Re: What's cooking in git.git (Mar 2016, #04; Wed, 23)

2 messages, 2 authors, 2016-06-15 · open the first message on its own page

Re: What's cooking in git.git (Mar 2016, #04; Wed, 23)

From: Junio C Hamano <hidden>
Date: 2016-06-15 23:08:56

"Philip Oakley" [off-list ref] writes:
quoted
The beginning of "split bundle", which could be one of the
ingredients to allow "git clone" traffic off of the core server
network to CDN.
...
Hi Junio,

I think there may be a concept clash between the ideals of a
sneakernet bundle' and the 'resumable clone'.
Notice the "could" above ;-)

Read the original thread and notice that the inclination is for the
first one the primary "clone priming" mechanism would likely to be a
packfile, not a bundle, even though use of "bundle" is not ruled out.

Re: What's cooking in git.git (Mar 2016, #04; Wed, 23)

From: Philip Oakley <hidden>
Date: 2016-06-15 23:08:56

From: "Junio C Hamano" <redacted>
"Philip Oakley" [off-list ref] writes:
quoted
quoted
The beginning of "split bundle", which could be one of the
ingredients to allow "git clone" traffic off of the core server
network to CDN.
...
Hi Junio,

I think there may be a concept clash between the ideals of a
sneakernet bundle' and the 'resumable clone'.
Notice the "could" above ;-)

Read the original thread and notice that the inclination is for the
first one the primary "clone priming" mechanism would likely to be a
packfile, not a bundle, even though use of "bundle" is not ruled out.
I'd seen the thread ($gmane/288380), and that it had developed toward a bare 
pack-file.

This was just clarifying that if a variant of the bundle format (# V3?) was 
used, that it must, if the name was retained(*), still work as a sneakernet 
transfer option. In that case the user would need to be told, or be able to 
find out via (e.g.) the 'verify' sub-command, where the other half of the 
split bundle (the pack) was located so that both halves could be copied for 
sneakernet transfer.

Then on reaching the destination, the user would need to appreciate where 
the two halves are to go (side by side?), and the code would need a way of 
knowing that it should use the local copy of the split pack, rather seeking 
to transfer a fresh copy (which obviously fails in an air-gapped scenario).

(*) Changing the user facing names for the new resumable transfers to avoid 
the potential user confusion is a simple solution, which could just be a 
tweak to the final patch 4/4 c34c9a9db65 "bundle v3: the beginning". Perhaps 
explicitly use "split-bundle", though choosing a good alternative isn't easy 
;-)

As you say that all could become academic if the independent pack transfer 
works as a better resumable clone.

--

Philip
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help