Merge package repositories
Merge request reports
Activity
added 2 commits
- Resolved by Levente Polyak
added 5 commits
-
1a3cc30c...69bda815 - 4 commits from branch
master
- 30b6ee12 - Add RFC about merging package repositories
-
1a3cc30c...69bda815 - 4 commits from branch
- Resolved by Levente Polyak
- Resolved by Levente Polyak
- Resolved by Sven-Hendrik Haase
I am in favour of giving "Trusted Users" (or whatever they are called now!) access to [extra], but much less in favour of merging [extra] and [community]. My reasons:
-
While it is clear the distinction between [community] and [extra] is blurry, there is still a distinction. [extra] contains a lot of while I will call mid-level system packages - X, wayland, major desktops, major programming languages. It is not perfect, but there is a distinction - I can have a semi useful desktop system without [community], but not without [extra]. As all (or at least the vast majority) of new packagers start as TUs, the damage from inexperience at packaging is still limited. Much like [core] is limited to people with extensive packaging experience, I think [extra] should require some experience too.
-
The [community] db is by far the largest when I do a repo update. Merging it with [extra] will make it even larger, and be by far the slowest to download. The joys of parallel downloads will be lost and the people will be sad.
Can I propose this proposal becomes more limited? My suggestion is this proposal becomes "Allow TUs access to the [extra] repository after 6 months". (The timeframe is arbitrary, but I think reasonable). This would achieve increasing our packager coverage, without the downsides I list above.
As a second stage, I propose we consider our repo layout. I know what such a discussion would entail, and see it being a lack of fun, but I am volunteering to lead this... Properly defining (and renaming?) [extra], and splitting [community] that becomes larger from the overflow. e.g. a [haskell] repo would remove 1000+ packages from the database that a large proportion of users do not use. That would also allow recruitment of packagers against specialist repos more easily (e.g. Felix could pick people for [haskell] with limited wider input), reducing the burden to becoming an Arch Linux Packager.
In summary, I think the intended outcome is good, but the approach is a step in the wrong direction.
-
- Resolved by Levente Polyak
As additional evidence of the still real split between [extra] and [community], I currently have the following numbers of packages installed:
[core] = 178/260 = 68% [extra] = 623/3033 = 20% [community] = 136/9390 = 1%
In addition, a quick grab from pkgstats indicates the median installation percent for packages in [community] is much lower than those in [extra]. There are clear exceptions.
I think the split needs to stay.
- Resolved by Levente Polyak
I agree with Allan at this point. [core] and [extra] are the main parts the distro is about. I don't see a gain in allowing pushing to [testing] only as benefit.
- Resolved by Levente Polyak
- Resolved by Levente Polyak
I'd like to suggest keeping
[core]
and[community]
instead, similar to alpine's[main]
and[community]
.
- Resolved by Levente Polyak
Just a thought about proprietary applications and/or binary packages that are in the repos (e.g. Steam, Discord, Opera, ...) - this might be an opportunity to move those to their own "non-free" or "partner" repo. I'm not sure whether anyone really cares about the difference, but then again it might be nice to know that the "main" repos are solely built from sources.
- Resolved by Levente Polyak
I'm a bit worried about the size of the resulting repository. Didn't this cause us issues before?
mentioned in commit 0c87cd1e
mentioned in issue infrastructure#135 (closed)
mentioned in work item dbscripts#28 (closed)
mentioned in issue dbscripts#19 (closed)
mentioned in issue devtools#95 (closed)
mentioned in issue infrastructure#478 (closed)
mentioned in work item infrastructure#481 (closed)
mentioned in work item infrastructure#483 (closed)
mentioned in commit dbscripts@c723a955
mentioned in commit dbscripts@5282fa3e
mentioned in commit dbscripts@9ea65153