Re: [PATCH v2 2/2] index-pack: kill union delta_base to save memory

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 27, 2015, 21:18 UTC
Message-ID: <xmqqzj7zhx31.fsf@gitster.dls.corp.google.com>
In-Reply-To: <1424947928-19396-3-git-send-email-pclouds@gmail.com>

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Notice that with "recent" Git versions, ofs-delta objects are
> preferred over ref-delta objects and ref-delta objects have no reason
> to be present in a clone pack.

It is true that we try to use ofs-delta as much as possible, but where does "have no reason to be present" come from?

When an object cannot be represented as an ofs-delta (which can only refer backwards), don't we use ref-delta, instead of storing it as a deflated-full object?

Probably "Not so ancient versions of Git tries to use ofs-delta encoding whenever possible, so it is expected that objects encoded using ref-delta are minority" may be closer to the truth. And that observation does justify why using two separate pools (one with 8-byte entries for ofs-delta, the other with 20-byte entries for ref-delta) is a better idean than using one pool with 20-byte entries for both kinds.