FFmpeg's future and resigning as leader
ffmpeg.org
ffmpeg.org
As a heavy user of FFmpeg/libav, for VLC (I'm the VideoLAN president), the drama has been huge, and more aggressive than most forks we've seen in the open source community.
Most of the information you find online is biased, wrong or down-right lying, including the usual blogposts that are quoted from one side or the other one. (like the pkh.me pasted or the Debian wiki one)
Whose fault is it? Well, it's a bit complex to answer, and there has been very dick moves from both sides. Especially since the beginning of the story started in private and was not reported publicly.
Who is doing more work between the forks? That's also very hard to answer, because most of the libavcodec and libavformat work has been done in libav, but most of the libavfilter, ffmpeg and tools work has been done in FFmpeg.
What you should know is that the current codebase of FFmpeg is messier compared to libav, but has quite a few more features, notably filters and more formats supported.
Choosing one or the other gives you different set of features, but mostly different set of bugs and level of support of codecs. It's a big mess...
> most of the libavcodec and libavformat work has been done in libav
Well, I believe this is a bit more complex. While it might be true for API changes and cleanups, FFmpeg has dozens of additionnal codecs and formats, as well as way more assembly/optimization code.
> What you should know is that the current codebase of FFmpeg is a huge mess compared to libav
Honestly, this is pure FUD and it's very surprising to read that from you. You seem to have no idea about the cleanups of FFmpeg because we do not advertise them (they are not relevant for the users). What is so much a huge mess? This is very insulting for the work done by FFmpeg developers these last years. Hey, even libmpcodecs (the MPlayer filter wrapper) is no more. A lot of work was also done on the documentation. Seriously, please stop doing FUD like this.
> Choosing one or the other gives you different set of features
Not exactly true.
I did, and many times. And I already said so.
> Well, I believe this is a bit more complex.
Unfortunately, it's more complex, yes, but it's not that much. Elenril, Koda and Martin are the one spending the more times and commits on libavcodec and libavformat.
> Honestly, this is pure FUD and it's very surprising to read that from you.
It's not. Unfortunately. Who has multiple implementations of the same codecs (2 prores decoders, at least 2 prores encoders) or libraries (2 resampling libraries) ? Who is spending hours removing the famous MPEG12EncContext from everywhere? And I don't even speak about the refusal to removing the non-working xvmc support for MPEG2.
And sorry, but you are one side, that joined after the fork. Attacking me is not really objective.
> Not exactly true.
Of course it is. Every time I do package VLC I need to choose one fork over the other and every time that means a different bug. This is documented over and over on our wiki and our bugtracker...
EDIT: I'm sorry you (and other) took that as an attack on FFmpeg. We use FFmpeg over libav in the current VLC windows and Mac version.
In the decades I'm developing software, I've learned that there is no developer out there liking somebody else's code. There's always something. You can think you're objective and unbiased, but you're not, no-one is. FFMpeg's code might be a mess due to a large chunk of legacy / backwards compatibility code, but very likely the other side's code base is a mess in other people's eyes.
The fork is hurting us a lot. It means a lot of work.
Given what I've read from the stories, and taking into account the rhetoric employed by one side versus the other, I have to say that, at least in tone, FFMpeg struck me as being more mature about the whole affair.
same here. And regardless of what their codebase smells like, their stuff works and does what it says on the tin, letting people play whatever format there is out there on a massive amount of different systems. I'm the first to admit the dialogs they're using to configure the codecs is one which is a yearly 'Ugliest UI' award winner, but luckily others have made front ends which make you avoid all that mess in the first place.
Because distributions have picked libav or FFmpeg. So we need compatibility with both.
Given how similar libav and ffmpeg still are in places, they might even consider replacing the vendored ffmpeg with the packaged libav (or vice versa, if done the other way).
But the fallout would be for VLC to deal with ("VLC breaks on my Debian, it must be crap").
I assume this is an unbiased opinion as much as it can be but could you elaborate on what about FFmpeg is a "mess."
I've used both these libraries in building MythTV from strictly a user perspective. Both code bases are open source. I'm wondering if there is so much turmoil here, why doesn't someone (like VLC) take the best features from both projects and create their own library?
I never understood the OSS communities. It just seems simple to me. If you don't like it, fork it and do your own.
<shrug>
Examples: 2 prores decoders, 2 prores encoders (or 3), 2 different resampling libraries, libstagefright integration that does not work, more global context structures compared to libav, etc... They also refused to remove some very outdated (non-working) code like xvmc.
> why doesn't someone (like VLC) take the best features from both projects and create their own library?
Because it's a lot of work. An enormous work.
Scott Meyers, talking of the various… quirks of C++. https://www.youtube.com/watch?v=48kP_Ssg2eY
Three different copies of the same code? Yeah, ok. That's ok. One day, when you can, merge them together. You can't merge them together? It's ok. Maybe they shouldn't be merged together.
Is it a barrier to entry for new developers? I feel that any developer who has worked on messy legacy code won't bat an eye. This sounds more like a problem experienced by a young coder fresh out of university.
You definitely can write clean code. Cleanliness can be a priority. There are some surprisingly clean and well-written large codebases out there. However, a lot of people just don't care. They are very smart, capable people who just bang out code that works and don't care to integrate in the proper clean way with the old code.
If you want clean code, an old, large, working C/C++ codebase is not what you want. We often can't spend time cleaning up old code, since there's always more features to implement, bugs to fix, etc. Learn to deal pragmatically with code like that. It's not like that out of malice, or some misguided attempt at writing the largest IOCCC winner. You can work with it. It's just a bit trickier.
That might not be quite as bad in a proprietary codebase since you have somewhat more of a captive audience, but in an open source project where you want a long-tail of contributions from many people, that is a serious problem.
The issue is the "you" implied there. So you fork it and start working on your own. What do the other contributors working on the previously-existing project do?
This seems to resolve itself when there's a clearly superior project and developers and distributions switch to the newer project.
However, if both projects continue to simultaneously develop...
Because that in itself would be a huge project, with moving targets as the source projects develop over time. If it were that easy the VLC people would just implement everything themselves and not need either library.
> I never understood the OSS communities. It just seems simple to me. If you don't like it, fork it and do your own.
The problem is, of course, the humans.
People sometimes take things very personally even when not meant that way. A significant fork for no good reason can distract from the work of the core project and dilute the user-base, and people will not always agree on what a good reason is/n't. Sometimes the problem is stated as poor code and/or procedure quality, and this can be taken as a rather direct insult.
Sometimes massive forks happen because of differing priorities between different groups of people and everything is perfectly amicable - you don't hear about these cases much (unless it inconveniences the userbase) because there is no drama to make a noise about. Sometimes these forks later re-merge, sometimes one becomes clearly superior and the other fades out, sometimes they go on to be completely separate projects.
It sounds like three are personal issues involved in this particular mess that we don't exactly know about so can#t hope to fully understand because the discussions started in private and have not been published and as we all know any argument, especially one that has gone on for some time, people can lose track of the true chronology of the issue and that can make reconciliation near impossible.
I'm a software developer too. In my career I've had to implement logic to work around buggy libraries, spend time looking into bugs in others code... It seems that at some point that the means don't justify the end and having my own code base that I'm familiar with would be easier than fixing theirs and waiting for disputed parties to figure out what they are going to do.
And I never claimed doing your own was easy. All I'm suggesting is that at some point it's easier than dealing with what you got.
I hope some big entity ends up throwing money at the situation, it seems fixable, just a lack of manpower to fix it. I mean, Google relies on it a whole lot for one.
For the merging to happen, there would have to be both some agreement on the technical direction of the two projects, as well as some sort of social resolution, which would likely involve some maintainers of one fork or the other leaving.
Regardless, from my point of view, it seems unsustainable to keep on merging from libav to ffmpeg at the pace that Michael has been doing, so the projects will end up diverging enough at one point that momentum ends up behind one or the other.
It's partially untrue, yes. There was an actual vote from all the active developers to remove Michael from leadership, and that included the infrastructure developers.
But yes, the fork was very badly done.
Then why did so many remain with the ffmpeg project instead of going with libav ?
"I think the statistics indicate that the situation is much worse than I expected. Please keep in mind that because of the daily merges, all commits that are made in Libav also appear in FFmpeg, but not the other way round. With a finger-in-the-wind estimate of subtracting the numbers in the Libav table from the FFmpeg table, the distance between Michael and the other FFmpeg developers because even more extreme."
"Reinhard asked whether FFmpeg is a one-developer project that would find itself in trouble should Michael stop working on it. "To me, this constitutes a serious bus-factor: Without Michael, (probably) nobody is able to replace him." He went on to suggest, though, that Michael's departure could do a lot to bring an end to the fork."
The reason? Yahoo! could not have existed without Perl.
FFmpeg has been used in so many profitable ventures (hint[1] hint[2] Netflix). I sincerely hope there is a business leader with the same level of consciousness and grace that will do the same for Michael. He's one of the heroes of the past decade. Internet video streaming would certainly not exist as soon as it did without him.
[1] https://twitter.com/nicolasweil/status/466248052454727680
[2] http://www.streamingmedia.com/Articles/Editorial/Featured-Ar...
This is why the community is broken and sadly some of us feel cornered to do Affero or just not open source. I used to be hardcore BSD, but times changed and had to adapt.
These the majority of the community is selfish, doesn't contribute back, doesn't even give credit to most of the tools behind their UIs. See:
Not sure if there was much contribution or acknowledgement for the longest time from any company (Google, FB, Vimeo, Netflix, Zencoder, etc) using it, and I remember the best they'd do for bug reports was say "yeah it crashes sometimes" and refuse to send sample files.
Eventually Google did have some people contributing full-time, of course. They paid me 1x GSoC (one week's engineer salary) for more than a year of work once. Um, it seemed like a good idea at the time.
Thinking about more than yourself never seems to work out in this world. I really, really wish it would. And I'll personally continue to follow my conscience over my wallet. But the wallet always wins in the public arena. In all facets of life.
I think he should have left out the bit about 'will I ever return', that might actually cause libav contributors to hold off from making the switch.
Still, it's great to see the recognition that in the end this is all about the product and its users.
There are more links in that post to fill you in on the details.
This post is from someone involved in libav but it refers to the different public e-mails that were exchanged when the fork was originated and they document factually things (at least, as far as I can understand)
I'm not saying the points aren't valid, but I stand by my statement that "libav is working to the same goal as ffmpeg. The split is due to idealism and politics, not due to the goal of the project."
>former FFmpeg leader Michael Niedermayer.
Former? Well, he resigned today, but this blog seems quite old.
>Since FFmpeg was a project famous for horrid code quality and sketchy and irregular API
Was it? Is it? All I know is a bunch of stuff uses it. Where is that fame?
>“The people behind Libav stole the FFmpeg infrastructure”, some of the admins even got questioned in their workplace if they really stole something. Pity that most of the people related in keeping the infrastructure functional were doing so on their own hardware and co-location, but obviously checking facts is hard, better go help shoveling manure on people.
Server infrastructure is one thing. You mention yourself that you guys didn't have the trademark, yet you took the domain name too? I dunno about the laws in all of the locations where this played out, but in the US at least, that'd be theft.
>started to merge daily everything Libav does more or less since the beginning, making FFmpeg effectively a strict derivative of Libav.
Again, tone, wording, etc. This statement does not stand on its own: If I have a 50 million line project that had a small subsection forked, and every day the fork's changes were merged, that does not make my project a strict derivative. I'm not saying libav is a small subsection, or making any sort of claim about the ratio of ffmpeg original work vs. libav merges, but you have provided zero information or evidence to support the claim one way or the other either.
>Now things are a little more fair and at least FFmpeg more or less states that is a derivative of Libav with additional features in their download page.
Ah, here you go. Kinda. But again, tone. And not much evidence. The ffmpeg download page seems to be open about including code from libav, but uh, additional features could be major. What's the difference in features? What is the actual volume of work in comparison?
>On the other hand I’m not cool at all in having an unfair competition, with a side piggy-backing on the other like it is happening: everything in Libav is merged inf FFmpeg, enjoying the fact the code is polished and cleaned before.
This is a feature, not a bug. Unfair competition? Piggy backing? You are developing an open source project. Do you only want things to be open if the people using it are people you like?
>sometimes the amount of nonsense thrown at you by those rabid fans not knowing anything is appalling.
You write this, and immediately follow with:
>Hopefully writing more about it might help defusing this situation.
Blog posts like this are very inflammatory. You hurl insults, make a lot of claims without providing evidence that backs them, and generally come across as someone who has a lot of personal bias on the matter.
That isn't to say that you're wrong: I have no idea who is right or wrong on this. I don't particularly care. But your blog post is about as biased as it gets. And to note, biased doesn't necessarily mean wrong - you can be biased and still be right.
He got demoted, as written in the blog post, did you read it with attention or you just went through looking for bits you could use? > Was it? Is it? All I know is a bunch of stuff uses it. Where is that fame?
You can ask the GStreamer people or the VLC people, I started being involved in MPlayer and FFmpeg because the code was not really working on PPC...
> Server infrastructure is one thing. You mention yourself that you guys didn't have the trademark, yet you took the domain name too? I dunno about the laws in all of the locations where this played out, but in the US at least, that'd be theft.
We did not took the domain, the owner of the trademark is not Michael Niedermayer.
You claim to know nothing yet you cut quite interesting judgement and you cherry pick in a quite interesting way.
"insults" is a quite loaded term btw.
It certainly seems like he had been leading the ffmpeg project until his resignation.
As for looking for bits I could use, I am quite specifically responding to your question about how the article comes across as biased. I was not looking for parts that seemed unbiased, because you were looking for the biased parts.
>We did not took the domain, the owner of the trademark is not Michael Niedermayer
Your own blog post says " the trademark FFmpeg had been given by the owner of it Fabrice Bellard to the former FFmpeg leader Michael Niedermayer"
Which is it? You say in your post he was given the trademark.
>You claim to know nothing yet you cut quite interesting judgement and you cherry pick in a quite interesting way.
You're defensive. You asked a question about how your article was biased. As an outsider, these are the parts that looked biased.
If you ask a question, you can reasonably expect an answer. If you don't want an answer, don't ask the question.
>"insults" is a quite loaded term btw.
Uh. Thanks for informing me, I guess? I'm not sure how you can in other places say "clean up monkey" is horribly offensive, but "rabid fan" is not an insult at all.
I doubt HN is a bastion of ffmpeg supremacists, and there's no organized movement to suppress your voice, so the reaction your getting should probably be taken as an indication that you should undergo some self introspection. Open source projects depend on community support to exist. Libav may be the best thing since sliced bread, but if you drive away potential users and supporters, you are harming the chances of the project's success.
In context he may consider the vote to keep Michael Niedermayer as team lead a farce. Several of the voters mentioned that they only voted for him to enjoy the resulting drama.
> This is a feature, not a bug. Unfair competition? Piggy backing? You are developing an open source project.
There is hate between the projects. libav was (iirc) started with the intent to be ffmpeg done right, something that is hard to show when the ffmpeg maintainers can simply copy paste every improvement until libav collapses from lack of support/donations.
Not going to take a side in this either, there seemed to be a need to grow up in both groups.
'..making FFmpeg effectively a strict derivative of Libav' is a pretty egregious display of bias, one that to some degree shows the distortion of the author's viewpoint.
I don't blame Michael for any of this situation, but this is not a win for the community, product, or users. Development is going to slow way down, bugs are going to go unfixed, and the overall quality will decline in the short to medium term. It is unfortunate that vitriol over the fork led to this situation, where neither libav nor FFmpeg win, and everyone loses.
and somehow leadership was still a thing for Michael at that time, which really contradicts with something he says now.
Sadly the people on the Libav side never spoke up and nobody really checked or asked since they are "EVIL".
This is completely incorrect. Most of his commits are merges from libav and other forks, and not code commits.
He's doing a great job at it, but he is not irreplaceable.
git shortlog -ns --no-merges v11..HEAD | head -n15
git shortlog -ns --no-merges n2.4..HEAD | head -n15
Michael Niedermayer is an EXTREMELY prolific dev who has been banging out FFMPEG code for more than a decade. It's by no means a one-man show, many other devs make large and valuable contributions - but in relative terms Niedermayer absolutely dominates FFMPEG, both in terms of personality and code output.
libsndfile -- another key component everyone takes for granted -- is in a similar position. Full of known problems, super-smart but slightly neurotic leader, and GPL issues that people are pretending not to notice in hopes he finally does a release after 5 years.
I've done major community work and man, is it ever exhausting. It's a shame we haven't found a way to get people the assistance they need. Some of which is psychological.
Should they? God no. No one can run that workload as a volunteer for a decade+.
But I can see how the default reality probably bends towards "Well, someone left and we need this job filled" -> "I'll just temporarily add that job to my workload" -> "It'd be more effort to move that job's responsibility to someone else, so I'll just keep doing it."
Hmm.
> I'm guilty as charged as well
You two are not being charged for the same thing.
> A leader has people skills above all else.
Leadership skills and interpersonal skills and management skills are completely separate. Amazing leaders/motivators who are downright shitty people are not uncommon in our industry.
This particular leader made the classic mistake of doing too much. "Nobody else wants to do it? Nobody wants to take responsibility for this? I will. I'll do it" is the definition of a bad manager. And admitting it in your farewell post doesn't demonstrate great people skills either.
Regardless, these people are all around us. And they deserve a beer and a mentor and some help. Next time you're looking for a project to contribute to, consider helping that guy.
No, they aren't. Leadership skills are a subset of interpersonal skills and management skills overlap interpersonal (especially the leadership subset) skills.
They are all different sets, but they aren't completely separate.
But, that quote above sounds to me like exactly what a leader is?
That's what leadership is.
There's probably some great popcorn in mailing lists.
Edit: libav have a big "personal conflict resolution" section here which is another hint:
All very sad but makes for interesting OSS history! I find FFMpeg very useful, but then again I find VLC and libav useful. In fact, it is all useful. I hope it all works out.
Somebody at Redmond is laughing at the OSS community for failing to provide functionality that comes out of the box on the latest Windows systems.
$ ffmpeg -codecs 2> /dev/null | wc -l
396
They can get new features out the door, though. One reason for the original fork in ffmpeg was that the very strict code review culture made it difficult to get anything in that was experimental or depended on personal taste.
Either way, yes. Elenril and others were doing good work cleaning up the API when I last paid attention years ago, but there's a lot of surprise technical issues esp. in libavformat that it has to cover up. (Like, PTS/DTS in avi. Like operations that are instant in some file types might be real-time minutes in others.)
As for the command line, major customers tended to not ever give feedback on it, and it was maintained by Michael who tended to use the same code-review standards on the UI code as on the inside of a video encoder. This made it quite difficult to ever get anywhere. That's one of the reasons there was a fork back then.
Of course, there's deep technical problems there too - you can tell it to do all kinds of things that are actually really hard, and it will kind of half-heartedly do them then fail. Try converting a Matroska file to MP4 without re-encoding it; you'll get a weird error. That one weird error would take more than a year of developer time to fix.
I have somewhere a much more prescriptive python library that sits on top of the command line API, but I think it'd be really great if there were a "strict" mode for the CLI -- no more implicits, no more silent rejiggering of options. If you don't totally account for your input and output, it's an error. Then, there could be a much more forgiving and intelligent front end to the front end that allowed for things like "-i foo.avi ./foo.mp4", making best guesses at each stage of the pipeline.
It may not be "pleasurable" but it's nothing to be trifled with.
Perhaps video is just a hard problem?
The problem as I see it is that media manipulation is not really suitable for a one-size-fits-all command-line solution, and the API was slave to the command-line tools. This may have changed recently, which would be great.
Windows 10 actually does support Matroska out of the box, albeit 11 years later than FFmpeg.
I know if I were ever in the position to contribute to general OSS video, libavcodec would be on the list of projects i would never be willing to support.
Hopefully these projects can reconcile and libavcodec and its shamed name can be put down and development can continue under the projects true name to honour its true roots.
I wouldn't put it that way, considering that the last distros still shipping libav are abandoning it in favor of ffmpeg.
and if he never liked the leadership role as he claims, he was given ample opportunities to get rid of it.
"I had hoped for a long time that the fork situation would resolve and both sides somehow merging back into one team. All the Libav developers joining FFmpeg again. But even now as the last distributions are preparing to remove Libav, still theres no hint of that happening. Maybe even the opposite."
and
"Do friendly merges, and if you like do hostile merges. Its all up to you now!"
That isn't "a ton of derogatory remarks".
libav is working to the same goal as ffmpeg. The split is due to idealism and politics, not due to the goal of the project. I'm not going to pick a side in this, as interpersonal relationships are impossible to accurately gauge from an external viewpoint.
I'm not sure if you have a personal stake in this argument, and I apologize, and sympathize if you do.
Best of luck in the future Mr Niedermayer. Thanks for all the work over the years.
that the gist of it, it's always the other side wanting him harm. he never ever acknowleged any wrongdoing on his side.
You can't claim that everyone that disagrees with you is being "derogatory".
EDIT: Example:
"I disagree with you" = not derogatory
"You're an idiot" = derogatory
Libav has a code of conduct and it is enforced so maybe you'd have much an harder time finding such gems.
http://article.gmane.org/gmane.comp.video.ffmpeg.devel/17927...
(Keep in mind at least one of the Libav supported got questioned on his dayjob regarding "thievery" because of those lovely examples and the legends the fans kept propagating)
Is that the line you're objecting to?
That sounds kind of jokey and affectionate to me.
>If noone has time ill fix the memleak and commit and let the libav cleanup monkeys clean the rest up for me.
Is the kind of inflammatory comment you are so up in arms over, I think you must have lived a very sheltered life. Especially considering this comment looks to have come shortly after the hostile split.
Libav is about rules and not leaders.
If your rules prevent you from keeping your software secure, you don't have enough manpower to do what is in your user's best interests.
posts like this one is why it went off the rails a long time ago. there is also the one where he suggested a libav dev should commit suicide and later labeled it as "Austrian humour"
Your team won, Michael stepped down. Of his own volition. What positive outcome could possibly be had by still spouting bitterness and negativity against him? Is it because he did not genuinely say "sorry" at some point? After reading [0] and [1], is it not possible to admit that maybe both sides have some genuine "sorries" to pass around? Now seems like the perfect opportunity to make a positive change for developers and users instead of venting frustrations that are only going to further polarize the current division.
(I'm not "automatically" labelling your posts as bitter. Almost all of yours in these comments objectively contain negative language, typically directed at Michael.)
[0] http://ffmpeg.org/pipermail/ffmpeg-devel/2011-January/106403...
[1] http://ffmpeg.org/pipermail/ffmpeg-devel/2011-January/106489...
Edit, to add:
I guess I echo this sentiment (from article's email chain): http://ffmpeg.org/pipermail/ffmpeg-devel/2015-July/176493.ht...
For the another batch of links and some more background.