Jason Dunsmore <emacs-orgm...@dunsmor.com> writes: > If you clone via git://, it does some optimizations during the transfer, > whereas cloning via http:// does not. If I do a "git gc" on the repo, > it reduces the size of the repo on the server and the time to download > via http://
I'd suggest that the repo on the server should get a regular garbage collect (daily, perhaps via cron) and on each release an aggressive garbage collect. It will be much easier on the server (reduced traffic and fewer requests to serve). Not only is the size of the repo much larger than it needs to be, the server also has to deal with lots of small requests that can't be efficiently transferred. Cloning the repo via http protocol produces 190MByte on disk, a simple and fast 'git gc' gets that down to 63MByte, just like cloning via git protocol (ironically, the server will do that work on each clone as long as you do not do a garbage collect on the repo itself). An aggressive garbage collect takes a while longer, but reduces the size of the repo down to 44MByte. An http clone on repo.or.cz currently delivers 72MByte, so it appears they do a regular garbage collect (or it simply comes courtesy of mirroring). Regards, Achim. -- +<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+ Factory and User Sound Singles for Waldorf rackAttack: http://Synth.Stromeko.net/Downloads.html#WaldorfSounds
