I’m suggesting the creation of a new release for ROS1 after noetic. I’m even suggesting the name “Obvious OwlOrlitia”.
I think that a good release ambition would be:
100% API compatibility with noetic
design to be “rolling release” from now on since it is community effort and without EOL
refactor on the core components to facilitate future changes without breaking ABI (for example use of d-pointer like in KDE or Qt), this supports the previous goal
deliver as many pending improvements as possible
the ROS1 to rule them all basically
Anybody else interested in making this?
And no “switch to ROS2” / “ROS1 vs ROS2” / etc here!
Supporting Ubuntu 22.04 a year from now is something I’ll be interested in- initially I’ll install it on a computer I use for development and will want to work on software that continues running on noetic in production. I expect minor fixes that allow compiling on 22.04 while not breaking 20.04 will get merged into noetic until EOL, but it otherwise it serves as a half-step test case.
I’m fine with compiling everything from source and would appreciate (and contribute to) a loosely coordinated effort to maintain branches somewhere on github with all the fixes which don’t make it into noetic, but would the aim here be binaries at least for the the core packages? (Could those go straight into Ubuntu? Maybe that is a longer term goal, looks like a lot of work, probably some history and discussion about previous efforts there to look over)
Merging in Unix Domain Socket Support or the observability changes from Formant would be interesting, I’m not sure how breaking those are.
Having a rolling release is great as a concept, but you have to baseline off of something. Unfortunately the size of Ubuntu or common distros is too large for a real robot deployment or fleet management. This is why I used and have used yocto to build custom operating systems from source. Having a rolling release compiled from source based upon a supported OS that has flexibility in its dependencies is important to me.
ROS distros are usually named after turtle species, so perhaps Orlitia instead of Owl?
More seriously though, I’m not sure that extending the life of ROS1 is the right solution. The risk of further dividing the ROS community into 1 vs 2 is too high a price. I think, as @adamsj pointed out, it’s all about helping companies currently using ROS1 to incrementally migrate to ROS2. It’s probably not the length of the migration path that matters, but the size of the biggest step within it. So any efforts that help smooth the transition and reduce the risk of migrating to close to zero would probably be very welcome by these companies and result in greater adoption.
From my understanding of the robotics industry, there is a ton of interested users. Do you have the resources to actually do it though? As Open Robotics has stated many times, this stuff is expensive to release and maintain. Open Robotics and most of the maintainers of ROS have rightly shifted their focus to ROS 2, in order to fix the blaring issues that everyone complains about in ROS 1.
Indeed! In my opinion a big mistake of the ROS 2 development strategy was rewriting everything from the ground up, rather than having every new distribution of ROS change approximatly one major feature of ROS. Total hypothetical that I’m sure will not happen due to limited resources, but what if ROS-O used colcon/ament and the new ros2 launch, but everything else stayed ROS 1? This would be a nice stepping stone to being on the ROS 2 bandwagon, and how ideally ROS 2 had been rolled out.
But what’s done is done, and I think the best strategy is to bite the bullet and upgrade to ROS 2.
Is there a gap/void left by Noetic that Orilita would bridge? The ROS2 transition from Open Robotics was huge, and not fun for people to transition to, but is there a blaring problem to address with Noetic being the last ROS1 distribution? Except of course community-driven LTS for Ubuntu 22.04
I agree that simply releasing Noetic as-is or in some “rolling” format on Ubuntu 22 would probably be least-cost and highest reward option. As it currently stands, ROS2 is still not all that close to ROS1 in terms of feature set and convenience tools. Giving people an option to use Ubuntu 22 with ROS1 while ROS2 continues to mature would be helpful.
EDIT: It would also be good from a ROS2 adoption/migration standpoint. ROS1 users using Noetic on Ubuntu 20 would have to upgrade to Ubuntu 22 on top of trying to migrate to the latest version of ROS2.
It would make ROS migrations easier to be able to run both ROS1 and ROS2 on Ubuntu 22, which would allow the use of the ros1_bridge, etc.
