Best method to pull *fast*?

[B Smith-Mannschott]

On Tue, Jan 19, 2010 at 16:20, John Arbash Meinel
<john at arbash-meinel.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Ali Sabil wrote:
>> On Tue, Jan 19, 2010 at 12:07 PM, Torsten Bronger
>> <bronger at physik.rwth-aachen.de> wrote:
>>> Hallöchen!
>>>
>>> I stay up-to-date with CVS Emacs by pulling from the inofficial
>>> Bzaar branch.  I don't contribute, I just want to synchronise
>>> unidirectionally.  However, an ordinary "pull" generates 140 times
>>> more network traffic than bytes changed on the disk.  In one case, I
>>> had a diff which was 500 kB big, but "bzr pull" reported to have
>>> transmitted 80 MB.
>>>
>>> Using a lightweight checkout and "update" didn't change much,
>>> however, I haven't really benchmarked it yet.
>>>
>>> Is there a faster way?
>>>
>>
>> Did you try to use the launchpad mirror (bzr pull lp:emacs) ?
>
> I would recommend the Launchpad site for now. Though you should also set
> up a user account. A smart protocol can have the server chop up the bits
> you actually need and send it to you, while branching over http or sftp
> (the only ways provided by Savannah, afaik) require the client to
> download the data and then chop it up itself.

I assume the OP is already using lp:emacs ("inofficial Bzaar branch"
[sic]), as I have been using that branch to mirror emacs myself and
saw just such an 80MB update come down the wire around the time when
copyright notices were updated to 2010 in most source files. (Though I
have *no* way of knowing if there's any causal relationship there.)

What combination of checkout/branch/--lightweight incantations is
likely to produce a branch that is inexpensive (in terms of bytes
transferred) to keep up to date?

// Ben