2011/6/7 Ruben Smits <ruben.smits@mech.kuleuven.be>

> On Tuesday 07 June 2011 11:08:12 Peter Soetens wrote:
> > On Friday 03 June 2011 09:23:36 Adolfo Rodríguez Tsouroukdissian wrote:
> > > On Fri, Jun 3, 2011 at 8:40 AM, Ruben Smits
> >
> > <ruben.smits@mech.kuleuven.be>wrote:
> > > > Hi Peter,
> > > >
> > > > I'm also forwarding to ros-mailinglists, to make sure we don't miss
> > > > any
> > > > feedback from that community ;)
> > > >
> > > > On Friday 03 June 2011 00:16:17 Peter Soetens wrote:
> > > > > I'd like to propose to restructure the orocos_toolchain_ros such
> > > > > that
> > > > > new&existing users can more easily find their way. It's mainly
> > > > > about
> > > >
> > > > > renaming packages:
> > > > This is indeed a issue that's waiting for a proposal.
> > > >
> > > > > 1. rtt_ros_integration -> rename to 'rtt_rosnode'
> > > > > -> an import("rtt_rosnode") makes your process a ros node
> > > >
> > > > Looks ok to me
> > > >
> > > > > 2. rtt_ros_integration_xyz_msgs -> rename to 'rtt_xyz_msgs'
> > > > > -> shorter notation, also makes it easier for users to update
> > > > > their
> > > > > manifest file, just prefix with 'rtt_'
> > > >
> > > > Or make rtt a suffix? xyz_msgs_rtt?? And maybe even put them in a
> > > > seperate stack
> > > > (We only provide typekites for the common_msgs stack)
> > > > -> common_msgs_rtt?
> > > >
> > > > > 3. rtt_ros_param -> rename to 'rtt_rosparam'
> > > > > -> consistent naming scheme, service is also named 'rosparam'
> > > > > and not
> > > > > 'ros_param'
> > > >
> > > > Look sane to me.
> > > >
> > > > > 4. rtt_ros_service -> ?
> > > > > -> a bit confusing about what it does, I wonder if the code
> > > > > shouldn't
> > > > > belong in rtt_rosnode instead, since it only provides the
> > > > > ros.topic()
> > > > > operations, which make only sense when running in a rosnode... I
> > > > > would also propose that this global 'ros' service is available
> > > > > from the moment rtt_rosnode is imported. Today you need an
> > > > > extra
> > > > > 'require("ros")' in scripting and something similar in lua.
> > > >
> > > > Maybe we could put the functionality of rospack, rosparam and the
> > > > rtt_ros_service, all in the rtt_rosnode package?
> > > >
> > > > > What do the current users/devs think ?
> > >
> > > +1 for succinct names, rtt_rosnode sounds fine.
> > >
> > > +1 for merging rosnode, topics, services and parameters in a single
> > > package. It makes sense to have a single package expose functionality
> > > available in ROS through a single entity, the ros::NodeHandle class.
> > > Also, +1 for allowing to import all of the functionality with a single
> > > statement. If there is  a significant overhead or bloat if you only
> > > want to use part of it (e.g., only topics), then also provide
> > > finer-grained import statements.
> > >
> > > +1 for separating messages into different stacks. This will definitely
> > > prevent dependency bloat. Again, I would aim for parallelism with the
> > > structure of the original ROS stacks/packages (rtt_common_msgs <->
> > > common_msgs). This will minimize mental transformations when mapping
> > > things between the Orocos and ROS worlds.
> > >
> > > Finally, if we're using rtt_* as prefix in most places, I'd rather
> write
> > > rtt_common_msgs than common_msgs_rtt. I'm open to be convinced
> > > otherwise,
> > > though.
> >
> > I agree here too. Let's stick to one prefix, and not start mixing with
> > suffixes.
> >
> > Target release for these changes ? We're in 0.x, so it can happen at any
> > time, but we should/must stick to 0.x == 2.x version mapping for clarity.
>
> I would like to do one more release without the renaming, 0.4.0
>
> We could do this in 0.4.1, but that seems odd, but I also do not want to
> wait
> until 2.5 comes out.
>
> Does anyone have a countervote to do it in 0.4.1?
>
>
ok for me


>
> > Peter
>
> -- Ruben
> --
> Orocos-Dev mailing list
> Orocos-Dev@lists.mech.kuleuven.be
> http://lists.mech.kuleuven.be/mailman/listinfo/orocos-dev
>