Re: [PATCH] [RFC] list-objects-filter: introduce new filter sparse:buffer=<spec>

From: ZheNing Hu <adlternative@gmail.com>
Date: Aug 9, 2022, 06:13 UTC
Message-ID: <CAOLTT8TSyKArOKTjbuGO=OkpciS6DH0mmVPSmiidOSHxo4thNQ@mail.gmail.com>
In-Reply-To: <xmqqczdau2yd.fsf@gitster.g>

Junio C Hamano <gitster@pobox.com> 于2022年8月9日周二 00:15写道：

Show 15 quoted lines

>
> "ZheNing Hu via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > From: ZheNing Hu <adlternative@gmail.com>
> >
> > Although we already had a `--filter=sparse:oid=<oid>` which
> > can used to clone a repository with limited objects which meet
> > filter rules in the file corresponding to the <oid> on the git
> > server. But it can only read filter rules which have been record
> > in the git server before.
>
> Was the reason why we have "we limit to an object we already have"
> restriction because we didn't want to blindly use a piece of
> uncontrolled arbigrary end-user data here?  Just wondering.
>

* An end-user's maybe doesn't even have write access to the repository,
so they can't config a filterspec file before git clone, what should they
do now?

* If there are  thousands of  different developers use the same git repo,
and they use "--filter=sparse:oid" to do different partial-clone, then how
many filterspec file should repo managers config first?

* Why not carefully check "uncontrolled arbigrary end-user data" here,
such as add a config like "partialclone.sparsebufferlimit" to limit transport
data size, or check if filterspec file is legal? Or if git server
don't trust its
user... we can use a config to ban this filter, And within some companies,
users can basically be trusted.

 * I'm sure it would be beneficial to let the filtering rules be configured
by the user, because now many people have such needs: download
only a few of files of directories of the repository.

* sparse-checkout + partial-clone is a good reference: we have a
".git/info/sparse-checkout" for record what we actually want to checkout to
work-tree, and it will fetch some missing git objects which record in
".git/info/sparse-checkout" from git server. I know it use <oid> to fetch
objects one by one instead of "path"... But In hindsight, its performance is
extraordinarily bad as a result...

Anyway, this patch represents some of my complaints about the current partial-clone feature and I hope the community will move forward with it.

Thanks.

ZheNing Hu