Le 30/11/2010 14:56, Guillaume Hoffmann a écrit : > Hi everyone, > > in August, Petr proposed a patch to have the commands ``convert > --import`` and ``convert --export``: > > http://lists.osuosl.org/pipermail/darcs-users/2010-August/024754.html > I don't think I answered this thread, but I'm think having these as a separate command is neither necesseray nor desirable. I think it is clearer if the darcs executable deals with darcs repositories, and tools that deal with other stuff stay in their own commands. This helps keeping darcs' consistency and manageability (by having separate manuals, and keeping the smallest set of commands ;-) ). Also, it is better for release management since, should we take darcs-convert under the darcs-team umbrella, we keep the possibility to release darcs-convert more often than darcs itself (if, for instance, the format for importing in git change, or we improve the sitution hg-wise).
> The answers to this were that it was better to keep this feature as a > separate tool. In particular (as Ganesh said), if people start relying > on this feature on a regular basis to keep darcs-git mirrors, we > should guarantee them some stability. > > Eric concluded: "Or maybe to develop a style of working on new > features where we let them stand alone for a little while (incubation > phase) and then fold them into Darcs as it becomes clear how important > they are and how to fit them in." > As long as we decide to maintain darcs-fastconvert and to advertise it officially, I don't see what advantages there are to having darcs-fastconvert included in darcs rather than separate. > So Petr wrote darcs-fastconvert and released it on Hackage: > http://lists.osuosl.org/pipermail/darcs-users/2010-August/024987.html > > Which is now a usable tool that has already helped a few people > convert and maintain mirrors of darcs repositories: > > * > http://www.serpentine.com/blog/2010/10/21/how-to-migrate-from-darcs-to-mercurial-or-git/ > * http://tab.snarc.org/blog/post/2010/11/17/darcs_mirrors_to_git/ > * > http://webcache.googleusercontent.com/search?q=cache:oFiMzMUCCC8J:sgenomics.org/~jtang/blog/ > > I think now there has been enough incubation and testing for this > feature, and it should be included in 2.8 as subcommands of > ``convert``. > If we include it, I'm strongly against using the convert subcommand for import-export, since it already has delicate semantics which would only be made even more byzantine: darcs convert works in place, while darcs-fastconvert deals with import/export streams, darcs convert needs to be done once centrally, a constraint we may or may not have with darcs-fastconvert. I think the team should "adopt" darcs-convert, but keep it a separate command, for the UI and release management reasons stated above. This means that if Petr stops working on it, we keep the maintenance. Also, we should accept darcs-convert's issues in bugs.darcs.net. > Why have it inside of Darcs? Well I have recently read the manual of > another SCM called Fossil ( http://www.fossil-scm.org ) and it does > have an integrated ``convert`` subcommand. And it makes sense: like > Darcs, Fossil has few users and has to live in a world where Git and > Github are very common. Switching to Darcs or Fossil has to be as easy > as possible for users, hence the integrated fastconvert feature makes > sense. Not everyone can afford to install the Haskell Platform and > build darcs-fastconvert themselves. > We certainly should have darcs-convert binaries for the platforms which have darcs binaries now. We have to have it packaged in debian, and so on. > If we agree on having this in 2.8, there is no hurry to do it now. > Maybe we can plan to have it in Darcs once the changes to drop writing > support for OF repositories are done. (By "we", I'm mostly involving > Petr because it would be easier for him to do the inclusion, and > probably not harder than maintaining fastconvert itself). > > What do you think? > > Now about the process: > > The fact that darcs-fastconvert was released as a separate package has > enabled users to try it outside of the Darcs release schedule. I'm > glad the feature took that path in the first place: it wouldn't have > been included in 2.5 anyway because of the freeze. It has worked > great. Also it made Petr use libdarcs, so we have another proof that > it's possible to write something on top of libdarcs :-) > That's another benefit of having darcs-fastconvert as a library program distributed by the darcs team: it makes a good darcslib/example.hs. Grumpily, Senilely and Manly yours, Florent _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
