On Mon, Feb 07, 2022 at 12:49:56PM +0100, Christian Weisgerber wrote:
> Sorry, I didn't get as far as testing the diff, since the comparison
> run with the old code already suffered problems:
> 
> * My first attempt ran out of /tmp space.  4.8G was not enough.

Caching deltas in compressed form in the future could help with this.

> * Related to that: With more /tmp space, "gotadmin pack" eventually
>   wrote a 4.7G pack file.  The original git clone of the repository
>   has only a 1.4G pack file.

Yes, deltas we generate are significantly larger than those generated
by Git. Reusing deltas generated by Git will help keep our packs a bit
smaller, for now. We can also look at ways of making our own deltas
smaller. With delta-reuse in place we may be able to afford additional
run-time cost if that would help with creating smaller deltas. There are
space/time trade-offs we can make by tuning deltification parameters.

However, the deltification algorithms implemented by Git and Got are not
the same. It is possible that a significant difference will always remain
unless we rewrite code inherited from git9 and use a different approach.