Prototype "improved_chk_index"

[Ian Clatworthy]

John Arbash Meinel wrote:

> Certainly there is still the question of 'good enough'. It currently
> takes ~2m15s to extract all of the ancestry of launchpad from my test
> repository. The above numbers hint that at least 10s of that is just
> processing the chk index. Not huge, but it would be an ~8% performance
> improvement for an actual operation.

"Good enough" in my mind means:

1. Storage efficiency equal to or better than Git

AND

2. Speed equal to or better than Mercurial.

We're teasingly close to that but not there yet.

I don't believe efficiency+performance is our top issue right now. OTOH,
I'd like to see us continue to beaver away in the background until the
above holds true across 95% of projects. While these things are fresh in
your mind and you're motivated to nail them, I'm ecstatic you're
tackling them. (In *my* opinion, if that means myself and others need to
do more in other areas, so be it.)

Ian C.