RMS on the Ogg Vorbis license (2001) | Hacker News Reader

RMS on the Ogg Vorbis license (2001)

lwn.net

100 pointsd99kris82 comments

It is interesting that 15 years later, with Ogg Vorbis not only providing better audio quality at a given bitrate, and Opus apparently even better, MP3 is still abundant.

I was especially annoyed when I still had an iPod, because substantial parts of my music library are in Ogg Vorbis and FLAC, both of which Apple could support without paying a single cent in licensing, yet iPod - at least a couple of years back - only supported MP3 and AAC. :-|

I would assume the patents on MP3 have expired by now, is anyone in a position to confirm or deny this?

GIF patents are expired I believe, but I dont' think MP3 is safe until 2018.

This patent, for example, doesn't expire until near the end of 2017: https://www.google.com/patents/US5703999?dq=5,703,999

Edit: This is supported by the wikipedia article: https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and...

> In the United States, the technology will be substantially patent-free on 31 December 2017 (see below). The majority of MP3 patents expired in the US between 2007 and 2015

The expiration date is wrong. Even before 1995, it was 17 years after publication or 20 years after priority date.

Fine, another example: https://www.google.com/patents/US5924060

This one was filed in 97 so +20 is 2017 (because the 20 years is from filing date, yes?)

I think it is priority date, which is often the filing date of the earliest patent in a chain of continuations. For fun trivia, look at https://en.wikipedia.org/wiki/Submarine_patent

It's from the priority date, which was Aug 29, 1986 making this expired.

https://en.wikipedia.org/wiki/Term_of_patent_in_the_United_S...

From 35 USC 154(a):

    (2)Term.—

    Subject to the payment of fees under this title,
    such grant shall be for a term beginning on the date
    on which the patent issues and ending 20 years from
    the date on which the application for the patent was
    filed in the United States or, if the application
    contains a specific reference to an earlier filed
    application or applications under section 120, 121,
    365(c), or 386(c), from the date on which the
    earliest such application was filed.

    (3)Priority.—

    Priority under section 119, 365(a), 365(b), 386(a),
    or 386(b) shall not be taken into account in
    determining the term of a patent.

For this patent, there is a chain of specific references to earlier filed applications, but I'm not sure which of those, if any, are under section 120, 121, 365(c) or 386(c) and so affect the expiration date of the patent.

Apple's needless introduction of ALAC, basically a slightly dumber FLAC, was inexcusable.

There were some vague reasons behind it. Apple claimed it improves battery life for mobile devices (possibly because it is 'dumber') and it's based on the mp4 container format used by other Apple-preferred formats.

https://en.wikipedia.org/wiki/Apple_Lossless#Codec

Using the same file extension for non-lossy and lossy music... ugh. If they wanted to sell DRM protected lossless files I'd even cut them some slack, but they didn't. This was just poorly thought out. Likewise both formats are trivial to decode; I'm not buying the power argument. The smaller .flac file likely wins anyway.

Unless you have a hardware decoder

According to rather old Rockbox benchmarks, their implementation of ALAC requires 3-4x the clockspeed as FLAC: http://www.rockbox.org/wiki/CodecPerformanceComparison

On a positive note, Apple later released a reference ALAC implementation under the Apache 2.0 license, which includes a patent grant, so the damage is mostly limited to consumer annoyance.

They could have stuck with the .mp4 extension, but they were attempting to reduce the potential for confusion. It's a file extension, once we embrace the fact that it is just a hint (and can be a lie), it doesn't really matter so much.

The MPEG-4 format was based on the QuickTime format, so it shouldn't be any surprise that Apple prefers to go with an MPEG-4 format even when an acceptable alternative is available. The reality is that Apple doesn't support very many alternatives to MPEG-4 formats, and they probably wouldn't support MP3 if it weren't for the fact that they wouldn't have been able to take a significant position in the market for digital media players without MP3 playback. The same will probably never be the case for Ogg or FLAC.

I'm aware of at least two (actual) MPEG-4 lossless formats (MPEG-4 ALS and MPEG-4 SLS). ALAC is just a random format worse than the existing one at the time, stuffed in a .m4a container. (Apparently .caf as well.)

MPEG-4 ALS and SLS were standardized after ALAC was released in OS X/QuickTime/iTunes, but yes, those are two more possibilities when you have an .mp4 or .m4a file, and the pile of other extensions they can use.

All media container formats can contain differently encoded data: avi, mov, mp4, ogg, mov, mkv.

>Apple claimed it improves battery life for mobile devices

Is there a source for that claim? I ask because FLAC is generally considered to be faster for decoding.

My guess is the MP4 container is the only reason they went for it (that and possible business/ecosystem reasons).

Vorbis and FLAC may be distributed under very permissive open-source licenses, but they may also infringe on software patents, which is ostensibly why Apple has avoided them. They also may not, but under the current US patent system we can't really know until a court case; it does make sense for Apple (or any other company trying to avoid paying huge sums to patent trolls) not to test these waters. Licensing AAC and MP3 from Fraunhofer was an easy way to pass the buck.

> Licensing AAC and MP3 from Fraunhofer was an easy way to pass the buck.

Not really. Someone else could pop up and claim to have a patent over AAC or MP3, just as easily as they could for any other codec. Licensing a patent from one party makes absolutely no guarantee that someone else couldn't pop up with another patent and either payment or destroy a product. And I seriously doubt that a patent license for AAC or MP3 came with indemnity against any third-party patent claim.

Not really. Someone else could pop up and claim to have a patent over AAC or MP3, just as easily as they could for any other codec.

And in fact, that's exactly what happened: https://en.wikipedia.org/wiki/Alcatel-Lucent_v._Microsoft_Co...

A $1.5 billion jury verdict was handed down against Microsoft, which was overturned by the judge.

> but they may also infringe on software patents, which is ostensibly why Apple has avoided them.

No, it's not the reason. Anything can infringe on unknown patents, even same AAC. Opus and Vorbis aren't any more risky, Opus is actually much less. Apple are just jerks who don't want that barriers for entry (requirement to use patented codecs) would disappear.

It's an argument against being the first major company to ship something; Vorbis has had nobody sue any of the major companies who have been shipping it for over a decade. The longer other major companies have been shipping without being sued the more likely it is nobody else will surface.

The claim that apple avoided FLAC due to parents doesn't hold up. Apples ALAC implements all the same techniques and so is open to all the same patent threats.

Some (Apple) staunchly refuse to support free codecs. They on purpose put blockers for wider adoption, so the above is not surprising.

I make it a point to buy music in FLAC, and then encode it in Opus for playback (that's why I try to avoid buying when FLAC / lossless is not available). But so far no store sells music in Opus optionally.

Probably because there's no widely available encoder. Why hasn't this been folded into ffmpeg/sox/etc?

Not sure what you mean. opusenc works just fine and is perfectly portable.

ffmpeg also supports it through libopus for quite some time.

See https://www.ffmpeg.org/ffmpeg-codecs.html#toc-libopus-1

It's decode only for some reason. For whatever reason sox and libsndfile won't touch it; ffmpeg won't encode it. And iOS and Android browsers won't play it. Shame because it's clearly the current winner.

ffmpeg should encode it if it was built with --enable-libopus. May be you are using one without it?

Firefox supports Opus everywhere. Chrome I think should too, since it's needed for WebRTC.

Safari obviously doesn't, as I said, Apple can't stand free codecs.

http://caniuse.com/#feat=opus

Strangely absent across all mobile platforms at the moment. And even Firefox won't play it as .webm only .ogg. (Huh?) Man, why is audio always so wonky?

I'm not so sure that the link has correct info.

On Android, support for OPUS is mandatory for 5.0 and newer (by CTS). I would be surprised if the browsers would implement codecs by themselves, instead of passing the streams to the system for decoding.

Firefox can play it as .opus

Recent changes to ffmpeg broke opus in webm in some cases, if you're using git master: https://trac.ffmpeg.org/ticket/5509

You forgot to build ffmpeg with --enable-libopus. ffmpeg has its own reimplemented decoder, but relies on libopus for encoding.

I said "widely available." When distros bundle an ffmpeg encoder, it's widely available. When libsndfile and sox support it, the format migrates toward common use. "build ffmpeg" is a nonstarter for most mortals.

What distros bundle it without libopus? Debian has it enabled for example.

ALAC is a free codec since 2011.

Since they made it themselves, they overlooked their usual negativity. But of course, they can't support FLAC (which was released in 2001) because NIH and because Apple.

Of course what? I don't get what you're trying to imply. Are you suggesting that inventing redundant incompatible codecs for the sake of it is de rigueur for Apple? Because I can't think of another example. As far as I can tell, the opposite is overwhelmingly true.

Apple has invented a handful of codecs and container formats in its time (I can think of MOV, ProRes, Pixlet and a few ancient/esoteric Quicktime codecs) but they all served a distinct purpose not sufficiently catered for by their contemporaries.

I'm willing to believe Apple when they say there was an engineering reason why they created ALAC. Perhaps an optimised implementation (not the reference implementation) may have offered some performance/battery life advantage on certain hardware. We don't know. And it doesn't matter much.

I don't have a handy source, but Apple claimed that they made ALAC because FLAC couldn't be decoded on an iPod in a power efficient way.

> And it doesn't matter much.

Yes, it doesn't because whatever reason they had then, they can't use it not to support FLAC today. Today it's clearly their arrogant NIH attitude.

> clearly their arrogant NIH attitude

Sure. (Except the opposite is almost always true.)

I thought Apple are notorious for NIH, and looking down at interoperability in general. I surely didn't notice that the opposite is almost always true.

Can you cite any examples of where Apple has eschewed standards or interoperability without a good technical reason? I suspect anything you list will be either:

• No interoperable standards were available at the time;

• Distinctly better than / more capable than contemporary interoperable standards;

• Maintaining a legacy decision made for one of the above reasons.

Practically everywhere. From USB in hardware to touch events, WebRTC and MSE in the browser.

Anyway, no interoperable standards were available at the time doesn't excuse them not adding these standards when they become available.

Maintaining a legacy decision made for one of the above reasons is exactly NIH and attempt to sabotage interoperability.

USB? You're not even being serious.

There really isn't much on iTunes in lossless, though. At least, there never seems to be when I want to buy something.

Yeah, I'm not using iTunes though. There is really no reason whatsoever not to provide lossless in digital form, especially since CDs did that for a long time in physical medium. Bandcamp always provides FLAC. That's where I buy music mostly.

Spotify uses Ogg Vorbis for all its streams, I believe. But I guess that doesn't really count when it's heavily DRM encumbered and delivered by unfree software.

It also found a niche market in video games, where it's very common to this day (there's an outdated list on the xiph wiki [1]). Netflix also used it on the Wii when they didn't have a system decoder available [2].

[1] https://wiki.xiph.org/index.php/Games_that_use_Vorbis

[2] http://blog.streamingmedia.com/wp-content/uploads/2013/07/20...

I wonder if they still use Vorbis now that so many use Spotify on their smartphones. Lack of hardware decoding might be an issue.

I think it "counts" in the sense that Spotify had the freedom to do that, which is exactly what Stallman was fighting for.

Android reads Ogg and Flac natively. It is being used by Spotify, and in most games internally, especially indie games. Ogg didn't win, but not far from it either.

At the time, Ogg Vorbis was difficult to hardware decode and its impact on the battery life was noticeable. Other decoders have been introduced since but I don't know if hardware decoding is here or not.

Tremor, the fixed-point decoder of Ogg Vorbis, came out in 2002. According to the Rockbox performance data, it requires about the same clock speed as AAC-LC. iPods and similar generally don't have fixed-function audio decoders. http://www.rockbox.org/wiki/CodecPerformanceComparison

   > Ogg Vorbis not only providing better audio quality at a given bitrate

That is not my experience at all. If I need to encode in a lossy format, LAME has always produced a smaller transparent encode than Vorbis. But if I've been bandwidth/space restricted, Vorbis and AAC have been better than MP3.

Calling MP3 "transparent" even at 320kbps is such a far reach. Especially in treble and bass, lack in detail, etc. It's just a limitation to the codec. In comparison, AAC produces a much higher quality encodes, which somewhat approach a "transparent" lossy encode.

I have yet to see anyone being able to tell the me what sample was the mp3 with any consistency in an ABX test. I don't know how other encoders perform, but LAME is producing really really good results.

There is no excuse for MP3 in 2016. Like parents who eagerly took "baby's first step" photos with the first iPhone camera... oops. As technologists and influencers we should help migrate people forward. Ya hear me, SoundCloud?

These days Ogg Opus is a better audio codec than any other lossless one.

you mean lossy?

yes

You simply don't need to be the best to win, you just need to be good enough. If your technic is good enough other soft factors are way more important.

In the case here I think it's mostly that mp3 was first to market: https://en.wikipedia.org/wiki/MP3#Going_public

Absolutely this....

What a lot of us techies fail to realise is that the world at large does not care about technical elegance and purity. They care about the end result.

Packaging and aesthetics do matter to the world at large... therefore enough features, nicely packaged and sold well.... will win.

> It is interesting that 15 years later, with Ogg Vorbis not only providing better audio quality at a given bitrate, and Opus apparently even better, MP3 is still abundant.

MP3 requires far less memory to decode than Ogg Vorbis does. This was a big deal at the time as RAM cells added cost to the MP3 decoders.

Now, with CELT (which later became the Opus codec) this was not true. But it took a long time to get there.

Comparing Vorbis to MP3 nowadays doesn't make a lot of sense. Compare it to AAC.

Opus, the a descendant of Vorbis, is standardized by the IETF and (correct me if I'm wrong) mandatory-to-implement for WebRTC.

And since WebRTC is supposed to take over our lives, perhaps Vorbvis did win, after all.

Opus is not descended from Vorbis. It's descended from CELT and SILK. CELT was developed by Xiph.org, SILK by Skype. The CELT portion handles complex, arbitrary audio like music and sound effects. The SILK part handles voice.

Xiph, Skype, and IETF (among others IIRC) pooled their expertise in CELT, SILK, and the standards track respectively. Xiph.org did develop Vorbis though.

I've heard numerous complaints from people with deep technical knowledge in this domain that Ogg is a terrible format - for example, see: http://hardwarebug.org/2010/03/03/ogg-objections/

MP3 is used where simplicity and compatibility are prioritised - Ogg is neither simple nor widely compatible. Where quality is prioritised, there are other alternatives providing at least as good quality.

So asking "Why is Ogg not used over MP3 because quality" misses the point - MP3 wasn't chosen for premium quality, and Ogg is awful at the things MP3 is good at. Apple supports MP3 for compatibility/simplicity, AAC for quality (lossy), and ALAC for quality (lossless). I don't really see how Ogg competes on a technical level with any of those, for their respective purpose.

Your link is to an argument against the ides of using Ogg as a general purpose container format (for video in particular). This is about something different; Vorbis encoded audio which traditionally comes in an Ogg container. I have never heard the argument that the Ogg container format somehow makes Vorbis inferior to other codecs in practice.

I think that's misrepresenting the article somewhat. It mentions video in several places, because a lot of the issues stem at least in part from the complexities introduced by trying to be a general purpose container, but there are many significant problems with use of it as an audio format besides these.

Ogg containers have this problem that to change the metadata you have to rewrite the whole file whereas with ID3 it's always the last bytes at the end and highly efficient to edit.

I've dealt with the ogg container format a little and from what I saw it seemed quite reasonable, at least for the codecs it is typically used for, but for the sake of argument I'll assume everything in that article is correct.

In that case, ogg competes because it beats AAC on quality per unit size (low bitrate vorbis supposedly beats out low bitrate AAC at least, and I guess opus is supposed to be really good), and FLAC, which usually comes in its own container format but sometimes is distributed in an ogg, has smaller file sizes than ALAC.

I don't think AAC had much traction before it became the iTunes Store format, and ALAC remains one of the least common codecs. Both of them would have lost out to vorbis and flac. Perhaps the reason they were chosen is because they have standardized DRM solutions (at least AAC).

I think people interested in AAC probably aren't using it at low bitrates, so I'm not sure how meaningful that is. At more reasonable bitrates for people who care about quality (and thus are using AAC over MP3 in the first place) I don't think there's enough difference to discern on most listening devices - but I've not done any sort of scientific testing between the two myself, so I could very well be wrong on this.

FLAC vs ALAC is a much stronger argument IMO - though elsewhere in these comments there's a suggestion that ALAC was chosen as it was cheaper to decode on Apple's hardware at the time, which if true would be a reasonable motivation.

The author of Ogg wrote a reply to that article: https://people.xiph.org/~xiphmont/lj-pseudocut/o-response-1....

Anyway, I highly doubt the design of the Ogg container had anything to do with the popularity of Vorbis vs MP3 - most users of Vorbis used the Ogg container, and some of the criticisms in the original article apply to MP3 (ever tried to seek a VBR MP3 file?)

Functionally, it doesn't make a difference if the patents are expired if licensing fees are still being extracted. And they are.

It's still way too expensive for the little-company or little-man to go to court over it, even if it's obviously false.

MP3 seems to be one of the cases where patents seem to have not hindered the spread of information. The inventor made a small amount of money, and decodes and encoders proliferated. MP3 dominated music. And there's no shortage of open source (which, I know, is not what RMS wants) code available.

This is not the catastrophic situation that I was told to expect from patentable file formats.

From https://en.wikipedia.org/wiki/MP3#Licensing.2C_ownership_and...

"MP3 license revenues generated about €100 million for the Fraunhofer Society in 2005."

Is that small amount of money ?

Compared to the value of the music/audio distributed through mp3? I'd say that's a pittance.

Buying a complete set of codecs from Fluendo for Linux costs 20$. That includes Windows Media, h264 and others. That's not too much I think.

http://www.oneplaydirect.com/oneplay/oneplay-codec-pack/

For many years, nobody shipped implementations of MP3 at all; there are still Linux distributions today that don't provide support for MP3 out of the box. Many don't support MP4/h264, either.

In fact, the MP3 patents are directly responsible for spurring the development of Ogg Vorbis, which is arguably superior in some ways (see other comments). From Wikipedia:

> Vorbis is a continuation of audio compression development started in 1993 by Chris Montgomery.[8][9] Intensive development began following a September 1998 letter from the Fraunhofer Society announcing plans to charge licensing fees for the MP3 audio format.[10][11]

This is interesting because "innovation through forced workarounds" has long been one of the ways patents are said to be beneficial. Ogg is a prime example.

I think the situation in the second to last paragraph is now the usual case, not the exceptional case, as defensibility in today's market has shifted from technical feats to network effects.

Reply on news.ycombinator.com