, it's quite fast.  I suspect that Github is what has set
my expectatations.

For day-to-day work, obviously, I use ssh.  But only a few people
in the world have ssh access to the gawk repo; the rest use https://
or git://.  I know that Github disabled git:// access some years
ago, I think out of security concerns --- so out of habit I always
use https:// for Git repos.

Hmmm...

> $ time git clone https://git.savannah.gnu.org/git/gawk.git

I don't think I knew about this option. I will update the gawk doc
with this URL.

Thanks,

Arnold

Bob Proulx <bob@proulx.com> wrote:

> arnold@skeeve.com wrote:
> > Things are better at the moment (it's ~ 2:30 AM east coast time).
> > But... Although an https clone no longer pegs my CPU at 100%, it still 
> > sucks:
> >
> > $ time git clone https://git.savannah.gnu.org/r/gawk.git
> > Cloning into 'gawk'...
> > Fetching objects: 61396, done.
> >
> > real        11m34.265s
> > user        0m39.351s
> > sys 0m6.825s
>
> There are two problems with the above.  One is that of course the
> https:// protocol has more overhead and must compete with all of the
> abuse agents that are always hammering on the web service.  It's
> probably never going to be as fast as using the ssh:// protocol.
>
> But two is that the above is using the raw web file access URL.  Why?
> That's never going to be as efficient as the git smart http protocol.
> That's intentionally choosing the worst possible source to clone.
>
> Example from my system in Colorado which should be a fair test going
> across the net from a distance using the git smart http protocol.
>
>     rwp@madness:/tmp/junk$ time git clone 
> https://git.savannah.gnu.org/git/gawk.git
>     Cloning into 'gawk'...
>     remote: Counting objects: 61396, done.
>     remote: Compressing objects: 100% (14478/14478), done.
>     remote: Total 61396 (delta 48459), reused 58396 (delta 46267)
>     Receiving objects: 100% (61396/61396), 66.18 MiB | 2.61 MiB/s, done.
>     Resolving deltas: 100% (48459/48459), done.
>
>     real    1m11.502s
>     user    0m41.194s
>     sys     0m0.970s
>
> I have sometimes thought that we should remove the /r/ raw access
> path.  But as you know breaking something that people are using upsets
> them because it breaks something they were working using.
>
> At one time I believe the /r/ path was needed for people behind
> restrictive firewalls and broken http proxies to be able access the
> repository regardless of their bad environments.  During the http://
> days.  But now with https:// use one can't proxy through and that
> already blocks bad corporate http proxies.  And bad firewalls already
> are unhappy that they can't see into encrypted https connections.
> Which makes this very odd niche access method pretty much obsolete.
>
> Bob



reply via email to

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