On Thu, Sep 2, 2010 at 4:58 PM, Jeff King <peff@xxxxxxxx> wrote:
> pack-objects will reuse previously found deltas. So the deltas you have
> in your existing packs matter. The deltas you have in your existing
> packs depend on many things. At least:
>
> 1. Options you used when packing (e.g., --depth and --window).
>
> 2. Probably exactly _when_ you packed. You could find a good delta
> from A to B. Later, object C comes into existence, and would
> provide a better delta base for B. I don't think we will ever try A
> against C, unless --no-reuse-delta is set.
>
> You have a different pack than somebody who packed after A, B, and
> C all existed.
>
> In practice, this tends not to happen much because the best deltas
> are usually going backwards in time to a previous version. But it
> can happen.
jeff, thanks for explaining (and to nicolas, i see, since beginning this)
mrhmfffh. just been reading Documentation/technical/pack-heuristics.txt.
so... options include:
* writing an alternative "canonical" pack-object algorithm. i'm
inclined to select "git format-patch"! :) but that would be the lazy
way....
* taking the seeder's pack-objects as the "canonical" ones,
regardless. cacheing of the results would, sadly, be virtually
unavoidable, given the situation (multi-threading etc.)
* throw away bittorrent entirely as a transport mechanism.
* force-feed one peer to be "the" provider of a particular given
pack. doesn't matter whom you contact to _obtain_ a pack from, as
long as you solely and exclusively get the pack from that particular
peer.
* slight improvement / variation on the above: if two peers just
coincidentally happen to create or have the same pack (as can be shown
by having the same SHA-1, and/or by having the same data in their
cache) then ta-daaa, you have a file-sharing network for that
particular pack.
i think.... i think i miiight be able, hmmm... i believe it would be
possible to implement this last option by creating separate .torrents
for packs (one each!). by splitting things down, so that pack objects
are named as {ref}-{objref}-{SHA-1}.torrent and by providing a "top
level" torrent which contains the refs/heads/* and the associate
rev-list(s)... each set of rev-lists would have the SHA-1 of the
pack-object that happened to be created (and shared) at that
particular time, from that particular client: you then genuinely don't
give a stuff about who has what, it's all the same, and...
hmmm, i feel a modification to / deviation from the bittorrent
protocol coming on :)
l.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html