unofficial mirror of guix-devel@gnu.org 
 help / color / mirror / code / Atom feed
From: "Ludovic Courtès" <ludo@gnu.org>
To: Brendan Tildesley <btild@mailbox.org>
Cc: "guix-devel@gnu.org" <guix-devel@gnu.org>
Subject: Re: What’s next?
Date: Mon, 17 May 2021 22:25:19 +0200	[thread overview]
Message-ID: <87y2cdcdlc.fsf@gnu.org> (raw)
In-Reply-To: <1367816066.282875.1621167873186@office.mailbox.org> (Brendan Tildesley's message of "Sun, 16 May 2021 14:24:33 +0200 (CEST)")

Hi Brendan,

Brendan Tildesley <btild@mailbox.org> skribis:

> Since you asked I'll dump my nebulous wishes here. Sorry that I don't have very
> concrete suggestions and solutions, just things I think could be better.  I
> should also state that 99% of my thoughts about Guix are through the filter of
> "I want to build a Guix based desktop distribution that can be installed on
> library, school, and home computers and Just Work better than Microsoft Windows"

This is great.  I agree that polishing what already exists should remain
a major goal.

> 1. Improve guix pull and updating reliability.
> Updating guix for me is often like this:
> $ guix pull
> retrieving git objects....
> TLS error in stream blah blah blah
> $ guix pull
> retrieving git objects..34%......... *dies but doesn't return to the prompt*
> ctrl-c

I’ve literally never experienced this.  Could you file a bug or two to
bug-guix@gnu.org, with at least the command and all the output?

> $ guix pull
> $ guix upgrade
> $ sudo guix system reconfigure...
> GUI desktop crashes back to TTY mysteriously.

Likewise, please!

> Often downloads to source tarballs will download at ~30KiB/s. maybe some
> axel-like system where both upstream an mirror urls are used might be
> good.

Could you report as many details when that happens?

> Also this may be wrong but from memory I think if the hash is wrong on
> a source tarball, it will not bother trying to use the guix ci mirror
> of it, only if it's 404. So people with --no-substitutes can get
> completely stuck.

True: <https://issues.guix.gnu.org/28659>.

> 2. 'guix system reconfigure' without instantiating the new system until reboot.
> I think the fact that reconfigure changes the whole system while its running is
> a bit of a party trick, like pulling the table cloth out without any glasses
> falling over.  For a Desktop GNU/Linux it kinda just accidentally works much of
> the time. I've always thought I'd prefer the default to simply install the new
> configuration in to GRUB and tell the user to reboot or add some flag like
> --instatiate-now.

I’m satisfied with the current default, but we could have an option to
delay activation until reboot.  Likewise, please file as a “wishlist”
item to bug-guix@gnu.org.

> 3. CLI consistency, e.g.,
> $ guix package => nothing
> $ guix system =>
> guix system: missing command name
> Try 'guix system --help' for more information.
>
> $ guix package --list-generations, but
> $ guix system list-generations ???\
>
> Also https://issues.guix.gnu.org/47618

All good points!

> 4. Make guix simpler and more performant.
>
> Guix is complex. It has features that make it theoretically superior in many
> ways, but in practice hasn't reached the reliability of simpler systems.  Every
> aspect of Guix seems to take more time, cpu, ram, hd space than say pacman.
> It's a bit beyond my skills to figure out how to actually improve these things
> though.
>
> $ guix system foobar
>
> takes 1.5-3.0 seconds on SSD, stats 2374 files, with 345 No such file or
> directories in order to determine the command doesn't exist and do nothing.

Agreed.  There have been improvements over time, such as the ld.so cache
in ‘core-updates’ (not merged yet), but there’s still a long way.

It’s frustrating for all of us, but one way to help is by pinpointing
specific bottlenecks and gathering as much data as possible so we have a
good starting point for optimization work.

> A fast guix could have 'guix time-machine -- environment --ad-hoc emacs --
> emacs' run instantly after its been ran once, and some kind of 'guix run'
> command could be added.

Yeah.

> 5. I think Guix should have some simple system to only update to the latest
> commit where behemoths like webkit/chromium already have substitutes available
> and that should maybe be the default way to update. It's not a big deal if we
> are making people build stuff that only takes 30 seconds and uses minimal
> ram. Doesn't need to be perfect. Ricardo made some simple scripts that did stuff
> like that in the past.

Yes, this is being worked on (you may have seen
‘channel-with-substitutes-available’ in 1.3.0, which is a first step in
that direction.)

Thanks for sharing!  I guess my message here is that all these should be
broken down into individual bug reports/wishes that can be more easily
addressed.  :-)

Ludo’.


  reply	other threads:[~2021-05-17 20:25 UTC|newest]

Thread overview: 71+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-05-16 12:24 What’s next? Brendan Tildesley
2021-05-17 20:25 ` Ludovic Courtès [this message]
  -- strict thread matches above, loose matches on Subject: below --
2021-05-15 17:47 Ludovic Courtès
2021-05-15 18:08 ` Julien Lepiller
2021-05-18 19:30   ` Leo Famulari
2021-05-18 21:19     ` Julien Lepiller
2021-05-18 20:25   ` Ludovic Courtès
2021-05-19 15:39     ` Katherine Cox-Buday
2021-05-19 16:22       ` Ricardo Wurmus
2021-05-15 20:24 ` Efraim Flashner
2021-05-16 18:25   ` raingloom
2021-05-16 22:06     ` Joshua Branson
2021-05-17 20:13   ` Ludovic Courtès
2021-05-21 11:07     ` Efraim Flashner
2021-05-26 13:26       ` Ludovic Courtès
2021-05-16  4:09 ` Maxim Cournoyer
2021-05-16  8:57   ` Pierre Neidhardt
2021-05-16 18:18     ` Christopher Lemmer Webber
2021-05-17  5:43       ` Pierre Neidhardt
2021-05-16 13:38 ` Maxime Devos
2021-05-16 16:08 ` Vagrant Cascadian
2021-05-16 16:26 ` Svante Signell
2021-05-17 14:45   ` Leo Famulari
2021-05-18  2:35     ` Joshua Branson
2021-05-18 14:05       ` Leo Famulari
2021-05-22 23:11         ` raingloom
2021-05-24 22:32           ` Joshua Branson
2021-05-17 12:36 ` zimoun
2021-05-18  3:37 ` Bone Baboon
2021-05-18 13:08   ` Bone Baboon
2021-05-18 20:24   ` Ludovic Courtès
2021-05-18 16:44 ` Maxime Devos
2017-05-24 13:11 Ludovic Courtès
2017-05-24 13:23 ` Ricardo Wurmus
2017-05-27 10:01   ` Ludovic Courtès
2017-05-27 21:44     ` Ricardo Wurmus
2017-05-28 20:44       ` Ludovic Courtès
2017-05-28 21:36         ` Ricardo Wurmus
2017-05-30 15:55           ` Ludovic Courtès
2017-05-24 15:52 ` Brendan Tildesley
2017-05-27 10:04   ` Ludovic Courtès
2017-05-28 20:41     ` Maxim Cournoyer
2017-05-30 15:17       ` Ludovic Courtès
2017-06-03 21:16         ` Maxim Cournoyer
2017-05-24 16:09 ` Catonano
2017-05-24 16:25 ` Jan Nieuwenhuizen
2017-05-24 18:40   ` Adonay Felipe Nogueira
2017-05-24 19:34   ` Catonano
2017-05-24 19:56     ` Ricardo Wurmus
2017-05-30  0:09       ` myglc2
2017-05-24 21:47     ` Leo Famulari
2017-05-24 21:45   ` Leo Famulari
2017-05-25  8:11     ` What???s next? Pjotr Prins
2017-05-27 10:16       ` Ludovic Courtès
2017-05-28  7:30         ` What's next? Pjotr Prins
2017-05-28 20:48           ` Ludovic Courtès
2017-05-28 22:05             ` Roel Janssen
2017-05-30 15:19               ` Ludovic Courtès
2017-05-30 20:15                 ` Pjotr Prins
2017-05-29  2:31             ` Maxim Cournoyer
2017-05-28 20:37         ` What???s next? Maxim Cournoyer
2017-05-28 21:34           ` Ricardo Wurmus
2017-05-30 15:14           ` Ludovic Courtès
2017-05-25 14:57     ` What’s next? Chris Marusich
2017-05-25 18:32       ` Leo Famulari
2017-05-25 20:01       ` Ricardo Wurmus
2017-05-25 20:41         ` Adonay Felipe Nogueira
2017-05-27 10:13         ` Ludovic Courtès
2017-05-29 23:28           ` myglc2
2017-06-08 14:35           ` Ricardo Wurmus
2017-05-27 10:09   ` Ludovic Courtès

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

  List information: https://guix.gnu.org/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87y2cdcdlc.fsf@gnu.org \
    --to=ludo@gnu.org \
    --cc=btild@mailbox.org \
    --cc=guix-devel@gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this public inbox

	https://git.savannah.gnu.org/cgit/guix.git

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).