[CentOS-devel] Before You Get Mad About The CentOS Stream Change, Think About…

On Wed, Dec 16, 2020 at 2:29 AM Neal Gompa <ngompa13 at gmail.com> wrote:
> On Wed, Dec 16, 2020 at 2:18 AM Mark Mielke <mark.mielke at gmail.com> wrote:
> > > The vast majority of self-supported deployments would be better off choosing CentOS Stream, and having both makes that a lot less clear.  Producing CentOS is very expensive, and provides no value for the vast majority of users.  It provides no value to Red Hat, either.  And saying so doesn't mean that Red Hat is making a cash grab.
> > If this was true - then RHEL could also abandon minor releases. This
> > is not how vendors certify that their products work with RHEL, and it
> > will be a major problem for CentOS 8 Stream, just as it would be a
> > major problem for an RHEL 8 Stream (which is what CentOS 8 Stream
> > should be!). I don't believe that the CentOS board or the Red Hat
> > management team are innocent and unaware of this. Messaging such as
> > "if you require a stable release, you must buy RHEL" makes it clear
> > what is really going on here. CentOS 8 is being eliminated as part of
> > a determined business strategy. It is predatory. Allowing CentOS 8
> > Stream to exist only if CentOS 8 is destroyed, under legal threat, is
> > essentially defeat. It is saying "you can only exist if you do not
> > provide the same product we do".
> Well, I would certainly be happy if Red Hat dropped minor releases
> from RHEL. Officially, Red Hat advises ISVs to not target specific
> minor releases already, and that compatibility is assured by following
> reasonable practices and the documentation around ABI guarantees in
> RHEL.

The problem is that 10 years is a very long time to require both
source and binary, forwards and backwards compatibility, and without
point releases - you essentially have no ability to introduce changes
to interfaces. And this includes, no ability to upgrade important
packages. It might be possible with some system whereby packages are
built statically, or multiple versions of libraries (both source and
binaries) are installed in parallel, or containers are used to run
programs in the environments they were designed for. But, that's not
what we have with RHEL or CentOS today. What we have today, is a
system where feature changes are introduced in new minor releases, and
important patches are introduced in channels. Point releases exist for
an important reason. They might introduce effort for contributors, but
they also reduce risk for users, and they provide a baseline for
certification.

I think *if* RHEL abandoned minor releases, then the CentOS problem
would disappear as a problem on its own. However, this would then
re-introduce the problem that:

1. CentOS 8 would be a direct competitor for RHEL 8, which is very
likely the reason why this change is being introduced in the first
place.
2. Users might not agree. RHEL 8 without point releases is not a valid
option, for the same reason that CentOS 8 without point releases is
not an option. Feature changes could arrive any day, and break
Enterprises every day.

I think abandoning point releases, is basically abandoning RHEL's
bread and butter. I can't see it happening. It doesn't matter what ISV
think or which path is easier for ISV.

Also, I don't mean that judgmentally on you... it also doesn't matter
what I think. My contribution here is only to ensure a few additional
points are captured. I have no confidence that my contribution will
change the outcome. :-)

-- 
Mark Mielke <mark.mielke at gmail.com>