From: Taylor Blau <me@ttaylorr.com>
To: git@vger.kernel.org
Subject: [NOTES 06/11] Repository maintenance long-term goals
Date: Mon, 6 Oct 2025 15:19:39 -0400 [thread overview]
Message-ID: <aOQWS+ZJdXr73PE0@nand.local> (raw)
In-Reply-To: <aOQVeVYY6zadPjln@nand.local>
Topic: Repository maintenance long-term goals
Leader: Taylor Blau
* Taylor's talk was limited towards the end. Could expand on that future work.
* Constant repacking into a single pack was historically the major problem.
* Doing that less (because of geometric repacking) helps, but it's still a
potential issue when it does occur. Gets them 98% of the way.
* Future items were geometric reachability, ?, best effort gc
* Previously during geometric, used to accumulate loose objects too. 6 months
ago they changed to an approach where the big cruft pack could be excluded
from the midx.
* Challenge would be to do a full complete repack without rewriting all of the
midx chain.
* Because bitmap is tied to object order in a pack, need something like
tombstones to not break the bitmaps. Need the tombstone to know that we don't
have the data.
* Unitary midx idea - Taylor designed the chained midx before he figured out the
repacking strategy. MIDX and pack index duplicate the data. No reason to
de-dup other than for space saving. Could even skip having idx, but plenty of
old git versions can't read midx.
* brian - there may be other implementations, such as git lfs, that don't use
midx and object id mappings in pack idx v3 aren't supported in midx either.
* Nothing preventing you from having two parallel repacks, one that's geometric
and one that's trying to do an all-into-one.
next prev parent reply other threads:[~2025-10-06 19:19 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-10-06 19:16 Notes from the Git Contributor's Summit, 2025 Taylor Blau
2025-10-06 19:18 ` [NOTES 01/11] SHA-256 and interoperability work Taylor Blau
2025-10-06 19:18 ` [NOTES 02/11] First-class conflicts in Git? Taylor Blau
2025-10-06 19:18 ` [NOTES 03/11] The future of history rewriting - rebase, replay and history (+Change-IDs) Taylor Blau
2025-10-06 19:18 ` [NOTES 04/11] Rust Taylor Blau
2025-10-06 19:19 ` [NOTES 05/11] Pluggable object databases Taylor Blau
2025-10-06 19:19 ` Taylor Blau [this message]
2025-10-06 19:19 ` [NOTES 07/11] Change-ID Header in Git Taylor Blau
2025-10-06 19:20 ` [NOTES 08/11] Resumable fetch / push Taylor Blau
2025-10-06 19:20 ` [NOTES 09/11] Git 3.0 Taylor Blau
2025-10-06 19:20 ` [NOTES 10/11] How can companies respectfully engage contractors to work on Git? Taylor Blau
2025-10-06 19:20 ` [NOTES 11/11] Conservancy 2025 updates Taylor Blau
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=aOQWS+ZJdXr73PE0@nand.local \
--to=me@ttaylorr.com \
--cc=git@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).