Proper Publishing of a Repository

Proper Publishing of a Repository

[Jon Loeliger]

Folks,

So, I feel like I missed a step in the grand
"How To Publish A Repository" scheme of things.

I made a repo visible over on jdl.com.  No problem.
But cloning it took forever.  So I ran "git-repack"
on it.  Now cloning only takes hours, not forever.

All this on the linux kernel over HTTP.

Did I miss a step somewhere?

Thanks,
jdl

Re: Proper Publishing of a Repository

[Timo Hirvonen]

On Wed, 22 Mar 2006 13:17:52 -0600
Jon Loeliger <jdl@freescale.com> wrote:

> Folks,
> 
> So, I feel like I missed a step in the grand
> "How To Publish A Repository" scheme of things.
> 
> I made a repo visible over on jdl.com.  No problem.
> But cloning it took forever.  So I ran "git-repack"
> on it.  Now cloning only takes hours, not forever.
> 
> All this on the linux kernel over HTTP.

Use git:// protocol (git-daemon) if possible.  It is much faster.

-- 
http://onion.dynserv.net/~timo/

Re: Proper Publishing of a Repository

[Jon Loeliger]

On Wed, 2006-03-22 at 14:16, Timo Hirvonen wrote:
> On Wed, 22 Mar 2006 13:17:52 -0600
> Jon Loeliger <jdl@freescale.com> wrote:
> 
> > Folks,
> > 
> > So, I feel like I missed a step in the grand
> > "How To Publish A Repository" scheme of things.
> > 
> > I made a repo visible over on jdl.com.  No problem.
> > But cloning it took forever.  So I ran "git-repack"
> > on it.  Now cloning only takes hours, not forever.
> > 
> > All this on the linux kernel over HTTP.
> 
> Use git:// protocol (git-daemon) if possible.  It is much faster.

Trust me, if it were an option, I would.  It isn't.

So, I think git-prune-packed was the answer here...
jdl

Re: Proper Publishing of a Repository

[Junio C Hamano]

Jon Loeliger <jdl@freescale.com> writes:

> On Wed, 2006-03-22 at 14:16, Timo Hirvonen wrote:
>> Jon Loeliger <jdl@freescale.com> wrote:
>>
>> > All this on the linux kernel over HTTP.
>> 
>> Use git:// protocol (git-daemon) if possible.  It is much faster.
>
> Trust me, if it were an option, I would.  It isn't.
>
> So, I think git-prune-packed was the answer here...

It really depends on who your audiences are.  By packing
everything in a single pack, you are optimizing for the initial
cloners but punishing incremental updaters.

Since your tree is derived from Linus tree, anybody who is
interested in your tree is very likely nterested in Linus tree
and chances are that she has a recent copy of it locally.  In
such a case, I would not be surprised if:

	$ git clone -l -s -n /local-linux-2.6.git/ jdl-linux.git
        $ cd jdl-linux.git
        $ git fetch http://jdl.com/jdl-linux.git master:refs/heads/jdl

goes a lot faster if you have loose objects around in the
repository.