threads / discuss / 11391

Pushing and fetching sha1s directly

Subject: Pushing and fetching sha1s directly

## tl;dr

3 messages between Dec 22, 2007 and Dec 23, 2007.

replies: 2people: 3as markdown or json

Finn Arne Gangstad· Dec 22, 2007, 21:13 UTC · lore

Currently there seems to be no way of pusing a sha1 directly, or to fetch a sha1 directly. When working with submodules, it is convenient to be able to work with detached HEADs, so it would be good if this could be supported also by fetch and push.

What would be a resonable syntax for this? I'm thinking something like git fetch --sha1 <repository> <sha1> git push --sha1 <repository> <sha1>

Where <sha1> could really be a <commit> I guess, but the option name "--commit" seems wrong somehow. Another option is to extend refspecs so sha1s can be allowed in there directly, so this would just work:

git fetch <repository> <sha1>

What do you prefer?

- Finn Arne

Junio C Hamano· Dec 22, 2007, 23:12 UTC · re: Finn Arne Gangstad · lore

Re: Pushing and fetching sha1s directly

Finn Arne Gangstad <finnag@pvv.org> writes:

> Currently there seems to be no way of pusing a sha1 directly, or to
> fetch a sha1 directly. When working with submodules, it is convenient
> to be able to work with detached HEADs, so it would be good if this
> could be supported also by fetch and push.

At least push of an arbitrary commit is already supported, I think.

For fetch, I would normally say "check the list archives", but it was very long time ago that it was proposed and discussed. It has some security and performance implications.

Johannes Schindelin· Dec 23, 2007, 20:04 UTC · re: Junio C Hamano · lore

Re: Pushing and fetching sha1s directly

Hi,

On Sat, 22 Dec 2007, Junio C Hamano wrote:

Show 10 quoted lines

> Finn Arne Gangstad <finnag@pvv.org> writes:
> 
> > Currently there seems to be no way of pusing a sha1 directly, or to 
> > fetch a sha1 directly. When working with submodules, it is convenient 
> > to be able to work with detached HEADs, so it would be good if this 
> > could be supported also by fetch and push.
> 
> For fetch, I would normally say "check the list archives", but it was 
> very long time ago that it was proposed and discussed. It has some 
> security and performance implications.

To enhance on that: sometimes you need to delete a branch in a hurry, because you publicised some secret code/file. In that case, you do _not_ want people being able to fetch arbitrary commits.

Ciao, Dscho