I'm a leecher with a passing interest in the techniques that go into encoding. Just ran through this thread, have accumulated some thoughts on the issues facing the whole "bring h264 to the masses" concept. In particular, what bottlenecks do and don't play a big role (both technical and psychological).
Issue #1: Connection speed isn't an issue.
1. Most people who have an Internet connection have something noticeably faster than dialup (one stat I found said above 80% of residential connections are at basic DSL or better). Even at a fairly basic connection like 768/192 DSL it's quite possible to set up a few torrents one evening, let 'em run overnight, and in the morning you have a couple of shows to watch the next evening. Rinse and repeat.
2. In rural areas like where my parents live now, that have most of the households without broadband access, the people with crappy Net access simply go watch anime with one of their friends that has access. Or watch Cartoon Network. Or buy DVDs. Which brings me to the next point...
3. People who watch fansubs regularly are a highly specialized audience. Going out on a limb here, but I'd say they are disproportionately college students with access to massive university pipes, suburban/urban dwellers who are in broadband markets, technophiles who who go out of their way more than most to get a good Net connection, etc.
4. Finally, there are those folks without much money who simply can't get good Net service. If they are downloading 30Mb craptastic videos off of YouTube on old wimpy machines, then they have no use for h264 anyways (or indeed high quality encoding of any type really). They aren't of primary concern to the torrent-oriented issues facing most people on this forum.
Issue #2: Why filesize shouldn't matter so much.
Lemme start this by saying that I utterly don't get why people would care about forcing all their videos into rigidly defined size brackets. If multipass encoding produces better-looking files, I'll take it. So Ep.1 is 170Mb and Ep.2 is 195Mb, who cares? If you wanna sell people on h264 looking better (which it clearly can), don't gimp your product for the sake of having pretty-looking uniform filesizes. Also, I dont think it's hard to explain to people that your h264 file is bigger because it's the same resolution as the original "HD" raw, while your XviD is smaller mostly because it's not as nice a picture. Just be sure to make it clear on the filenames when you're releasing a version with a higher resolution so folks who care about image quality can go for the better one.
Issue #3: Processor speed is definitely a problem for many.
I think a lot of people in places like this tend to lose sight of the fact that regular upgrading is becoming increasingly unnecessary as computers become capable of doing more and more without straining. My parents have an eight-year-old computer that handles fairly well all the email/web/home office/non-polygon games they care to throw at it. Only thing it won't do is play stuff like h264... it handles youtube and most DivX playback all right. What's going to force many people to buy new machines these days is when their machine is too old to even install the latest OS upgrades... and that timeframe keeps getting longer and longer. 5 years isn't really legacy anymore. I think this is one reason why the XviD downloads continue to outpace the h264, there are an awful lot of people out there whose gear has trouble with the power needed. This bottleneck will fade out over time though... you can build a machine around a 2.16GHz C2D chip for under $500. Powerful PCs for the people!
Issue #4: Software and codecs oh my!
At this point I need to say that I have a Mac, so a) I know I'm kind of gimped by not having a true equivalent to CCCP, and b) a lot of specific names aren't going to carry over. My big issues though are more with usability in general. While it may be true that "push butan, receive XviD" is a sign of crappy encoding practice, for most users downloading software really ought to be that simple. Google codec, click link, click download, drag and drop into appropriate folder, doubleclick video and watch. Any additional steps *required* beyond that mean the maker of the codec or software in question needs to stop pawning off their alpha test as useable. God knows any player whose instructions for use involve the phrase "download additional libraries" shouldn't be touched, let alone be recommended in order to view your fansubs. Quite frankly even asking for folks to use multiple players in order to view your extra-cool-and-funky fansub creates a burden of one more piece of software that needs to be learned and updated. I really don't want to need Mplayer for a handful of files that are bursting with karaoke SFX, let alone have to compile it myself.
I'm a Mac guy, I want one piece of software that can do everything without demanding that I learn everything about it, that becomes the de facto standard because it simply works better. This seems to me to be the source of much of the difficulty surrounding shifting to h264, especially with the matroska wrapper. For things that have been around for years already, the codec and the matroska players seem like they're still in alpha/beta testing, not one solid and widepsread piece of software that cover 99.9% of all files intended for public distribution. If a viewer needs three different players and weekly updates of their codec pack in order to watch your file, that is fail.
I think my biggest beef, both with sloppy encoders and with people making software, is a lack of robustness. Quality checking should involve machines with different codec versions, preferably across multiple platforms. Likewise a player has no business being offered to the public when fairly minor errors in the source material causes it to behave all sorts of erratically (like piling subtitles on top of one another, or completely losing track of timetracks). I think this problem is particularly severe with softsubs, we've got three different major versions that the various playback methods have varying levels of success with. I have softsub files that work in VLC right now... and some that I can't even open with it. I don't think it's just VLC's fault, the matroska wrapper quite frankly seems to allow too many different ways to do too many things. The folks who make Perian, OSX's closest thing to CCCP, have been months working on an update... and the delay is basically trying to find an answer to a lack of timecode enforcement with matroska. Not that I fully understand the problem, far from it, but it seems like their problem is just one more on the list of things that make matroska hard to implement because it's trying to incorporate every feature under the sun. This is particularly an issue with the many viewers who want to download a fansub and have it "just work", people will get rapidly fed up if they download 200Mb files that require a bunch of work to get 'em to play (or they end up trashing unwatched because the highly experimental new softsub method crashes their players).
On a more positive note, I'd like to say that I love h264 compared to XviD. The fundamental leap in image quality that I see is that h264 scales much more smoothly than any .avi I've ever seen is capable of. The problem I run into is that videos done at the current "HD" resolution (somewhere around 864X434? Wider than 700) are kind of small on my 1600X1200 monitor, but doubling them in size makes 'em bigger than my monitor. Making an XviD file fit to the window causes all manner of weird blocking effects that simply aren't there with h264. That's a huge difference beyond the inherent size savings to me. So I'll be downloading every mp4 and mkv I can get from now on (especially now that I found an avc1 codec for QT7 that overrides the *&*%*^ POS h264 codec that Apple provides).
Phew, longpost. thank you for your time, and keep up the good work.
