[Robotics] Code repository?

[Herman Bruyninckx]

On Mon, 27 Apr 2009, Séverin Lemaignan wrote:

> Back on the repository question.
>
>> Maybe we should put a bit more effort in "design" before such a repository
>> is created. My personal suggestion is to think about where we want to go
>> with this repository in the medium and long term, and structure the
>> repository accordingly. I think that the "long term" goal is a complete
>> vertical robotic software stack, comparable to the ones that exist for
>> mobile telephony (Limo, Moblin, Android, ...) or for web services (the
>> Apache ecosystem being a major representative). "ROS" from Willow Garage
>> _is_ such a vision, LAAS' "OpenRobots robotics framework" is another one,
>> so maybe we should see how the repository that we think of would fit in
>> their code tree structures and policies, such that later integration will
>> become easier...
>
> I agree, of course. The problem is that I think it's pretty hard to
> figure out _now_ how all the pieces will fit altoghter 
Well, it _is_ possible to come up with design that do not make many
assumptions about how exactly components will be interacting at runtime.
That's the core criterium behind "component-based development' :-)

On the other hand, it is also not so difficult to identify _now_ a set of
things that _have_ to be standardized, and that are _mature_ enough to make
a suggestion for such a standard without loosing sight of all (technical)
aspects: all kinematics and control algorithms for example.

> (btw, it could be
> worthy of summarizing on the wiki the emails exchanged over the last 2
> weeks, since a lot of insightful ideas where expressed).

> Herman, as soon as the project become a bit more precise, we can move
> everything to a better structure, easier to integrate into other
> robotics "stacks" (this idea of stack is extermely interesting, and at
> the same time, probably quite far away :-)

If we take, once more, the easy road, and just put source code in a
repository without the integrated "stack" vision, it is not difficult to
predict where we are going to end: with lots of code that cannot be used
together at all. And people will start putting effort in extending parts of
that non-interoperable code, which means that the desired goal is only
being delayed even more...

> and I need a code hosting
> solution now, to exchange patches and chuncks of (beta) code).

> Concerning solutions for code hosting, I was a bit fast last time
> concerning what the LAAS could offer, and it appears we won't be able to
> set up rapidly a GIT repo with numerous external contributors (mostly
> because of the technical constraints of our authentification
> architecture). Till then, I've set up a branch on Launchpad (reusing the
> "OpenRobots" project I already created some weeks ago: don't take it as
> an attempt to cannibalize common efforts! If someone want to create a
> dedicated "blender-for-robotics" project, I'll be happy to move the code
> there).

I support this approach.

> If you want to checkout/commit, first register on Launchpad, then add
> yourself to the "blender-for-robotics" team
> (http://launchpad.net/~blender-for-robotics), then install Bazaar (sudo
> apt-get/yum/... install bzr) and checkout the code:
> bzr checkout
> bzr+ssh://<login>@bazaar.launchpad.net/~blender-for-robotics/openrobots-simulator/trunk
>
> (more details here:
> https://help.launchpad.net/FeatureHighlights/TeamBranches)
>
> Currently, I've only put the few Python scripts I used for the
> Yarp+Blender little demo + the corresponding Blender scene.
>
> By the way: does someone is in touch with ROS developpers? are they
> interested in the project? in some kind of integration? It's great to
> have Paul amongst us for Yarp, I'd like to have someone from ROS (or
> other middleware) to ensure, as you said, Herman, "middleware agnosticity".

I am not aware of active ROS involvement. But I will probaly visit them
next month or so, and try to discuss the relevance of Blender.

Herman