Re: How does team maintenace of python module works?
On 02/21/2013 01:59 AM, Scott Kitterman wrote:
> I don't keep local copies of things I don't work on regularly, so for team packages I would be downloading all of it.
I thought about it for a long time. Then I tried again today, and now,
I don't buy into this "it takes too much time to download" argument.
Currently, with SVN, givent the size of the repository, it takes really
a huge amount of time to clone a package *debian folder only*.
>From a server in a data center in Seattle, it took me 90 seconds
to download the packaging sources of python-eventlet. Compare
this to the 4 seconds it takes me to clone python-warlock from
Alioth, upstream source included !!!
Then because this was only a small package with only a very small
amount of upstream code, I tried again to benchmark how much
time it took with a bigger project. Well, with nova, which I consider
a quite big project, it took me 32 seconds to get all of it. And about
10 seconds for glance, which I would consider a medium size project.
Then I repeated the experience from my (very) slow Chinese
ADSL connection. Even there, it took me 42 seconds to clone
glance from Alioth (a 9.4 MB repository).
Scott, I would be *very* surprised if you had an internet connection
slower than mine, considering where you live (if db.debian.org still
has accurate information).
The problem is that with SVN, it takes so much time and space
(as each tag will generate some files), so you might have been
fooled into thinking it would also with Git. But the reality simply
not that at all.
If you could suffer the slowness of SVN for so many years with the
python module team, I'm sure you will like the upgrade to Git which
will make clones so much faster.
So in the end, I don't buy into your point that having an upstream
source branch type of repository is too slow and too big. It's just
simply not the case.
Cheers,
Thomas Goirand
Reply to: