LWN: Comments on "Upheaval in the Debian Live project"

Upheaval in the Debian Live project

jschrod — Mon, 07 Dec 2015 12:12:18 +0000

> There's *no* reason to bring this topic back from the dead now. The whole thing with
> Daniel was painful for everyone, we don't need to be constantly reminded about it.

LWN.net is one of the mainstays of information what happens in Linux communities. *YOU* might have known about that "Daniel thing". Others, especially those not involved in Debian, didn't. And for them, that information is valuable if they want to know what happens in our diverse distribution communities. For me, it was.

Upheaval in the Debian Live project

ras — Sat, 05 Dec 2015 13:52:13 +0000

Depends on your definition of replacement I suppose.

An ITP adds a new packages to Debian. They do not remove existing packages. So if it goes ahead both live-build and live-build-ng (now renamed to live-wrapper because of Daniels request) will be part of Debian.

I expect the people introducing it will be replacing live-build with live-wrapper in their projects, otherwise there would not be much point. The rest of us can continue to use either, although someone will have to pick up live.debian.net if live-build is to continue to remain useful.

Upheaval in the Debian Live project

plugwash — Fri, 04 Dec 2015 14:48:52 +0000

Umm, unless i'm misreading things, the package in question is a replacement not a fork.

Upheaval in the Debian Live project

mstone_ — Fri, 04 Dec 2015 13:58:57 +0000

yes, adding a -ng is lazy: it takes far less time and wasted effort than bikeshedding a new name, by simply using an existing convention--and that's a good thing. I count more than 20 -ng packages in debian (e.g., syslog-ng, dput-ng, linux-wlan-ng, aircrack-ng, fakeroot-ng, etc.) and I don't recall any of them leading to such sturm und drang/rage quitting. There's even a grand tradition of -ng projects starting to scratch an itch, then disappearing while the original project continues on. (mutt-ng comes immediately to mind, but there have been others.)

I think it's reasonable to assume that the person who came up with live-build-ng expected past to be prologue, and was not actually trying to start something.

Upheaval in the Debian Live project

ras — Thu, 03 Dec 2015 22:55:52 +0000

> what really matters was what happened after the ITP was filed.

The Debian-CD team forked Daniel's package, live-build and called it live-build-ng. They forked it because it wasn't well suited to what they were doing, and Daniel wasn't receptive to the changes they wanted. Forks are a biggish deal in open source, but no one is forced out by a fork - the original package continues as it was. No one was threatening to take over, shut down, or otherwise interfere with Daniel's work on live. The only change is it had some additional competition.

What happened after the ITP was filed is Daniel demanded the package name be changed, and after 58 hours of the Debian-CD team saying they weren't going to do so he quit.

Spats like this crop up within Debian from time to time. Normally they brew and fester over months and months. The bad ones are sorted by the FTP team, the really bad ones by the technical committee, and all out wars as happened over systemd require a General Resolution or two, were the whole project gets to have a say.

> this was a nicely orchestrated operation by a few people to bully him out.

If that is indeed what happened, I got to hand it to them - they did it extraordinarily well. Getting someone to quit the project after 58 hours of starting is a record that will stand for a long, long time. And using a page name as the trigger! It wasn't even a fight over who gets to use a package name, it was a package name no one else was using - unbelievable.

Upheaval in the Debian Live project

flussence — Thu, 03 Dec 2015 22:28:56 +0000

Going off topic a bit: “Winners compare their achievements to their goals; losers compare their achievements to those of others” is often predictive when it comes to things like software project naming and public relations.

It doesn't apply at all in this case, but it's not hard to find correlations elsewhere between people/companies openly bashing their competition and producing bad products themselves.

live.debian.net deleted?

tzafrir — Thu, 03 Dec 2015 13:10:36 +0000

It seems that http://live-systems.org/ is likewise down.

Alternatively, see https://wiki.debian.org/DebianLive .

The git repositories are now in Alioth: https://anonscm.debian.org/cgit/debian-live/

Copies of the original repositories were mirrored to https://github.com/debian-live/

The site was built from the live.debian.net repository. As it happens, pages there seem to be just markdown, and Github renders markdown. So, for instance, the list of downstream distributions (projects using Debian Live) is at: https://github.com/debian-live/live.debian.net/blob/maste...

Can we please stop the slanging match now?

Martin.Schwenke — Thu, 03 Dec 2015 09:02:30 +0000

I simply posted a comment saying that I was unaware of any of the situation, so the article was useful to me. At this point, as a long term Debian (and Debian Live) user I will now know to look elsewhere when I find that live.debian.net no longer resolves. I will also understand that there are complex reasons behind any changes I see.

However, nobody is really gaining anything from the public slanging match that is now taking place. It sounds like this has already been discussed elsewhere - those who are interested will find the previous discussion in list archives somewhere. I don't think anyone is going to change anyone else's mind by posting more and more details here... :-)

Upheaval in the Debian Live project

petur — Thu, 03 Dec 2015 08:40:54 +0000

I've read a fair bunch of mailing list posts and bug reports, and the only feeling I get left is that this was a nicely orchestrated operation by a few people to bully him out. Blowing issues out of proportion, for example.

And as for your nice description of ITP, what really matters was what happened after the ITP was filed. And then blame Daniel for not being a team player. Yeah, right.

Upheaval in the Debian Live project

ras — Thu, 03 Dec 2015 05:48:23 +0000

> You've sort of confirmed my point, you cannot have "people who do not trust him much"and and the same team

Yes, it is true, people did not trust Daniel. This is a serious issue for Debian as it depends on around 1000 people who barely know each other, mostly doing things unsupervised, yet all these unsupervised actions must push Debian in the direction of getting a working, stable Debian release out door. Trust is the lubrication that makes it work. Without it, every Debian package upload would have to be checked by another DD which would slow things down and increase the man power required. Manpower is a limited resource, so Debian simply could not be what it is today if developers did not trust each other to get the job done. Daniel broke that trust on a couple of occasions by making and not addressing changes that delayed the release. Eg: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=699808;...

In the end Daniel was demoted from DD. It seems to be fashionable to attribute that to personality clashes, and yes there is no doubt there were and are personality clashes. But the reality is far more mundane - Debian was ejecting someone that had become on impediment to achieving a core reason for the projects existence. He wasn't ejected completely, instead he was made a DM. A DM could be said to be a supervised DD. Notice how that fixes the trust issue, while still allowing him to be part of Debian and continued with all the personality clashes that existed before. He was even trusted continue to administer the live.debian.net domain, which he has now rewarded by shutting it down with almost no notice.

> uploading packages

A package was not uploaded. ITP means "intent to package". It is the name of the formal procedure Debian uses to advise the entire project someone would like to introduce a new package into the distribution. It is actually a "wishlist" bug report that is automatically CC'ed to the entire project. Discussion on merits of introducing the package happen on that bug.

Sending an ITP isn't enforced mechanically - despite being official policy. It's one of those trust things. A DD could in principle introduce a new package without doing a ITP. Daniel did precisely that when he was a DD, to the chagrin of some.

> working behind his back

Daniel rage quit because someone discussed the idea introducing a new package into Debian in a forum Daniel monitored. The new package was a fork of a package Daniel maintains.

> with misleading names

The name is not misleading. xxx-ng generally means a new version of package xxx maintained by someone else and taken in a new direction. That accurately describes the ITP'ed package.

> But then again, I only know what I read on LWN, so my ignorance might be showing here...

Yes, well LWN isn't unique in promoting the "tribe expels member because they don't like him" angle. It makes for a good copy I guess.

The reality is rather different, as anybody who inhabits the mailing lists knows. Debian regularly has spectacular clashes on it's mailing lists - both personal and technical. I consider people so passionate about something they push it a little too far a sign of a healthy project. Despite all the complaints about how unpleasant Debian's mailing lists are, it is rare for Debian to take punitive measures. When it does, it's usually just banning an email address.

Despite how it's being reported this isn't an example personality conflicts running roughshod over Debian's internal governance systems. I see it as more of an example of how Debian's tendency to be fair even under the most trying circumstances can be costly at times. Regardless, personally I would not have it any other way.

Upheaval in the Debian Live project

juliank — Thu, 03 Dec 2015 00:21:26 +0000

This is certainly true. There was some overlap here though, between the team responsible for (live, and other) images and the team that maintained the tools for building live images.

If your team is responsible for building live images, it seems like a natural choice to name your next generation tool for that this way. After all, live building is their responsibility, and they replace the existing tool.

That was unfortunate, but I don't think anybody really thought much about naming here, and just chose the first thing to come to their minds.

After all, we are mostly a group of software developers, not social relations experts.

Upheaval in the Debian Live project

philomelus — Thu, 03 Dec 2015 00:06:23 +0000

If I am the maintainer of package foo, and without any notice see a package named foo-ng, my first reaction would be to question why someone is writing a next gen package of something I maintain without coordinating with me? Posting bug reports, etc. is not the same as telling someone that you are working on a new version of the package they are responsible for (the operative words there are 'responsible for'.)

Upheaval in the Debian Live project

fandingo — Thu, 03 Dec 2015 00:01:09 +0000

I'm glad they changed from "live-build-ng." Is there anything in software development lazier than throwing a -ng on the end of the previous implementation's name. How long is it "next generation?" Once it's reached release status for the public, doesn't it necessarily become "this generation" by definition?

Upheaval in the Debian Live project

juliank — Wed, 02 Dec 2015 23:37:04 +0000

That does not mean that the less trust is really subjective, it's a result of past incidents. Let me give you some background.

1. Daniel has always done too much work, and got overworked. He already had to reduce his work around 2007 - I was personally affected by this, but I'm sure there are sources somewhere to show that. After his initial reduction of his maintenance workload, he somewhat quickly managed to get back to an unmanageable state, though, if I recall correctly.

2. He also demonstrated lack to team skills by actively worked delayed a debian-installer release candidate (the syslinux incident) - just because his argument was technically correct does not mean that it's a good idea to just break the workflow of a core team, and he should have helped to do what is needed to help releasing things.

3. He also broke almost all systems using btrfs as / at around the same time and blamed others for the issue because it worked for him (https://bugs.debian.org/701936), while he could have easily reverted the change to prevent breaking user systems. Remember that the social contract says: "4. Our priorities are our users [...]", "We will place their interests first in our priorities." - he clearly did not do so in that case.

I think with this history, you will understand why the trust was lost and why he was "demoted" from a Debian Developer to a Debian Maintainer, that is from unrestricted upload access to any package in the archive to upload access to a set of packages people trust him to work on (any DD can give upload rights for any package to any DM).

I don't think there were any passive aggressive moves. The team responsible for building the official live images (and all other images) found live-build to unstable to use, so they wrote a new one to replace it. They named it live-build-ng, as it was the next-generation of live building tools used for official Debian releases. People do not always think about what effect their comment/name choice/etc. will have on other people, so sometimes unfortunate things happen.

The systemd debate was a much more brutal debate. Not only internally, but also with many people from the outside trying to push their own agenda (less people then apparent, because many persons were in fact the same person). It was decided in a democratic way which is the second best outcome there can be (the best being achieving consensus). In contrast, there really was no huge debate here at all. A new package was announced, Daniel did not like the name, the name was changed after some insisting, and then Daniel suddenly and unexpectedly announced the live project as dead.

Upheaval in the Debian Live project

MatejLach — Wed, 02 Dec 2015 22:57:10 +0000

> There was never any intention to force Daniel to stop working on the live packages.

> It's obvious that it's hard for him to work with people who do not trust him much, but that does not mean that he had to completely pull back his involvement

You've sort of confirmed my point, you cannot have "people who do not trust him much", uploading packages with misleading names and working behind his back as part of one and the same team, that simply doesn't work, (look at Microsoft's politics).

Maybe "There was never any intention to force Daniel to stop working on the live packages", but it certainly seems to me like there were some passive aggressive moves being made by some of the same people who seemed to be making such moves with the systemd debate - my point being this doesn't seem to be an isolated incident of such behaviour in the Debian community.

But then again, I only know what I read on LWN, so my ignorance might be showing here...

live.debian.net deleted?

laarmen — Wed, 02 Dec 2015 22:57:03 +0000

I just tried to follow the link to the downstream project list hosted on live.debian.net, and I get a DNS lookup error (other debian.net domains such as mozilla.debian.net work fine AFAICT). You could perhaps change the appropriate links to the Internet Archive snapshots ?

Upheaval in the Debian Live project

juliank — Wed, 02 Dec 2015 22:46:02 +0000

And that's the part people outside cannot understand: Daniel was not forced out "due to social issues and high egos of others".

There were repeatedly serious problems with live-build not working correctly when a Debian release was being made[0], thus the team that builds the Debian images wanted something more (consistently) reliable to replace it, wrote something new - based on another tool to create Debian systems, and announced it to the world.

If you had one tool that is repeatedly broken the time you need to use it, and another reliable tool that does almost everything you need, you probably would do the same thing and extend the other tool instead of trying to fix the current tool all the time.

There was never any intention to force Daniel to stop working on the live packages. In fact, the new tool uses those exact packages.

Daniel overreacted on this issue, and I do not really understand why. It's obvious that it's hard for him to work with people who do not trust him much, but that does not mean that he had to completely pull back his involvement in live system building.

[0] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=804315#65

Upheaval in the Debian Live project

MatejLach — Wed, 02 Dec 2015 22:31:51 +0000

Same here, sadly it seems to me, (as an Debian outsider), that some of the same elements that made the whole systemd Debian debate a lot more toxic than it needed to be, (again, from the outside, it certainly seems that way), which is sad when it forces good developers out.
We cannot keep loosing good developers due to social issues and high egos of others, we rally can't.

P.S. Just an opinion, no flame please.

Upheaval in the Debian Live project

skelbley — Wed, 02 Dec 2015 19:45:26 +0000

Likewise, I hadn't heard the story behind this yet.

Upheaval in the Debian Live project

Martin.Schwenke — Wed, 02 Dec 2015 19:42:38 +0000

While I'm not one of those who sent email, I'm grateful for the information in this article. I had no idea that any of this was going on. Thanks!

Upheaval in the Debian Live project

juliank — Wed, 02 Dec 2015 19:39:32 +0000

The thing is; I'm only active in Debian today because of Daniel. He uploaded my first packages, as he did with many others. He was always involved a lot in Debian, and that became too much. So at some point, he had to stop doing some of the work he had done before, because he could not do the work in a reasonable time anymore. That was in 2007 or 2008.

The package name choice was unfortunate, but that always happens in free software. Just look how the apt namespace is invaded by third parties without any coordination with the APT team. It's not nice to have tons of other software with similar names and users being confused by it, and then asking in the wrong channels for help, because they think your team is responsible for it. That aside from the issue of building upon the reputation of the project you base your name on.

The syslinux incident and the demotion to a maintainer shows you that there were problems in the past with how Daniel was working and interacting with other members, and there are more issues that led to the creation of "live-builder-ng" and the notion that "Debian Live" is not official, but it's not up to me to comment on those.

I'm glad for all the work Daniel has done, for the packages he sponsored for me; and I am sad about the problems, but the best we can do now is to look forward instead of backward, and focus on making future Debian releases better than ever (the APT 1.1 release will make package management in the next release much better, BTW).

Upheaval in the Debian Live project

corbet — Wed, 02 Dec 2015 19:12:42 +0000

We have had email from a number of readers who disagree with you. You may be current on what happened here, but a lot of other people are not.

Upheaval in the Debian Live project

juliank — Wed, 02 Dec 2015 19:08:02 +0000

There's *no* reason to bring this topic back from the dead now. The whole thing with Daniel was painful for everyone, we don't need to be constantly reminded about it.