Who controls glibc? [LWN.net]

Who controls glibc?

Posted May 8, 2018 16:57 UTC (Tue) by k8to (guest, #15413)
In reply to: Who controls glibc? by dractyl
Parent article: Who controls glibc?

It's important to remember that consensus is not a majority vote.

If all you do is "most people want X, we do X", then you actively leave other people out in the cold and foster discontent. This can be minimized if everyone accepts this is the governance model, but that's not at all what consensus means.

It's apparent that at least some project members do not agree with the removal. The ideal next step is for all parties to work to better understand the nature of the disagreement, and for individuals to consider their reasons for their viewpoints in light of the viewpoints of others etc. Essentially, in consesnsus process, the decision is not made, and may take some time to achieve.

That said, there are definitely some participants who are not really trying to follow that model here.

to post comments

Who controls glibc?

Posted May 10, 2018 9:33 UTC (Thu) by dractyl (guest, #26334) [Link] (7 responses)

The glibc wiki has a whole page devoted to illustrating the concept (https://sourceware.org/glibc/wiki/Consensus), but the definition of consensus is only important if that's the model you're working under.

If it's not actually a consensus driven model and RMS can assert his authority whenever he pleases without regard to the option of others, which is clearly what he thinks in this case, then the definition of consensus doesn't matter either.

Having looked over some of the (rather impressive for an OSS/FS project) GNU governance documentation, RMS is probably technically in the right, although I didn't do a deep dive. I mostly base that opinion on a) that there is formal and detailed governance documentation, and b) that he's the sort who knows the rules inside and out (probably wrote most of them) and wouldn't say what he did if he didn't think he could back it up in triplicate. Nonetheless, I'm sure that will be part of the ensuing discussion.

That said, legal and wise are not necessarily the same thing. This seems like an odd thing to rarely assert one's authority over, at least from a practical standpoint. I'm sure he at least partially sees this passage as illustrating some principle of free speech, but there are many other opportunities for such expression that don't come with in-fighting and risk of escalation attached.

He likely has the authority to do it, but it seems like a bit of a bad call.

Who controls glibc?

Posted May 11, 2018 16:19 UTC (Fri) by civodul (guest, #58311) [Link] (6 responses)

I don't think it's a matter of "right": there's no "law" governing what RMS can and cannot do within the project.

However, as a GNU maintainer, I disagree with this authoritarian governance model. It's at odds with the way many (most?) GNU packages work on a day-to-day basis, focusing on consensus similarly to what glibc does. Many GNU packages do provide a welcoming environment. What RMS is doing here is painting GNU as an unwelcoming project, one where it's okay to dismiss other people's views; *that* is something I cannot accept.

Free software now goes far beyond GNU, but I think GNU still has a key role in the movement; we shouldn't let such behavior harm it.

Who controls glibc?

Posted May 12, 2018 4:30 UTC (Sat) by dractyl (guest, #26334) [Link] (5 responses)

This goes back to my offhand remark in my first post: English nobility vs the King.

RMS clearly believes he's the King and that being King means he can enforce his will on everyone else if he chooses to do so. He can probably even point to a document saying that's so. The fact that he doesn't do it very often is irrelevant. Indeed, he's said just this, only with slightly different words.

You disagree, as do many others with you. You also have a document of sorts ("uses a consensus-based community-driven development model"), but more importantly the reality of day-to-day contemporary practice built on the culture and beliefs of the active participants.

Long leash dictatorship, or consensus governance?

That is the basis of the situation. Everything else is fluff; a distraction.

In the OSS/FS world, the inevitable conclusion of failed negotiations over political differences is the fork. Then the matter is decided by the very democratic method of the developers choosing sides once the schism has formed. I'm fine with that. That seems very much the OSS/FS way.

I am reminded of a principle in negotiation: BATNA, or Best Alternative To a Negotiated Agreement. This is the principle that lets apparently weak individuals crush multinationals at the negotiating table. It talks about what happens if no agreement can be reached through discussion. Talks are over, there's no deal. Then what happens?

Someone else mentioned that GNU needs glibc more than glibc needs GNU. This is true.

RMSs BATNA is the loss of a major, high-profile project when the developers fork, not to mention loss of face, influence and reputation. It would be hard to qualify the loss of glibc as anything other than a huge net negative. The problem is that I doubt RMS will see it in this light, and thus will not negotiate accordingly. He might eventually find, to his lasting discomfort, that yesterday's god is today's out of touch has-been. I also don't imagine he ever truly envisioned the freedom-guaranteeing power of the fork being used against him personally in this way or at this scale.

The dissenting developers' BATNA, assuming they can maintain cohesion, is they probably lose a minority of talented developers, waste some time coming up with a new name and laying down new infra, and otherwise coping with temporary disruption. If they were smart, they'd also formalize their new governance, incorporating lessons learned. Other than that, they lose little; life proceeds much as before. There are several high profile examples of this happening in the history of OSS/FS.

The distributions' BATNA, as outsiders, is to switch their glibc upstream. I don't imagine that would be especially problematic for them. If they did systemd, they can manage this.

I doubt the commoners, such as myself, will care at the end of the day, or indeed even know. We're just too far removed from palace intrigue for it to make a difference.

Normal politics would muddle through with some kind of kludge or lower the heat until everyone moved on; perhaps with the help of a well timed distraction. I'm not sure that'll the case here. The point of contention is too clear (you can't smudge it up, or say more time is needed to understand the problem) and RMS just doesn't strike me as the kind of person to back down at all.

I think, overall, the community zeitgeist favors the democratic approach and RMS has not built up sufficient goodwill beyond his core supporters to overcome this in the way someone like Guido or Linus might.

Who controls glibc?

Posted May 17, 2018 10:56 UTC (Thu) by Zolko (guest, #99166) [Link] (4 responses)

"The distributions' BATNA, as outsiders, is to switch their glibc upstream. I don't imagine that would be especially problematic for them. If they did systemd, they can manage this."

I know I'm late to the party, but I can't let this one sail: if the result of forking glibc is as successful as the catastrophe that systemd is, and considering the Devuan fork, it is quite clear that the oldtimers (including Stallman) are right. Technically speaking a fork is probably possible, but because of copyright issues the nightmare would be huge, in no possible comparison to the small virtual benefit of removing a potential negative emotional effect.

In other words: this is an easter egg, FFS !!! The kind of things that make life funny and enjoyable for must of us. This has provable positive emotional effects on many people: it makes us smile ! And nobody here has actually come out and shown a provable negative effect, only some hypothetical, potential, made-up stuff.

The GNU project had political motives from the beginning, nobody can now come and pretend they didn't know. Reversing that political motive should not be allowed with a one-line diff.

Who controls glibc?

Posted May 17, 2018 11:38 UTC (Thu) by excors (subscriber, #95769) [Link] (1 responses)

Excellent, I was worried this article wasn't going to reach 200 comments even with the culture war stuff, so a systemd flamewar is absolutely what we need!

(In other words: Perhaps you should have let this one sail, it's a distraction from the more relevant topic of glibc's governance.)

Who controls glibc?

Posted May 17, 2018 13:09 UTC (Thu) by sdalley (subscriber, #18550) [Link]

I *think* Zolko was just being heavily satirical...

Who controls glibc?

Posted May 19, 2018 10:27 UTC (Sat) by dractyl (guest, #26334) [Link] (1 responses)

(Sorry for the delay in my reply; I'm traveling ATM)

I don't see what copyright problem would be here, but perhaps I am not up to date on GPL legalities.

IIRC glibc, under GNU policies, has copyright assignment as part of their process. Therefore there aren't a million little copyright holders all over the place. It doesn't seem especially messy in this regard.

GNU licenses the software (shockingly) under LGPL, so an outside party has every right to fork it.

Therefore, if a new entity is created, they can fork it under the GPL in a very tidy way and since they're bound by terms to keep licensing it the same way, there is no downstream change. I never envisioned licensing as a major issue when I first wrote my comment.

I was referring to potential logistical issues; downstream disruption. But if mostly the same people are woking on the same code, with the same license, I can't see an issue there either. I see XFree86 and maybe LibreOffice as examples of this pattern.

All in all, the barrier to action arising from disruption for the glibc developers is low.

As for systemd, I am asserting no opinion of it's quality or suitability for purpose one way or the other because it's not relevant to my point. I used systemd merely as an example of something that necessitated a lot downstream changes in order to accommodate it.

The point of whether the joke was appropriate or not, whether is was funny or not, whether someone might be offended by it or not or whether it improved the world with levity or not, isn't relevant or material to the underlying issue in any way. It could have just as easily been some arcane compiler optimization.

Whether the GNU project's politics (as in policy and governing structure, rather than philosophy) are still relevant and acceptable to the governed is a key issue to be resolved. I don't see how you can leave it unresolved now that it's come up.

This is not about reversing GNU's fundamental raison d'être. I don't see "RMS is our BDFL" being a core tenet of the Free Software ideals upon which it was founded.

If the organization is ultimately unable to evolve and mature, someone will come along to pick up the torch and run with it. Perhaps that someone, individual or group, will have a stronger commitment to democratic ideals as well.

It would be a shame if that became necessary, but at some level, isn't that part of the beauty of Free Software?

Who controls glibc?

Posted May 20, 2018 2:23 UTC (Sun) by tdalman (guest, #41971) [Link]

dractyl,

thank you very much for sharing your thoughts as they match very well with my own.

Being a watcher of the glibc mailinglist for quite some time, I have seen a very positive transformation of this community in the last years. Carlos, Siddhesh, Joseph, and the others are doing a great job not only technically but also by are carefully leading people through constructive discussions.

I was very surprised to see such a vehement discussion virtually on non-topic (technically, that is; I understand the political intent by Richard, as well as the arguments against the "joke" being part of the manual). Likewise, I was surprised by Richards dominant decision because he seemed to play no role in the glibc mailinglist for years (IIRC, the only thread he really participated was about a malloc hook to be removed and its use in emacs, but maybe I missed something).