[ros-users] ROS 2.0 Strategy review

[Ken Conley]

That's not quite how I remember it.

There were no tight business goals motivating ROS or Willow Garage.
The REP process also didn't come along well into the life of ROS, so
most of the baked in decisions were done prior to that. There were
also several open-source robot frameworks at the time, so it was
controversial to push forward with yet another framework. And we did
most of the middleware before there were any
perception/manipulation/navigation teams to speak of.

We readily assumed the middleware transport would be redone once we
had more time to go back and fix it. As open-source middleware started
to prosper again, it became obvious that we could just throw away our
hasty work and collaborate with larger communities.

It's shocking to me that the original middleware is there in any form
8 years later. It probably points to some bad design decisions
(probably by me) early on. It was easier to make bad decisions back
then because we had lots of WG resources to fix them later.

We seem to be trapped in a paradox of choice now: too many choices
over which middleware to use and choices over the development path
(tick-tock iterate in place, or migrate via bridge). With respect to
the development path, we used both to good effect over the course of
ROS. The 'bridge' strategy is probably the #1 way most people entered
the ROS community.

So, I have sympathy for the OSRF, which has been tasked with selecting
which color of paint to use (for the record, grey always works).

 - Ken



On Thu, Oct 1, 2015 at 6:51 AM, Thibault Kruse via ros-users
<ros-users at lists.ros.org> wrote:
> On Tue, Sep 29, 2015 at 7:23 PM, Bill Smart via ros-users
> <ros-users at lists.ros.org> wrote:
>> Also, I was thinking about the (perceived) autocratic behaviour of OSRF, and
>> about the origins of ROS. [...] Then Willow Garage came along,
>> asked a bunch of people in the community what they thought, and implemented
>> ROS. [...] Sometimes we have to stop talking and just do it, even if
>> it's not the optimal strategy.
>
> Hi Bill,
>
> A difference between then and now is that Willow Garage did
> not previously had any community that it could divide/hurt by creating ROS.
>
> Also Willow Garage, despite having tight business goals (10K robots
> in US households by 2015 IIRC), and despite having enough resources to
> sustain it's own software, made the effort to nurture an open-source
> community.
> OSRF, without the pressure of business goals, and desperately needing
> open-source contributions, instead ignores the open-source process (REPs), and
> plans to make a release that will divide the community into sub-communities.
>
> Willow Garage had strong and immediate pressure to provide a core that
> the perception / manipulation / navigation / supervision teams inside Willow
> Garage could easily use.
> OSRF, not having a robot plattform to produce or support, does not have
> this immediate feedback. OSRF could substitute the lack thereof by making small
> increments to ROS1 that the large ROS1 community can validate in a lot of
> real-world projects, but OSRF decided to rather validate ROS2 features with
> some hello-world packages.
>
> The behavior of OSRF does not only influence how quickly ROS2 will be
> ready (if ever) or what technical features it provides. The behavior of OSRF
> also influences how motivated anyone will feel to contribute to ROS.
> And what value has ROS without it's contributors?
>
> Nobody needs ROS2 to use DDS. There are open-source DDS libraries to use.
> Not much is gained by delivering a ROS2 that's just a paper-thin layer on
> top of DDS, but lacks the community momentum of ROS. That's why the
> behavior of OSRF and the strategy for ROS2 matter. Ideally ROS2 should
> not divide the community, and ROS2 should not make potential contributors
> feel like their opinion does not matter.
>
> regards,
>   Thibault
> _______________________________________________
> ros-users mailing list
> ros-users at lists.ros.org
> http://lists.ros.org/mailman/listinfo/ros-users