I've tried it out and it's very convenient.
I found a few usability issues. The visibility of the ROS_WORKSPACE
variable is low. I created a workspace in one directory, then went into
another. I called init and then add and it added it to the first, which was
very non-intuitive. I think this could be fixed with putting some printouts
in the setup.sh/bash scripts which state what they are doing. Also a nice
"Now source your setup.(ba)sh" instruction before rosws init quits would
help.
The other issues I had was that I added navigation and then pr2_navigation
and when I removed navigation pr2_navigation went away too. This is
indicitative of not actually knowing what the users wants, just the state of
the system. Also related to this is that it implicitly relies on having the
stacks preinstalled in the binary tree.
In thinking it through there are three inputs the user needs to provide to
define their workspace. The list of stacks they want from development
branches(currently done by adding), the list of stacks they want from
released sources(using --released option, useful for providing patches), and
the total list of capabilities which they want available(currently implied
by the binary installation). From these three lists the workspace can be
constructed.
The other two elephants in the room are support for custom rosinstall
snippets and for binary package installation. Right now the system relies
on the binaries already being present and the source system just overlaying
on top. This could be extended to support anything in the distro file
relatively easily, and install the binary packages underneath if they are
available.
As for custom rosinstall snippets I don't think need to be handled
explicitly but we need to make sure not to clobber the custom rosinstall
snippets when "deleting" elements.
Thoughts?
Tully
On Tue, Jul 19, 2011 at 1:58 PM, Brian Gerkey <
gerkey@willowgarage.com>wrote:
> Ok, I changed the name to `rosws`, queued it up for inclusion in the
> next `rosinstall` release (0.5.17), and documented it here:
>
> http://www.ros.org/wiki/rosws
>
> Further feedback is most welcome.
>
> brian.
>
> On Wed, Jul 13, 2011 at 2:17 AM, Brian Gerkey <gerkey@willowgarage.com>
> wrote:
> > On Wed, Jul 13, 2011 at 1:50 AM, Geoffrey Biggs
> > <geoffrey.biggs@aist.go.jp> wrote:
> >> It needs a shorter command. rosws?
> >
> > hi Geoff,
> >
> > I like rosws, too. My only concern is that there's already a thing
> > called rosws, which is ROS over websockets. So I was worried about
> > the name collision. But it's likely worth it to shorten the command.
> >
> > brian.
> >
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
--
Tully Foote
Systems Engineer
Willow Garage, Inc.
tfoote@willowgarage.com
(650) 475-2827
(text/plain)