Hi,
my thoughts:
- The switch from bzr to git was supposed to potentially attract new contributors because git is more widely used than bazaar. Even if it's true, I don't think people interested in contributing to inkscape will have any problem finding the code. Inkscape is an insanely popular and known project, and whatever option we choose, the code will never be further away than a google search for "inkscape official code repository"*. Many projects are on code repos way less known than gitlab, for instance git.gnome.org for gimp, and that will not prevent any potential developper to find in 15 seconds she will have to type "git clone git://git.gnome.org/gimp" to get started. So the "popularity" of github vs gitlab should, imo, not matter much.
- I think we are right to pride ourselves for being FLOSS. In my opinion, advocating in favor of open source, being part of SFC, etc, means I would always favor using a FLOSS tool over a not-opensourced one. So, philosophically speaking, Gitlab is a clear winner (but I'm not religious about it, I won't stop contributing if we decide on github).
- Also, I do think the "big fish in a small pond" would actually be an argument *in favor* of gitlab. It means that we might have more possibilities of interactions with the gitlab people, more opportunities to talk with them about our needs and what works great and works less great, while if we find any quirks with github, we'll probably have to stay with "that's how it works". That may also help gitlab itself improve. So I agree we might find some things less "polished" than github, but (1) we can do something about it, and/or (2) we probably can talk to the people taking care of it.
- Whatever the choice, migrating after a while if it ends being a bad choice (say, if we go to gitlab, but it timeouts all the time and no one cares) will be easy.
- The big plus of github afaiu is/was the Travis CI, but if gitlab has it too now ... maybe we should try it ?
- "Git is a *D*VCS, there is no real need to keep only one upstream copy." => I'm a bit curious about how it would work if there are CI stuff involved... can github reject a commit while testing it on travis, even if the commit was originally pushed to lp and gitlab's tests accepted it ?