A Blog From Behind the Trenches

Great explanations

*Ahem* {Didn't Microsoft lose that FAT clawback case?}

Legal precedent is building the opinion that if something is "Free" enough, passed-around enough, you don't GET to take it back and request compensation. MPEG-LA is destroying its own position here.

Well said.

Originally posted by hellspork:

Legal precedent is building the opinion that if something is "Free" enough, passed-around enough, you don't GET to take it back and request compensation. MPEG-LA is destroying its own position here.

No it isn't. It isn't TOTALLY FREE. It's only free for non-commercial use. But that's irrelevant. The owner of a product can choose to make it free and then charge later, no problem at all.

Well, other than all those pesky times when a corporation overextended and couldn't get its own IP back. Non-commercial use is a broad definition, but "free client use" is the draw for producers of hardware and software that consume content. And if half the consumers have CD players, you'd best have CDs to sell to them.

We'll see how the chips fell three years from now, and we'll see the hard line five years from now.

You seem to be confusing trademark law with copyright or patent law. Trademarks can be lost. Copyrights and patents can't (until they expire by themselves, and not out of inaction on part of the "owner").

I'm not sure how anyone could fail to see through MPEG LA's "altruistic" actions here. The drug dealer analogy is pretty apt.

OTOH, H.264 is already becoming "the next big standard" for HD in the broadcasting industry so it's not like the MPEG people are banking on a windfall later; they are earning licencing money right now. There is a chance they will continue to make it free for the web because the collection hassle in 2015 will be too complicated.

But you're right, it's definitely not a safe bet.

As I read it this promise only covered content providers. So browsers wishing to implement the decoding technology still need to pay the MPEG LA much cash, right?

Here's something interesting I found:

Nuanti brings HTML5 and Ogg Theora video to Silverlight

Originally posted by ars technica:

Nuanti, a company that develops Web browsing technologies, has produced a high-performance Ogg Theora decoder for Microsoft's Silverlight browser plugin. Nuanti's Highgate Media Suite will enable support for standards-based HTML5 video streaming with Theora in browsers that have Silverlight. It works entirely without requiring the users to install any additional software.

According to Nuanti developer Alp Toker, the company plans to open the source code in order to enable broader adoption of open and unencumbered video technology on the Web.

Well that's good news. But I am talking about h.264 IP itself. MPEG-LA may never be able to charge for the playback side again, after this. Provider fees are another matter. Also, this term is currently for the next five years. Perhaps MPEG-LA will be trying to charge us for use of the "next" method.

Please bear in mind that 1) improvements in data efficiency will soon end and 2) improvements in calculating efficiency have similar constraints and 3) nearly EVERY basic theory of video compression turns out to have been preempted by ANOTHER basic theory of video compression.

So the race for a better codec theory will soon end, and the growing list of "soft" patents has resulted in more of them being thrown out. I believe I do agree with Opera's "open patents, guarded sources".

Originally posted by hellspork:

MPEG-LA may never be able to charge for the playback side again, after this.

Stop this nonsense. They can stop charging and start again as much as they want. You are confusing copyright law with trademark law. TRADEMARKS are lost if you don't protect them. Copyright is not.

I believe the fear in this article and shared by some people is largely unsound and not particularly forward-thinking. In all honesty, h.264 probably won't be the codec of choice in 2016 so I don't see the worry for video providers who are looking to stay on the forefront of video quality. 6 years isn't "a few years" in the internet, it is like an eon in internet time. What was popular in online video 6 years ago? Hard to say because YouTube didn't even exist in 2004. Mpeg-4 ASP, h.263 had been around for a little bit but I'm not sure how much they were used. vp6 was the new kid on the block but I'm not sure it was used much that far back. I mean did Flash even decode vp6 back then? I recall RealVideo and WMV were still really popular. OK, chinese users will debate me there as apparently rv40 is still popular in china for some reason.

By 2016 we'll at least have the free VC-2 (though the theoretical efficacy of wavelet based encoding is debatable), h.265 (exected ~25% efficiency gain over h.264 and likely another period of free consumer licensing), and maybe another On2 codec depending on what Google does.

One of the great things about not setting a specific codec for html5 is that it allows the technology used with the video tag to progress faster than the html standard. Imagine having to wait for a new standard to be able to watch video with a new codec. Imagine the internet video landscape being the same in 2016 as it is now. That's like someone in 2004 expecting people to still be watching 320x240 videos using RealPlayer in 2010. Setting a single video codec standard, be it theora or h.264 or any other codec, that would sit stagnant for 6+ years would be bad for the internet.

Originally posted by Dillon:

One of the great things about not setting a specific codec for html5

It's not setting one specific codec as the only codec, it's saying at least this specific codec has to be supported.

Originally posted by Dillon:

That's like someone in 2004 expecting people to still be watching 320x240 videos using RealPlayer in 2010.

That's still pretty much the default YouTube video quality (albeit with Flash). More to the point, Theora won't be the same in 2016 as it is now either so you're pretty much doing what you're accusing "some people" of.

The Theora codec has certain technical limitations that make it impossible to reach the quality of other more modern video codecs. As far as I know, it could about reach the quality of mpeg-4 ASP and that's being generous. This is largely because it has to avoid using patented encoding methods like b-frames. Sure it will be better by 2016, but only to a limit and an encoder like x264 will have also been making progress during that period, while other, possibly better codecs have been released.

Theora matches h264 for YouTube playback. And this is web streaming we're talking about, not downloadable videos.

But stop it with the quality insanity. Theora only needs to be good enough, and that's it.

Also, Theora would just be the baseline codec.

And you don't know what happens in 2016. Also, they could easily change the license terms. Also, the free period is just for FREE STREAMING videos. Everything else will cost craploads of money.

Really? Also even if that were true for Theora, Vorbis is an outstanding audio codec. H.264 more or less comes coupled with AAC.

Edit: I hadn't seen Purdi's reply while I posted this and he was ahead of me. Regard this as an addendum to his post.

Well Theora, similar to certain XviD filters, is best with animation and not as good with live action. Definitely not on the level of h.264 or x264 in data efficiency, better on CPU but not as much as one may hope. That's what I've found in the last few days.

Any "minimums" for the tag need to be free on the playback side, or they won't qualify to be part of the standard. That explains the clarification of h.264 playback rights. So any company should be able to embed a decode-only core for free.

The uncomfortable reality of Flash is that it works on Mac/PC/*NIX already, provides a stable and backward-compatible set of features and methods, and comes in a feature-limited type for weaker hardware (though promises abound of HW-accel and ActionScript coming soon for mobile). To compare, most W3C standards are not equally (and certainly not completely) implemented across browsers. I know the pages on Wikipedia are never quite accurate, but the features comparison and standards comparison serve to demonstrate what I'm talking about.

For that matter, last I checked not even ONE browser fully supports the SVG methods yet (ANY single SVG method has unsupported components as of the moment) ....

Actually, Theora requires LESS CPU for decoding/playback than h264.

Flash is security hell, and it will die.

Originally posted by hellspork:

better on CPU but not as much as one may hope

Quoted for ShutUpShutUp appeal. Depending the settings, DEPENDING THE SETTINGS. There are cases where h.264 can be much more data-efficient with the same rez and quality, and only slightly higher CPU usage. However once you enable VFR and force composited decoding at 15/24/30/60fps for different parts of the picture at the same time, CPU usage exponentially grows.

I also did not say Flash was safe, nor did I say it offered a free or open playing field on the supply side. Though it has many problems, Flash is a well-known cross-platform plugin and IT IS FREE TO THE VIEWER. That's all it took for Flash to stay on top until now. Flash can die when we've successfully replaced it, just as ActiveX is a neglected relic.

Originally posted by Cutting Spoon:

The uncomfortable reality of Flash is that it works on Mac/PC/*NIX already, provides a stable and backward-compatible set of features and methods

It doesn't provide the other kind of stable, though.

The trouble is that theora isn't "good enough". That one comparison by Xiph is the only comparison I've ever seen that made theora look even comparable to h.264 and it was produced by Xiph, who leads the ogg group. Using one non-independent source of information probably not a good idea. A real-world example like the (now, sadly non-existent) tinyvid.tv website demonstrates that theora requires at least 2x the bitrate of x264 to reach comparable quality. Here's an even better idea, get x264 and get the theora encoder and compare yourself. Look up what Youtube bitrates are for different resolutions and encode to those bitrates (this may be hard since the rate control in theora doesn't always work).

Here's a piece from the same discussion thread about Theora:
http://forum.doom9.org/showthread.php?p=1335958#post1335958
Notable people are Ben Waggoner and Dark Shikari though a lot of people on doom9 know what they're talking about. Ben Waggoner is an authority on video codecs and works on Silverlight. Dark Shikari is also an authority on video codecs and he's a lead programmer for x264. That certainly gives him a bit of bias but note that he knows more about video codecs than Xiph and the ogg team. He's even given them pointers about things to add to the theora encoder - one of which has been recently implemented (adaptive lambda). The thread also goes into some of the patent issues with theora. There's some doubt about how much the patent status of vp3 was researched because apparently Xiph just assumed On2 had done all the research before releasing vp3 to them. Word is there's at least one company that still claims patents on vp3, so that would be a big issue. I'm curious if Opera, Mozilla, or other groups for whom the free-ness of theora is such a concern have hired lawyers to research the matter or are they accepting Xiph's/On2's word that theora is free.

I'd like to add that the current theora decoder implementation (there's only one) is actually slower than many of those for x264 since the theora decoder is not multithreaded and otherwise inefficient. Hopefully this is resolved though since frame-based multithreading probably wouldn't be too difficult to implement and someone could put the work in to add assembly optimizations. That being said, the current implementation of decoding and presenting in Opera (despite some known issues) seems to have much better performance than that of firefox, which is good Hopefully browser developers will learn quickly about the foreign subject of video processing and presentation.

Originally posted by Astrophizz:

The trouble is that theora isn't "good enough".

This is a blatant lie, and just more FUD. Also, you are dishonestly using straw men and red herrings. I specifically said GOOD ENOUGH. I never said the quality needs to be perfect or exactly as good as everything else. I CLEARLY pointed out that quality is one of the least important things to most people.

Originally posted by Dillon:

The trouble is that theora isn't "good enough". That one comparison by Xiph is the only comparison I've ever seen that made theora look even comparable to h.264 and it was produced by Xiph, who leads the ogg group. Using one non-independent source of information probably not a good idea. A real-world example like the (now, sadly non-existent) tinyvid.tv website demonstrates that theora requires at least 2x the bitrate of x264 to reach comparable quality. Here's an even better idea, get x264 and get the theora encoder and compare yourself. Look up what Youtube bitrates are for different resolutions and encode to those bitrates (this may be hard since the rate control in theora doesn't always work).

My phone can play videos in a certain format. Some kind of MPEG-4 I believe, or maybe it's even some implementation of h.264. Anyway, that doesn't really matter. The point is the quality of the videos I encoded to watch on my phone can be quite awful and I won't be bothered too much. When I tried to save some space by reducing the audio to 22kHz, now THAT was unacceptable.

In other words:
High quality video, high quality audio: awesome.
Mediocre quality video, high quality audio (this is where we'd put Theora compared to h.264 at similar bitrates I'd say): definitely good enough.
Low quality video, high quality audio: Definitely acceptable.
Mediocre quality video, mediocre quality audio: Probably acceptable, but more annoying than low quality video.
Anything with low quality audio: unacceptable.

Of course this is just my opinion, but I think the threshold for "good enough" for video is much lower than for audio (at least speaking in compression compared to the full signal).

I mean, you keep saying Theora isn't good enough because it's about the same quality as DivX/XviD, but tell you what, I loved streaming those DivX videos back in '05 or '06 or so. Was still better quality than what YouTube gives me now.

Also, that discussion you linked to highlights precisely why we need Theora. That doesn't mean they shouldn't keep improving it or perhaps design a new patent-free codec from the ground up, but Theora is here now as the patent-free alternative to the others.

Originally posted by Frenzie:

It doesn't provide the other kind of stable, though.

Yes! This is true, but at least it's all-or-nothing. Bad side in *NIX is the system dependencies that you don't have with Windows. Take for example the lack of Flash in Opera between that Adobe update and the release of Opera 9.5; nothing could be done about that. Still not the headache of non-overlapping HTML/CSS/SVG, or the different methods of GIF/ZIP/PNG that force me to use different programs for the same task. This is why Opera advocated Capability-sniffing as a better alternative to simple useragent parsing.

As for the supply-side licensing of h.264, $5M sure sounds like a hell of a lot of money. But look at the cost of operating YouTube, a paltry $5M is the least of Google's concerns. Let's assume that "good enough" requires 1.5x the data in Theora. If all files available in HD required 1.5x the storage capacity and 1.5x the bandwidth, would it cost MORE or LESS than $5M to cover the extra capacity needed? I'd say it would cost MORE. If h.264 was discarded and high-quality video was 1.5x bigger, how much would it add to the cost of iTunes' video division?

In many cases, individuals will not be pushing the volume of downloads needed to make h.264 "cheap"; but established companies currenty HAVE the volume to make it cheaper. The growth potential in Theora is entry-level video distribution, where royalties would be prohibitive to a low-cost market.

But you know what? You can sign a storage contract with Rackspace, pay the 20c/file h.264 distribution royalty, and still make 50% net profit without owning any hardware directly. Free is fine, Open is fine, but the TCO is all that matters to companies. They won't use it if there isn't money in it, and they can only code for a "known" playback target. Flash has many limitations but "video for Flash" will (mostly) work with any recent Flash in any browser. This is why most media players natively support .FLV and .MP4 files now.

On the playback side: if Opera needed to pay 1c per download of its browser in royalties, that would be $100K/10M downloads. That's not TERRIBLE but Opera couldn't afford to pay individual royalties for every feature the browser supports. The ideal for providers of free software is to use other providers' free code to add new features. Barring that, they would like to develop their own code to support a royalty-free method or IP. The carrot now being dangled to Opera, Mozilla etc is 'royalty-free decoding', which would grow the supply-side (volume distribution licensing) market.

With GStreamer or a similar cross-platform core, Opera will have options for the future (especially Opera Devices and eventually Opera Mobile). After years of providing a highly extensible framework, Opera will now be able to bundle a media-playback toolkit into this framework; this could grow in the direction of a browser-resident media player, dedicated home-theater widgets or cheaper multimedia support for set-top boxes and media tablets.

So Mozilla will be in a bind unless Fennec can grow their device market, but Opera has the potential to offset some multimedia royalties by including them in the cost of third-party products designed using the Opera Devices SDK. A Devices based in-car computer could use Opera to play DVDs and digital media, a revamped Internet Channel on Wii (Local Media Channel? Multimedia Channel?) could play video from memory sticks. Adding direct, baseline media support APIs will improve the marketability of Opera's non-desktop products; it could also 'push' even more new features to the free desktop browser as a worldwide testing ground.


Like it? I take checks.

Originally posted by hellspork:

If all files available in HD required 1.5x the storage capacity and 1.5x the bandwidth, would it cost MORE or LESS than $5M to cover the extra capacity needed?

This is nonsense considering the fact that they are going to make licensing more expensive. And if they get a monopoly they can pretty much make is as expensive as they want to, and let the money flood in.

Free is fine, Open is fine, but the TCO is all that matters to companies.

Only a silly man indeed will ignore the fact that giving a commercial company a monopoly and the ability to dictate the terms to their liking is a crappy idea. TCO needs to take into account what happens over the next few years, not just what things look like today.

So you are not disputing that current royalty is acceptable on a large enough scale? If anything I would expect MPEG-LA to possible reduce the per-unit and increase the cap. This would make entry-level operations more feasible, while increasing the revenue from major operators like YouTube.

The current royalty is not acceptable, but that's not the point. The point is that giving these people a monopoly lets them to whatever they want to.

There's an old maxim that goes, "beware of greeks bearing gifts". (a reference to the Trojan war)
That is the point here. It doesn't matter which standard is better in quality. If we adopt a non-free codec as a web standard, we give the owner of that IP a licence to exploit users of the internet.

Unfortunately this h.264 nonsense is a monopoly based on technical merit. The largest operators all use it for their high-def because they can't afford the alternative.

Theora has excellent merits in the entry-level market, but it is not as robust for streaming and it is not as data-efficient. I see no reason that both could not coexist for now. If Opera can add h.264 to its GStreamer implementation without fear of royalties for the next five years, I say they should do it. Then we can all meet back here in 3-4 years to flame each other over the next new compression method.

Originally posted by hellspork:

Theora has excellent merits in the entry-level market, but it is not as robust for streaming

BS.

If Opera can add h.264 to its GStreamer implementation without fear of royalties for the next five years, I say they should do it.

Again, BS. This would give the H.264 morons a monopoly, and they could start charging even more outrageous prices at any time.

You just don't get it, do you?

They are extending the free period for free streamed content to fool gullible people into thinking that everything will be all right!

Purdi, I hope you realize your statements are just your suppositions. There's no factual context to your statements. If there is, then please post a link - I'd be interested in reading about Mpeg-LA's grand scheme of deceit.

Something that may interest people: http://doom10.org/index.php?topic=170.msg1471#msg1471
One highlight: theora devs themselves don't consider theora to be nearly as good as h.264

Originally posted by Dillon:

Purdi, I hope you realize your statements are just your suppositions. There's no factual context to your statements. If there is, then please post a link - I'd be interested in reading about Mpeg-LA's grand scheme of deceit.

haavard's post on top of all the comments? http://my.opera.com/haavard/blog/2010/02/04/h264-trickery

Originally posted by Dillon:

One highlight: theora devs themselves don't consider theora to be nearly as good as h.264

They'd be crazy if they did, lol.

Originally posted by Astrophizz:

Purdi, I hope you realize your statements are just your suppositions. There's no factual context to your statements.

WRONG. The fact is that they can change the pricing at any time. So only a complete and utter moron would fall for their games.

If there is, then please post a link - I'd be interested in reading about Mpeg-LA's grand scheme of deceit.

You are incredibly naive if you think they are into this for anything but making money. The way to make as much money as possible is to try to kill Theora. If they manage to do that, they will have a monopoly, and can charge even more outrageous sums than today.

One highlight: theora devs themselves don't consider theora to be nearly as good as h.264

Theora doesn't have to be as good as h.264. It only needs to be GOOD ENOUGH. Pay attention instead of spouting nonsense.

So you would advocate a noticeable regression in online video quality in order to use an open video codec?

Since h.264 isn't or is barely being used properly yet it'd be more like maintaining the status quo than having a noticeable regression. x264 is the only reason I like the format for personal use, the typical Apple h.264 junk is, well, junk. But I'd personally go for x264 in an MKV container with Vorbis audio, or even if I went Theora + Vorbis I'd still prefer an MKV container. For an open Internet, I'll gladly take the "horrible quality" of Theora.

Originally posted by Astrophizz:

So you would advocate a noticeable regression in online video quality in order to use an open video codec?

There wouldn't be one. Look at the YouTube comparison.

People will not notice it.

So stop spreading FUD.

Originally posted by Astrophizz:

So you would advocate a noticeable regression in online video quality in order to use an open video codec?

It worked for Mozilla and most of what they've done with Firefox. /rimshot

If MPEG-LA got out of hand, Google would probably throw money at the problem and acquire/open VP8 in future. Simply put, the quality and bitrate of hand-tuned Theora examples are not very impressive. The online viewing experience is currently buggy and reminiscent of Windows Media plugins in the early 2000's.

Mozilla desperately wishes to avoid paying for any of this, which may shed some light on the enticement. With both Mozilla, Opera and Internet Explorer currently not supporting h.264 in any native form, the fact that Mozilla and Opera now support Theora is a valid threat. So h.264 playback must be free for browser vendors, or it will not gain ground.

On the supply side, Theora is NOT a very attractive proposition and it only continues to age. One might say that Theora currently bears the trappings of a stillborn codec, never even reaching a RealPlayer-style peak before sinking under its own primitive weight.

MPEG-LA has thrown a fortune into expanding the range of playback support, all to expand the sales market for its encoder. Pushing h.264 as a native HTML solution takes customers away from Adobe's Flash, and forges the opportunity for expanded direct licensing of the h.264 encoder. It's fine if the decoder is free on 1Bn desktops and devices, if every supplier must then distribute (and pay for) the master format.

Theora does not (will not, can not) have all of that.

Originally posted by Cutting Spoon:

Theora is NOT a very attractive proposition and it only continues to age.

?

There was an upgrade just a few months ago. It isn't like there has not been an upgrade in years.

I think Cutting Spoon is referring to the theora specification, which is frozen and based on VP3 which was already a bit old before it became theora. The dated specification creates certain limitations that the theora devs have to work around. This is well-understood by the devs themselves, though misinformed people outside of the coding community seem to believe that theora can be and do anything.

As for the encoder, they've started a new dev branch again, called "ptalarbvorm" (I have no clue where they get these names). They added a suggestion for better AQ from a lead x264 developer but the update is at a really early stage. It helps quality a bit though.

Originally posted by Dillon:

called "ptalarbvorm"

Where did you hear about that?

Originally posted by hellspork:

If MPEG-LA got out of hand, Google would

Ah, you are betting the future of the web on the hope that it will magically solve itself somehow! Not exactly a good tactic.

Simply put, the quality and bitrate of hand-tuned Theora examples are not very impressive.

This is a lie. Pure FUD. Theora is more than good enough.

Mozilla desperately wishes to avoid paying for any of this

No, Mozilla, like Opera, wants the foundation of the web to be free and open.

On the supply side, Theora is NOT a very attractive proposition and it only continues to age.

Wrong. It's more than good enough, and can be improved even further.

One might say that Theora currently bears the trappings of a stillborn codec, never even reaching a RealPlayer-style peak before sinking under its own primitive weight.

More FUD. Will you never stop spreading FUD about Theora?

Originally posted by Astrophizz:

This is well-understood by the devs themselves, though misinformed people outside of the coding community seem to believe that theora can be and do anything.

Ah, the straw man. Sign of desperation.

@Daniel Hendrycks:
http://svn.xiph.org/experimental/derf/theora-ptalarbvorm/lib/internal.h
Under "/*This library's version.*/" (or in the url ).

@Purdi: I'm not desperate, but fine let's say there's noone who believes theora is capable of scaling with other more modern codecs. No more "straw man". Yes theora can be improved further but I would argue that it is not "good enough" and I don't think it can close the gap with the current theora specification. Maybe if they extended theora but I don't know if that will happen - I don't think there are any plans to do this.

Originally posted by Dillon:

I'm not desperate, but fine let's say there's noone who believes theora is capable of scaling with other more modern codecs. No more "straw man".

Even if unspecified random people thought that, do you really think people at Opera think that? Now Mozilla's Asa I'm not so sure of...

Originally posted by Dillon:

Yes theora can be improved further but I would argue that it is not "good enough"

Amd why not? XviD always served me well, although x264 serves me slightly better. I haven't tried Dirac so far.

Originally posted by Frans:

I haven't tried Dirac so far.

It is very slooow. While Schroedinger is fast, but crappy. If they would dump the hybridness they wouldn't have to develop for a lossy and lossless side. We already have a lossy codec, Theora. Let's just get a lossless codec that's not a hybrid. How about something like "FLAV". Free lossless agile video codec. It sounds like FLAC, the lossless audio codec.

Originally posted by Purdi:

This is a lie. Pure FUD. Theora is more than good enough.

And "Good Enough" is by definition "Not. Very. Impressive." We're getting to the point where $300-400 netbooks possess the extra hardware to play high-def h.264, which is VERY impressive considering that none of them could do it last year. The year before, many new laptops couldn't do it. And the year before THAT, many new PCs couldn't do it. Theora's requirements have always been much lower, but Theora's quality has not evolved to match h.264's improvements in efficiency. There's no reason that BOTH could not coexist - Theora LQ for systems with older hardware - h.264 HQ for newer machines. In that case the data rates would be similar, providing more predictable bandwidth consumption.

Now, I am quite partial to the occasional foreign film or television show. These videos are commmonly coded by students attending college or university. There are two primary formats of distribution, often supplied in parallel.

1) .AVI, coded by DivX or XviD spec, permanent subtitles, stereo sound, often 720 or 800 pixels wide.

2) .MKV, coded by h.264(common) or x264(uncommon), overlay subtitles (higher system requirement, better quality, no recompression or artifacting), AAC 5.1 surround sound, 1280x720 or 1920x1080 resolution.

*) Since 2004, I have seen exactly ONE(1) series coded with the .OGM container. It had permanent subtitles and mp3 stereo, 2.5GB for 13 episodes at 640x480. Roughly equivalent to most DivX releases in quality, bitrate, CPU.

Concerning the above, what I find to be astounding is that titles released in both formats are NEARLY THE SAME SIZE. Further, competently coded h.264 is silk-smooth in 1280x720p on most N270 (or better) netbooks with a decent decoder. So if 720x480 hardsubbed video in DivX or Theora takes 200MB per half-hour, and 1920x1080 softsubbed video takes 300MB, what's the point in accepting anything less? A few groups have even switched to h.264+AAC for their 720p LQ releases, achieving good video quality in less than 150MB/half-hour.

I appreciate Theora, I think it is a valuable weapon; I will NOT elevate it to the status of a direct competitor to h.264 OR wmv. My hope is to see both h.264 and Theora in ALL major browsers within the next year. Anything less will hold back the revolution of online video.