[kde-community] Future Git Plans

[Jeff Mitchell]

(Sending from the proper account this time)

Hello KDE Community,

Several years ago when transitioning to Git the sysadmins evaluated 
several possible options, including GitLab, Gitorious, Gitolite, and 
Gerrit. At the time, GitLab was quite immature in terms of code, 
community, and documentation. After evaluating options, we chose 
Gitolite, and built a suite of other services and helpers on top of it.

Gitorious has come along since being purchased by Powow AS, but still 
lags behind other solutions. Gerrit is still geared towards very 
specific code-review workflows and has an interface that is very 
difficult to grasp for more novice users. Gitolite is absolutely 
wonderful software, but it does one specific thing; we've set up a large 
number of other scripts, web servers, and web applications to address 
the things it doesn't do but that our community needs.

However, in the intervening years, GitLab (https://www.gitlab.com/) has 
advanced tremendously, due in part to the thousands of eyeballs it's had 
on its source code while gaining major footholds into the Git hosting 
market. It has gained documentation, features, an API, and a great deal 
of stability and maturity to its code base.

Due to its feature set, GitLab alone could take the place of at least 
projects.kde.org, commits.kde.org, quickgit.kde.org, and -- due to the 
built-in merge request workflow -- reviewboard.kde.org, drastically 
easing management and maintenance burden for the sysadmins. If the 
built-in wiki and issue tracking capabilities are enabled (which can be 
managed per-project), then projects (especially self-contained ones, 
such as Extragear projects) that desire a highly integrated workflow 
could migrate those functions to GitLab as well (note that this is not a 
statement indicating that we are planning to ditch Bugzilla any time 
soon!).

Replacing Gitolite with GitLab would have a number of benefits from the 
user perspective too, versus having all of those features spread across 
multiple different sites. I strongly encourage interested parties to go 
to http://demo.gitlab.com/ and browse around. (One benefit that is not 
immediately apparent is support for HTTPS as a protocol for pushing 
commits).

This replacement could require some minor changes to workflows based on 
the models that GitLab vs. Gitolite supports. However, the sysadmins 
feel that the benefits outweigh the burden of transition. The concrete 
and specific difference that is likely to cause the most pain is that 
personal forks and private repositories are handled very differently 
from Gitolite; during the transition period, we would likely require 
users to themselves migrate any clone/scratch repositories to the new 
system that they would like to keep. A benefit to this is that a 
majority of the clone and scratch repositories are currently forgotten 
and unused, and this would act as a spring cleaning.

This email serves two purposes: one, to inform the community of the 
direction we would like to go with KDE's Git hosting and request 
feedback; two, to ask for volunteer projects that are willing to act as 
crash test dummies for the new system, helping us figure out the best 
way to set it up, work out kinks, etc. Due to the bleeding-edge nature, 
we're currently limiting this to self-contained projects, such as those 
in Extragear.

Thanks,
Jeff