[Kde-pim] Problems with infrastructure

Jeff Mitchell mitchell at kde.org
Tue Dec 16 22:11:29 GMT 2014

Don't want to weigh in on Gerrit as I don't know it well enough, but as 
for Phabricator, Ben may have forgotten but we did evaluate it a while 
back. It was neat but had a very serious problem: you needed an account 
to even view anything (no public access), and once you got into it 
everything was completely open. There was no real way not to give 
everyone the keys to the kingdom. This was getting tracked in 
https://secure.phabricator.com/T603 which is now marked as resolved. So 
it may be a viable candidate again. But it's definitely a very 
opinionated way of doing things.

To be honest, I actually think that this discussion needs to start with 
a more fundamental question than the technical specifics of the various 
tools, because at this point we probably have several potential tools 
that can meet our needs. So the question I think is fundamental is: what 
is our goal in a tool?

- If the goal is to make it comfortable to those that are used to and 
like using GitHub, we should be looking at Gogs (I'm comfortable with Go 
so could even code in some missing features, potentially).

- If the goal is to put code review front-and-center, mandatory even, we 
should look at Gerrit.

- If the goal is to have a Phabricator-style hub (not sure what else to 
call it), we should look at Phabricator.

I don't know the best way to choose from that list -- and this is part 
of the reason that trials stalled after Phabricator and GitLab. As far 
as I'm concerned, we can put it to a poll or vote, although I think 
capturing the general sentiment of the KDE community rather than a 
smaller subset of interested people is likely to be difficult. I ran the 
trials of Phabricator and GitLab in the past and I'm happy to do trials 
of Gogs and Phabricator now (Gerrit obviously doesn't need this as it's 
currently being trialled).

Regardless, any move to a different tool will require compromises, and 
will make some users happy and others unhappy. But it's not reasonable 
to expect the sysadmins to support multiple parallel systems, and I 
really do think that until we figure out what we actually want, and by 
we I mean the general developer community, it's somewhat wasted time -- 
especially because until we know what our ideal end goal is, we won't 
know if the compromises we need to make in pursuit of that end goal are 
worth making, which makes it easy to keep discarding solutions that 
don't quite measure up.


More information about the kde-core-devel mailing list