[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
måndagen den 21 juni 2010 16.00.25 skrev Shawn Pearce:
> On Mon, Jun 21, 2010 at 2:31 AM, Alex Blewitt <alex.blewitt@xxxxxxxxx>
wrote:
> > Is there a reason why we use git+ssh as the protocol for remote
> > connections when the Git documentation only talks of ssh connections?
>
> Long story.
>
> Originally Git only used rsync for transport, and the argument given
> to git was passed down to rsync as-is. So "user@host:path" was the
> typical style for SSH connections, as that is what rsync uses to open
> an SSH connection to the remote host.
>
> Then anonymous git:// came along, because rsync wasn't ideal for Git
> transport. This lead rise to having a protocol name as part of the
> URL.
>
> Somewhere along the line we learned ssh:// to mean use an SSH
> connection. But I think another system was already using git+ssh://
> (Cogito?), and users confused the order, so ssh+git:// was also
> accepted... and now its a big pile of synonyms for the same thing.
>
> As far as preference goes, I think Linus Torvalds, Junio Hamano and
> myself all prefer "user@host:path" for SSH URIs. But not all tools do
> that. Gerrit Code Review for instance has to advertise ssh:// URIs
> because it needs to embed the non-standard port number into the URI,
> and I think ssh:// is cleaner to read than git+ssh.
>
> As for why EGit prefers git+ssh over ssh, I don't know. My memory is
> fuzzy, but I think its Robin's preference to use git+ssh to mean an
> SSH URI that is speaking Git over it.
Nah, I've always use ssh://. I don't think I've every typed the
combination "git+ssh" (before this mail). The code for cloning has been
written and rewritten many times and one of these "someone" (not me!) felt the
need to have only one name for the ssh protcol and chose "git+ssh". And the
rest of us were so excited about the GUI's so we did not see that or forgot
about it during review.
-- robin
