git, future directions - ESOE Development

Message from discussion git, future directions
Received-SPF: pass (google.com: domain of bedd...@intient.com designates 209.85.219.161 as permitted sender) client-ip=209.85.219.161;
Message-Id: <93B07508-1163-48CA-8060-468E602E26AA@intient.com>
From: Bradley Beddoes <bedd...@intient.com>
To: esoe-dev@googlegroups.com
In-Reply-To: <90ad28f40902282324s1fea9e1aj5ff753acafc15efa@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: quoted-printable
Subject: Re: [esoe-dev] Re: git, future directions
Date: Thu, 5 Mar 2009 10:49:15 +1000
References: <EF21CAFB-258E-4DFB-8986-85981E155D03@intient.com> <BE7C2473-6DC7-4EB3-B9B4-EAC63FEF8A87@gmail.com> <90ad28f40902282324s1fea9e1aj5ff753acafc15efa@mail.gmail.com>


On 01/03/2009, at 5:24 PM, Shaun Mangelsdorf wrote:

> No argument from me. The only thing I would be concerned about is
> making sure that the tooling supports it. The project needs continuous
> integration and a repository inspector, and ideally an issue tracker
> linked into both of these. I have used the Atlassian tool set to do
> this in the past which works well, but I don't think they are
> supporting git yet=85 yep just checked and they still aren't though they
> plan to.
>
> Although I see your point, I think we can find suitable substitutes =20
> that (while not as tightly integrated as the Atlassian suite) will =20
> give us what we need and support git natively.

Hudson is a good example here. Where possible I'd like to stick to =20
other open source solutions to support the project. We've been using =20
Hudson in the past with great success and flexibility.

>
>
> This might warrant some further investigation. If it turns out that =20
> git is too widely unsupported, we may need to stay with subversion.

Hudson at least seems to have pretty decent GIT support
http://wiki.hudson-ci.org/display/HUDSON/Git+Plugin

>
>
> It makes sense to me to move this project to Codehaus or even Apache.
> Although Apache is quite strict on the projects they host so Codehaus
> might be a better choice. They have all the tools and admin interfaces
> to manage them, not to mention free bandwidth. It also adds a certain
> aspect of credibility to the project by having it hosted by a well
> known hoster of open source projects. Plus, it allows the developers
> to focus on developing rather than installing and configuring =20
> software.
>
> We'd have to look into the copyright issues surrounding the =20
> ownership of the code. Apache for one requires that the code be =20
> owned by the foundation.
>
> Not sure what Codehaus' barrier to entry is like, but on first =20
> inspection they seem to have the Atlassian suite set up, so it's =20
> worth considering what we want to do for SCM before we decide where =20
> to host the project long term.

I've had at length discussions with the Apache folks before we got =20
semi close but there are some major blockers that we couldn't get =20
around, particularly in the C++ side.

I for one am actually happier with the project standing on its own, =20
the less we need to deal with other groups political processes the =20
more time we spend on the many various facets that make up ESOE.

>
>
> I am not a hardcore git user so I may not have seen the light yet, but
> I would rather stick with svn because of the support for it from
> FishEye and Bamboo for the time being. Individual developers can
> always use git+svn locally to get _some_ of the features.

I'd really rather one or the other and not be trying to support some =20
hybrid. Common sense would have to tell you that eventually that kind =20
of arrangement will come unstuck and cause headaches.

>
> git+svn would be an option if we're stuck and can't find the =20
> necessary tools, but I think a move to git would be good for the =20
> project as a whole. Git's branching and merging features (among =20
> others) are second to none in my experience; and there are a great =20
> deal of features which will benefit a project under distributed =20
> development, as ESOE is and will undoubtedly continue to be.
>

Totally agree.

Bradley

> Keen to hear more thoughts on this.
>
> Shaun
>
>
> >

Bradley Beddoes
Lead Software Architect
Intient Pty Ltd

http://www.intient.com
Twitter: @bradleybeddoes