> On 2010-10-18 3:32 PM, Brian Warner wrote: >> - darcs is slow, hard to publish branches, hard to use local >> branches,
Zooko just pointed me at the following tidbit: http://darcs.net/manual/Best_practices.html#SECTION00530000000000000000 to turn on the global patch cache. The version of darcs that I've been using (2.1.0) does not happen to enable this by default, whereas newer versions do. It makes a phenomenal improvement in speed. Cloning a new tree (by "downloading" it: darcs get --lazy http://tahoe-lafs.org/source/tahoe/trunk) now finishes in roughly *one second*, because it's re-using patches that were fetched earlier and stored in ~/.darcs . That's about 3x faster than Git (with a hot cache). My other criticisms still stand, and I am still disappointed by the speed of e.g. 'darcs whatsnew -l' and 'darcs diff', and the need to create a new tree for each branch (and therefore to rebuild dependencies). But the really painful delay of creating a new tree was an artifact of an old version, and no longer an issue for me or anyone else (who runs a modern version or learns how to enable the cache). As for Darcs being in Haskell, the only problem there is popularity: few people know the language, so the population of eligible authors and improvers and maintainers for Darcs is smaller. Popularity of the toolchain should never be the sole driver of technology choices (else we'd all be using CVS on windows), but there are hazards to being *too* far off the beaten path, and not being able to get darcs running on new platforms (because GHC is too hard to build) is one of them. cheers, -Brian _______________________________________________ tahoe-dev mailing list [email protected] http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
