Core Toolchain Infrastructure - Update on glibc service SOW

[Zack Weinberg]

On Fri, Jul 5, 2024, at 5:42 PM, Carlos O'Donell wrote:
> On 6/30/24 3:45 AM, Sam James wrote:
>> I have to ask at what point this process stops one way or another. I
>> don't really see evidence of consensus any time this comes up.
>
> The process of switching infrastructure, for me, would stop when we've
> supported all of the GNU Toolchain moving to paid services for the
> stated goals of CTI.
>
> Some part of the community are vocal opponents to the change in
> service providers, just like some were vocal opponents to the
> adoption of DCO.
>
> I hear those voices, but for the benefit of the project I will act to
> have glibc switch to CTI services because I'm empowered to support
> that change for what I evaluate to be the good of the glibc project
> and the GNU Toolchain.
>
>> What would make it clear to you whether there is, or isn't,
>> consensus?
>
> A clearly communicated problem from a concerned interest would show
> that we don't have consensus.
>
> Right now I see that some in the community do not like this change,
> but I need to evaluate beyond that position for the sake of the
> project.

As a very occasional contributor who will not be affected regardless of
what happens, I've been staying out of this argument, but at this point
it looks like it's careening towards a process failure of catastrophic
proportions, the kind that leaves people permanently not speaking to
each other and key contributors dropping out of the project.

I have not been able to attend GNU Toolchain in-person meetings for a
long time, so everything I know about this is what's been posted on this
mailing list. That is also the position that external observers who are
making decisions about, for instance, whether glibc is a project that
can be relied upon as essential infrastructure for *other* projects,
will be in, so I think it is reasonable to assess the state of the
process from this perspective.

I do not see consensus in favor of the move.

I see sustained objections from Andrew Pinski and Paul Eggert, who are
key contributors, and also from Alexandre Oliva and Alfred Szmidt, who
are not contributors as far as I can remember but have important roles
within the larger GNU organization.  There may be other sustained
objections that I missed.  Several other people have also expressed less
strongly held concerns, which may or may not have been adequately
addressed in their opinion.  Carlos, I understand that *you* think you
have done everything that is reasonable to address all the concerns that
have been raised, but that's not enough for consensus building -- it's
necessary for the *objectors* to believe that you have done everything
that is reasonable.

If I were the person organizing this move, I would consider it necessary
to get the *positive buy-in* of everyone who has raised a hard objection
before proceeding any further.  To ensure that the people who raised
less strongly held concerns also feel completely heard, I would then
bring the entire matter back to the community for a straw poll.

Since a critical aspect of all the objections I've seen has been that
this is a stitch-up -- that a decision was made among a closed group and
that group is now forcing it upon everyone else, with no true attempt to
seek approval from everyone involved -- I would conduct all further
discussion on the matter *exclusively* on this mailing list. No more
synchronous meetings of any kind.  No more private asynchronous
conversations.  There could perhaps be *one* shared, communally editable
document on the glibc wiki, or somewhere else 100% public, laying out
pros and cons.

> The next best question is: How are decisions made without community
> consensus?

This particular decision must not, IMO, be made in the absence of
community consensus, lest it split the community permanently.

zw