How will Stream work for the numerous projects and tools (some open source some other-wise) that have targeted RHEL/CentOS X.Y versions for compatibility? Things such as Lustre, IBM Spectrum Scale, OpenZFS, Mellanox OFED, Intel Omni-Path, etc. (these are ones I’m very familiar with being in HPC, but I’m sure there are others). Will there be mechanisms within Stream to set things at a release version of RHEL (maybe via a set of repos?) so that Stream can be used without having to carefully manage package versions via yum/dnf version lock, etc?
A lot of the above tools/projects don’t bring support for an X.Y version of RHEL until 1-4 weeks after the RHEL release drops, which has historically worked out well anyway for CentOS because of its inherent delay also. For instance I’m not sure I could get Mellanox drivers for the 8.4 kernel (that I understand is currently in stream). I’ve been using most of the above mentioned tools for the better part of the past 7 years and they don’t survive across an update from CentOS X.Y —> X.(Y+1) very often, if at all.
JD
On Dec 15, 2020, at 17:30, Jason Brooks jbrooks@redhat.com wrote:
On Tue, Dec 15, 2020 at 1:03 PM Jim Jagielski jim@jimjag.com wrote:
On Dec 15, 2020, at 3:50 PM, Mike McGrath mmcgrath@redhat.com wrote:
You may not like it, but the CentOS community didn't evolve in any way with the industry. When I think about the talent on this list, and in IRC, I can't help but wonder what went wrong. For whatever reason, CentOS never grew beyond a community of users
Whose fault is that? And, to be honest, I never recall such an expectation ever being vocalized during my tenure @ RedHat (FTR: I was one of the people inside OSAS who drove the CentOS "acquisition" along w/ Carl Trieloff)
The whole intent back then was "as long as there is going to be this huge community of 'free-loading' users out there, they might as well be under the RHEL/Fedora umbrella, rather than Canonical or elsewhere." I guess somewhere along the line that changed. The issue isn't that the situation changed but rather that up until very recently, promises were still being made and then RedHat backed out of those promises.
No. I was on that team too, and growing CentOS beyond just consumption and into contribution was something we emphasized throughout. Our primary intent, the reason the whole thing got started, was that we needed to provide our layered projects with a slow-moving community distro to layer atop. That's why we put so much effort into the SIGs, and into opening up the build processes and tools. Even with that work done, until we opened up RHEL development itself, contributions to the core of CentOS were basically blocked. Now, in addition to the layered project need, which hasn't gone away, we need a distro to open up RHEL development, and CentOS Stream is that distro.
I know that "same package set that RH ships to customers" has been an effective shorthand for "good enough for me," and that our assurances of trusting the automated ci and the power of community collaboration may be inspiring less trust right now, but I am confident that CentOS Stream will be a good, stable distro for current users. My team that hosts infra for various open source projects has already begun converting some of our production services from CentOS 8 to CentOS Stream.
-- Jason Brooks He/Him Manager, Community Architects & Infrastructure Red Hat Open Source Program Office (OSPO) https://community.redhat.com | https://osci.io
CentOS-devel mailing list CentOS-devel@centos.org https://lists.centos.org/mailman/listinfo/centos-devel