On Thu, 2012-04-12 at 03:07 +0000, Edwards, Shaun M. wrote: ...
> Is anybody open to this idea?  Is this a good idea?  Has anybody else
> seen/expressed this need?  Is ROS at the point where this makes sense?
> What do you think?

We are shipping ROS on DVDs and have been working through parts of this
problem space. So overall I think this is a good idea, but I think it
might be a little early to start working on an enterprise
long-term-support version of ROS.

I think a good first step is to focus on the hardware drivers. This
works through many of the issues and breaks the problems down into small
enough pieces that they become tractable. A hardware certification
program would be helpful for manufacturers and customers to ensure that
the hardware they buy can work with ROS and will also help build
awareness of ROS. It also encourages manufacturers to become involved
and dedicate resources to maintaining ROS drivers for their hardware.

I believe that the following three tier certification program would
serve both commercial and research interests while being opensource
compatible. As a side note, each of these would need some sort of logo
that manufacturers can put on their site.

* 'Designed for ROS' - A self certification that the product should work
with the current version of ROS.

* 'ROS Compatible'   - A self certification that the product is capable
of running a community maintained compatibility test and provides a
minimum set of features. This ensures the messages are correctly
produced so that the device is compatible with algorithms that depends
on those topics, services, etc.

* 'ROS Certified'    - A third party certification that ensures the
above and that the device can operate in a safe and stable way with some
quantifiable measure of reliability. This may be a good place for a ROS
foundation to focus on. The foundation could charge to test closed
source hardware and offer free testing for popular opensource hardware.

If we start with certifying hardware this will lay the groundwork
towards certifying algorithms and I think the process of developing a
stable enterprise version of ROS will become much easier. 
One other area that may be a good place to apply effort is to develop
benchmarks to measure the performance of a given hardware configuration.
This will help troubleshoot practical issues. For example, where a
camera might have a higher frame rate plugged into one USB port than
another. The end result of this could be some sort of ROSMark score that
could be used to determine which laptops provide the best performance
for ROS. This could also be useful for a mapping algorithm to specify a
minimum suggested score for realtime performance.

As for this, we have the beginnings of a rough pluginlib benchmark
framework and would be interested in collaborating to make this
generally useful.
https://github.com/IHeartRobotics/iheart-ros-pkg/tree/master/ihe_benchmark

-- 
Bill Morris <bill@iheartengineering.com>
I Heart Engineering
http://www.iheartengineering.com
<3