Transition from svn to git

classic Classic list List threaded Threaded
25 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Re: Transition from svn to git

Martin Chilvers
G'day,

I'm sure this was covered earlier, but unless git really is technically
superior (for our use cases!), then I like the idea of mercurial because:-

1) We are Python advocates and it would seem consistent for us to
support (and be seen to support) a Python DVCS (dog food and all that)...

2) I get a little worried when I hear about Windows installation issues
- they give off a slight aroma of "half-baked"!

3) If we're considering the morals of bitbucket plagiarizing git then we
should also probably review our use of PySide ;^)

But, like I said, if there are strong technical advantages to git then
we should use it, and just live with (or help fix!) the Windows issues...

Martin

On 19/01/2011 20:21, Gael Varoquaux wrote:

> On Wed, Jan 19, 2011 at 01:35:50PM -0600, bryce hendrix wrote:
>>     We had a lengthy discussion (again) during lunch. I was asking why github
>>     is better than bitbucket. I came across this article from someone on the
>>     github team:
>>     [1]http://dev.pocoo.org/~blackbird/github-vs-bitbucket/bitbucket.html
>>     The arguments why github are as follows:
>>     �1. bitbucket copied their website feature for feature.
>>     �2. to fork on bitbucket you need to fill out a form
>>     �3. github offers additional Post Services on jabber and IRC
>>     �4. github has more users
>>     I was expecting to see some technical advantage, but I'm not seeing it. Is
>>     there something I am missing?
>
> The two aspects in favor of github are:
>
>      1. Popularity: having a project on github gives it visibility to a
>         bunch of coders. These are the people we want to interest. On top
>         of that, in the scipy micro-community, people use github a lot,
>         so it will be easier to draw from that lot.
>
>      2. Usability:  Details like the forkqueue where you can integrate
>         patches with one click, displaying the readme (formated if it's
>         rest), serving svn, and (the killer feature, IMHO) commenting
>         directly on commit.
>
> 1 is more important than 2, but in my experience 2 is nice.
>
> It's funny, a year ago I was arguing against a switch to git+github on
> the nipy and related projects. Looking back I can say that I strongly
> dislike git, but the gains that come from github beat the pain from git.
>
> My 2 cents,
>
> Ga�l
>
>
>
> _______________________________________________
> Enthought-Dev mailing list
> [hidden email]
> https://mail.enthought.com/mailman/listinfo/enthought-dev

_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev
Reply | Threaded
Open this post in threaded view
|

Re: Transition from svn to git

Ilan Schnell
> I'm sure this was covered earlier, but unless git really is technically
> superior (for our use cases!), then I like the idea of mercurial because:-
>
> 1) We are Python advocates and it would seem consistent for us to
> support (and be seen to support) a Python DVCS (dog food and all that)...
On the other hand numpy, ipython (and soon scipy) are on git.

> 2) I get a little worried when I hear about Windows installation issues
> - they give off a slight aroma of "half-baked"!
>
> 3) If we're considering the morals of bitbucket plagiarizing git then we
> should also probably review our use of PySide ;^)
>
> But, like I said, if there are strong technical advantages to git then
> we should use it, and just live with (or help fix!) the Windows issues...
As mentioned earlier by Chris, speed is a strong technical advantage.

- Ilan
_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev
Reply | Threaded
Open this post in threaded view
|

Re: Transition from svn to git

william ratcliff
Also,

For windows, tortoisegit is rather nice and doesn't require cygwin.  

William

On Thu, Jan 20, 2011 at 3:10 PM, Ilan Schnell <[hidden email]> wrote:
> I'm sure this was covered earlier, but unless git really is technically
> superior (for our use cases!), then I like the idea of mercurial because:-
>
> 1) We are Python advocates and it would seem consistent for us to
> support (and be seen to support) a Python DVCS (dog food and all that)...
On the other hand numpy, ipython (and soon scipy) are on git.

> 2) I get a little worried when I hear about Windows installation issues
> - they give off a slight aroma of "half-baked"!
>
> 3) If we're considering the morals of bitbucket plagiarizing git then we
> should also probably review our use of PySide ;^)
>
> But, like I said, if there are strong technical advantages to git then
> we should use it, and just live with (or help fix!) the Windows issues...
As mentioned earlier by Chris, speed is a strong technical advantage.

- Ilan
_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev


_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev
Reply | Threaded
Open this post in threaded view
|

Re: Transition from svn to git

william ratcliff
In reply to this post by Ilan Schnell
Also,

For windows, tortoisegit is rather nice and doesn't require cygwin.  

William

On Thu, Jan 20, 2011 at 3:10 PM, Ilan Schnell <[hidden email]> wrote:
> I'm sure this was covered earlier, but unless git really is technically
> superior (for our use cases!), then I like the idea of mercurial because:-
>
> 1) We are Python advocates and it would seem consistent for us to
> support (and be seen to support) a Python DVCS (dog food and all that)...
On the other hand numpy, ipython (and soon scipy) are on git.

> 2) I get a little worried when I hear about Windows installation issues
> - they give off a slight aroma of "half-baked"!
>
> 3) If we're considering the morals of bitbucket plagiarizing git then we
> should also probably review our use of PySide ;^)
>
> But, like I said, if there are strong technical advantages to git then
> we should use it, and just live with (or help fix!) the Windows issues...
As mentioned earlier by Chris, speed is a strong technical advantage.

- Ilan
_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev


_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev
Reply | Threaded
Open this post in threaded view
|

Re: Transition from svn to git

Robert Kern
In reply to this post by Martin Chilvers
On Thu, Jan 20, 2011 at 1:53 PM, Martin Chilvers
<[hidden email]> wrote:
> G'day,
>
> I'm sure this was covered earlier,

Yes, but mostly internally because we are deciding about moving all of
our development, not just the open source parts.

> but unless git really is technically
> superior (for our use cases!)

I hate to step into this again, but here are the main technical
arguments in favor of git+github over mercurial+bitbucket/Google
Code/RhodeCode/other:

1. Efficient local branches.
2. github's pull requests.
3. SVN interface through github for users who prefer it.
   Particularly, this is important for our corporate customers who
tend to use SVN already and are really just consuming our repository,
not developing in it.
4. hg-git for a Mercurial interface for those who prefer to work with Mercurial.
   There are tools for going the other way, but this appears rather
more robust and featureful.
5. Rebasing for cleaner history.

Non-technical considerations:

1. The popularity of git+github in the numpy community and beyond.

--
Robert Kern
Enthought
_______________________________________________
Enthought-Dev mailing list
[hidden email]
https://mail.enthought.com/mailman/listinfo/enthought-dev
12