Hello. On 08/10/2018 08:09 PM, Mike Blumenkrantz wrote: > > https://gitlab-prototype.s-opensource.org/
Thanks to the intern without name to get a real PoC out for this. While people have advocating for such a move no one before her/him spent the actual time to get this demonstrated! This will help us to understand the details of a potential move way better. > > Some notes: > * This is read-only for now > * User creation is disabled, don't bother trying > * Issues with their comments have been imported Cool > * Patch submissions have been imported (the intern screwed up some of the > early imports so there are a few patches without the diff inlined) > - Comments on patch submissions cannot be imported because Phabricator > has no API for retrieving comments on patch review That is a bit of a pity. One could think of scraping the original diffusion web pages for the comments. Not sure if it would really be worth the effort spent on a script doing that. If we are able to clear out our patch queue enough this issue would minor. > * Wiki pages are not imported since some decision-making is required > > As is easily noticeable, not all projects have been imported by my intern. > Importing the repo takes some time on its own, and then running the > migration script takes a variable amount of time on top of that depending > on the size of the project (EFL was estimated to take 10+ hours to fully > import). As a first demonstration this helps already. If the community wants to go this way we can have further steps where we import more projects and check for consistency and sanity. I would expect there would be several of such cycles before we are happy and would make a final switch > Wiki pages have not been imported. On Gitlab, a wiki is project-specific > and so it is impossible to do a 1:1 copy unless we decided to stick > everything onto a specific project. We would have to decide how we want to > do this. Hmm. The way we used the phab wiki was really for the overall community and not individual projects. Having said that I would think that most of the wiki pages could be attached to efl, EFL or Terminology. The rest will most likely be pages on process (commits guidelines, releases, etc) There will also be a ton of outdated pages which could simply be removed. In the end we would need to go through all of them and decide what to do. e.g move process docs into dokuwiki, remove outdated ones, move to a specific project. If we should do this sortign before or after me moved all pages over I am on the fence. It would be cleaner to only import the sorted pages but that can delay the move for some time. > If we decided to switch to Gitlab, there would be a number of questions > that need to be answered: > Q: How do we migrate? > A: Gitlab cannot accurately mirror all of Phabricator, it can only do a > one-time migration of projects. This means we would at some point lock phab > and then begin migrating, likely over a weekend for the major projects with > the remainders being added later. As written above I would expect that we would have more trial runs on the migration while fine tuning the script and reviewing the results. If we are happy with what we have for a project we could lock it for a weekend and move the projects over. > Q: What happens to phab? > A: We would likely want to keep phab in read-only mode for a while after > the migration since all the migrated tickets/patches will provide links to > it. We can later evaluate if we need to keep it running. With all the reference to tickets and differentials we might need keep it around read only for a long time. The alternative would be to have all the old issues and differentials imported as well and have a re-write mapping for the links in our http server. Not very attractive either. > Q: Where would this be hosted? > A: The provided link here is a cloud service which will be funded for the > foreseeable future. This is a crucial point here. Business decisions change and the community has no influence on this. With my community hat on I appreciate that there would be a sponsoring of a cloud service, but I truly think we should not depend on this mid or long term (having it run there for a few month of migration would not worry me). Even if it would be more paperwork having the sponsorship going to the foundation and the service being paid out from there would be the right way. We can acknowledge the sponsorship on our sponsors page. regards Stefan Schmidt ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
