Chiariglione Is Right

I guess everybody else has reacted on his post about MPEG crisis so I can do that as well. So, $postname—most people just don’t understand his outlook. If you interpret his words from his point of view it’s clear he’s right for most of the things.

His posts starts with the list of their achievements in terms of standards produced. Some people claim that most of them don’t have much to do with MPEG: MPEG-1 video is based on H.261, MPEG-2 video is joint effort (and it is H.262), MPEG audio layer III was developed at Fraunh*fer, MP4 looks suspiciously like A**le MOV format etc etc. For that I’d like to quote Zvirmarillion (one of the best parodies ever written but since it’s in Russian hardly anybody would read it):

…First of all two lampposts were made, Aulë crafted two lamps, Varda filled them with kerosene, and Manwë signed the act of acceptance.

You may not seem it as such but from bureaucratic point of view proper paperwork is much more important than actual work done. And the words “business model” occur in the text shortly after that (and I’m sure the quotes are in the original text because MPEG is not for-profit organisation) that explain it: MPEG participants pool their research, they pick the bits for the “best performance” standard (this is true but it’s always a moving goal) so the result can be accepted by the industry for its merit and give profit to the MPEG participants that “donated” the technology (profit in forms of licensing patents or exploiting their intimate expertise in that area, whatever). So it’s more about the business than technology really.

And if you think about it, then indeed the great danger to this model is third party patent holders. From MPEG participants’ point of view they are either companies who own crucial technology used in a standard and could not agree with each other on licensing terms (it’s no secret that their relations are complicated) or parasites who try to profit off MPEG work because they own a blanket patent. From outsider’s point of view there is no difference—they all are greedy bastards without a trace of common sense. Actually they are not different from many other companies, e.g. D*lby Labs Licensing Corporation (you’d be surprised to find out how many H.264 or AAC decoders you can license for a single AC-3 decoder), except that some ITU H.EVC patent cesspools want you to pay by usage (i.e. streams encoded/distributed). Obviously that’s where any sane entity would start to search (or make) alternatives.

The solution he proposes makes perfect sense for the goals he sees before him (serving interests of MPEG members who allow him to travel around the world at somebody else’s expense). The fact that consumers want cheaper devices and less hassle with playing videos from various sources is a completely different problem outside of their scope (the standards are made for the vendors after all).

Now the bit about AOM being a threat for the whole industry is worth discussing further. First of all, remember ITU T.81 (aka JPEG). It was accepted in 1992 and it was arranged that everybody can get royalty-free patent license for it—except for Q-coder owned by !BM—and guess why it got widespread (except for arithmetic coding mode). Of course companies lose any incentive to research image compression further, funding to the universities was slashed and nobody has worked on image compression ever again. But if you look at it from standardisation point of view then this is exactly what has happened: JPEG is still alive and kicking while later standards like JPEG-2000, JPEG-XR or JPEG-WTF failed to supplant it. Same for audio: there are countless variations on AAC theme (including xHE-AAC which is almost completely unlike HE-AAC) despite people using Vorbis and nowadays Opus too.

One would argue that it’s harder and harder to create new codecs. But the building blocks are common so most of modern video codecs follow the same scheme as ITU H.EVC (still waiting for PERSEUS2 BTW). And people still keep creating screen capture or lossless audio codecs simply because they can.

Now one can remind of industry funding of research. I’d say they still have to finance research even in other areas (look at the companies participating in MPEG—most of them known for many products that have not much to do with multimedia—so they’ll survive regardless). And research is such a tricky thing that it will happen in one area even you were looking for something else. But it may be committee funding being slashed and obviously when you’re a chairman of one you care.

The other fun thing to remind is that many companies poured money into creating their own multimedia format ecosystems and while it seems those days are over, they are not. RealNetworks is still buffering, Baidu is trying to make its own ecosystem out of former On2 codecs etc etc (now it’s called AOM). And it will probably remain the same: there will be few accepted formats per ecosystem (like Baidu VP9/WebMKV or H.264+AAC for A**le) and there are means to recode almost any input into anything else so you can accept any user input. And if it gives you a competitive advantage (aka vendor lock-in), why not do it?

Of course the part of industry that depends on interoperability the most will be hurt the most too. But if you care about broadcasting then you know they’re completely fine with living on legacy, so it’s no big change for them.

In the conclusion I want to say that there are many outlooks on the same thing and they may be quite different, the same facts making sense in one system of the world but not in another. And it’s often quite useful to try and have a look from another point. For example, MP3 design is weird with MDCT applied after QMF which is hard to explain from technical point of view; but an anecdote (i.e. I have no proofs of that story being true) says that certain lamp producing company had patents on QMF technology and forced its inclusion into the codec—see, now it makes sense! And that’s why reading Cyril Northcote Parkinson works may be as beneficial as reading some textbook in signal processing.

Leave a Reply