On Wed, Oct 21, 2015 at 6:52 AM, Sylvain Thénault
Hi,
at Logilab, we've been involved for years with mercurial development and we fight daily against the centralized model offered by github, among others. The latter explains why we've self-hosted pylint for such a long time even if the UI didn't reach the level of other well-known platforms. Unfortunaly, I still fail to see a lot of counter arguments against the rationales exposed by Florian and we already have to admit that the switch to bitbucket has been a success, as it has allowed a new generation of developpers to take over pylint's development and making it even more successful. So if you, people who make the project alive today, feel like it has to be done, go with it. Even if it makes me, and maybe a few others, sad.
Still, if there is some options to use e.g. gitlab with a mirroring to github or alike which keeps the expected benefit but without the centralized model, I would like to have them considered. This maybe of interest for others PyCQA projects.
Using GitLab and mirroring to GitHub is a definite option. PyCQA already has an organization there and mirrors a few repos to GitHub. That said, I maintain a Jenkins infrastructure for those projects so I can run CI on the merge requests received through GitLab. GitLab only recently launched their own CI service and I haven't had time to investigate it at all. Travis CI is most certainly a lot less to manage if they move to GitHub. (Note that Flake8 has Travis enabled in the event someone sends a PR through GitHub.) Like I said, I won't try to sway either way, but I would vastly prefer if Travis supported GitLab as well, but they don't and probably never will. Cheers, Ian