Re: [CentOS] Upstream and downstream (was Re: What are the differences between systemd and non-systemd Linux distros?)
On Fri Oct 19 00:52:12 UTC 2018 Japheth Cleaver wrote:
This brings to mind a video I was pointed to not long ago of Brendan Conoboy's talk at a Dojo recently:
Hey, that's me! Hi. By the way, Jim Perrin did an updated version of this talk *today* at CERN in my absence (thanks Jim!). Hopefully the video will be posted soon. I expect we'll be doing updated versions of these at Devconf, future Dojos, etc- as things progress.
For quite a long time, many (perhaps most) folks had assumed that Fedora functioned more or less directly as the internal alpha for RHEL, with a branch at some point occurring, followed by pruning of packages, hardening, vendor testing, and release.
This is roughly true for new releases (plus or minus the kernel), but not for subsequent minor release updates. It is a shame because so much great work happens in Fedora between major RHEL releases.
Subsequently, CentOS (even after the RH integration) functioned *strictly* as a clean-room downstream rebuild, with the ability to do unsupported things, like alternate architectures, or heavier kernels, restricted to what could be done while maintaining a 100% binary compatible rebuild. Any contributions back up where taken to be incidental, from CentOS users reporting bugs that could be verified against RHEL.
Conoboy, on the other hand, takes great pains during the speech to describe a much more fluid and complex interaction between CentOS and its upstream, and puts forth CentOS as a mechanism (perhaps the best mechanism) for the winder EL community to contribute (something?) back into RHEL's future. He also gives clear signals that various Fedora steps have been in directions that Red Hat did not want EL necessarily going, and that the simplistic assumptions we've commonly been making aren't really correct.
You might be reading into this more than is there. It's not so much that things are fluid as it is that they are undefined. There is no clear, consistent way for a member of the Fedora or CentOS communities, who create something great, to have that thing make its way into an update of an existing RHEL major release. Defining that path, making it possible, would be win for all.
Obviously, there seems to be a bit of a discrepancy there.
The wider EL community is trapped between a rock and a hard place somewhat. If you try to direct Fedora into the needs of EL users, you stand a good chance of getting told to pound stand, and that EL is getting in the way of bleeding-edge progress. Traditionally, CentOS has had its hands tied since it aims to be 100% compatible with upstream. Red Hat (and Red-Hat-as-a-sponsor-of-CentOS) might do well to clarify just what type of back-and-forth it wants out of the wider EL-using community. Does it want direct feedback in the form of tickets? Should people form SIGs? Obviously RHEL7 is not changing init systems, but where should one talk about the future?
Man, it breaks my heart when I read things like this. There might be some historic truth to the above, but it doesn't have to be the future. The objective I mentioned near the end of the talk has been posted, but not yet voted on:
https://fedoraproject.org/wiki/User:Pfrields/Lifecycle_Objective
The beauty of community is that it can grow and shift according to the needs of its members. To me it looks like the lifecycle objective may be a partial answer to how Fedora, RHEL, and CentOS communities can reach a state of fluidity, a virtuous cycle. The thing that makes it the most likely to succeed is if members of the Fedora, RHEL, and CentOS communities work on it together. I hope those reading this who are interested in that join in.
On 10/19/2018 9:10 PM, Brendan Conoboy wrote:
On Fri Oct 19 00:52:12 UTC 2018 Japheth Cleaver wrote:
This brings to mind a video I was pointed to not long ago of Brendan Conoboy's talk at a Dojo recently:
Hey, that's me! Hi. By the way, Jim Perrin did an updated version of this talk *today* at CERN in my absence (thanks Jim!). Hopefully the video will be posted soon. I expect we'll be doing updated versions of these at Devconf, future Dojos, etc- as things progress.
Thanks for responding!
Conoboy, on the other hand, takes great pains during the speech to describe a much more fluid and complex interaction between CentOS and its upstream, and puts forth CentOS as a mechanism (perhaps the best mechanism) for the winder EL community to contribute (something?) back into RHEL's future. He also gives clear signals that various Fedora steps have been in directions that Red Hat did not want EL necessarily going, and that the simplistic assumptions we've commonly been making aren't really correct.
You might be reading into this more than is there. It's not so much that things are fluid as it is that they are undefined. There is no clear, consistent way for a member of the Fedora or CentOS communities, who create something great, to have that thing make its way into an update of an existing RHEL major release. Defining that path, making it possible, would be win for all.
*snip*
Red Hat (and Red-Hat-as-a-sponsor-of-CentOS) might do well to clarify just what type of back-and-forth it wants out of the wider EL-using community. Does it want direct feedback in the form of tickets? Should people form SIGs? Obviously RHEL7 is not changing init systems, but where should one talk about the future?
Man, it breaks my heart when I read things like this. There might be some historic truth to the above, but it doesn't have to be the future. The objective I mentioned near the end of the talk has been posted, but not yet voted on:
https://fedoraproject.org/wiki/User:Pfrields/Lifecycle_Objective
The beauty of community is that it can grow and shift according to the needs of its members. To me it looks like the lifecycle objective may be a partial answer to how Fedora, RHEL, and CentOS communities can reach a state of fluidity, a virtuous cycle. The thing that makes it the most likely to succeed is if members of the Fedora, RHEL, and CentOS communities work on it together. I hope those reading this who are interested in that join in.
While I do believe that's important -- especially in helping to prioritize re-basing decisions, if not architectural ones, for updates -- I feel like things are still more open to interpretation for the lead-up *to* a major release. Modularity, software collections, and the like can be used alongside native EL point updates or a more flexible EPEL policy to incorporate new tech, but the impression is that by the time a RHEL beta makes it out, it's already a bit late for a community-suggested major changes. Bug reports? Yes. Design changes? Not as much. Having a stable platform OS design is a key principle for EL users, and Beta->0 seems late in the game.
(Nevertheless, the lifecycle stream discussion is absolutely one that does need to be had, and I'm glad that there's that out there!)
-jc
-
Brendan Conoboy -
Japheth Cleaver