Melodic has been supported on Debian for a while now, albeit only as “recommended”, and this has been great. Meanwhile ROS 2 releases have focused on supporting the most recent Ubuntu LTS only, with Debian left as tier 3 support, which I have found to often lead me into dependency hell.
With ROS 2 targeting commercial uses, and Debian being more suitable for use in products than Ubuntu in general due to the way it is set up, I’m wondering when Debian will become a tier 1 platform for ROS 2? We’ve had problems recently in prototype commercial applications that we have traced back to features of Ubuntu clearly oriented towards desktop systems, such as NetworkManager and kernel automatic updates. I agree that we can (and do) turn such features off, but if we were able to use Debian reliably then we wouldn’t have such features to begin with.
Has the ROS 2 core team given any thought to when they want to support Debian as tier 1?
We (Open Robotics / ROS 2 core team) don’t have a strong opinion one way or the other, except to recognize that each additional Tier 1 platform adds effort for everyone involved in the project, both during regular development and pre-release QA. So it’s not something that should be done lightly, or without a strong commitment to keep it working.
So I’d turn that around to you and others in the community and ask, when are you ready to contribute the resources (primarily a mix of upfront and ongoing engineering labor) to make it happen?
If that time is “now”, then the first step would be to move from Tier 3 to Tier 2. A good example to follow would be what happened during the Dashing cycle for armhf / ubuntu. In that case a group that was very motivated to upgrade the support level for that platform came to us with a proposal to make it happen, including helpful context, including that (a) they’d already tested on that platform themselves and found things to be working well and (b) their team would be available to address issues that arise. We brought up new armhf nightly jobs on the farm and since then we’ve been collaborating with the folks who wanted the upgraded support to get things working and to keep them working (studious observers will have noticed that our initial bringup of the armhf nightlies were actually just arm64 by a different name, but we’ll get that fixed :).
That sounds like a proposal I might be able to make internally here. Let me talk to relevant people and see if we have resources and skillsets available.