[vorbis] Dvorak Interviews Monty

[Brian Zisk]

DVORAK:  A lot of people don't realize that the MP3 music standard is
something of a patented technology in some way, shape or form in a way that
requires certain people who make those programs that turn music into MP3's,
they have to pay a licensing fee to some guys in Germany.  And it's a very
unusual situation, and it's starting to become a problem for people who
just want to do hobby-like things with these compressors, these things that
compress music, codec, they're called.  One of the solutions out there is a
thing that's just starting to show up, it's the called the OGG Project,
actually.  And we're going to talk to Monte Montgomery, who is a lead
architect of a company called Xiphophorus that is responsible for
developing a new public domain open source codec that they hope everybody
will end up using.  Monte, welcome to Real Computing.

MONTE MONTGOMERY:  Thank you.

DVORAK:  All right.  Let's go over what you got here.  Give us a background
on what you guys are trying to do to begin with, and then let's talk about
some of these interest projects you have going.

MONTGOMERY:  The largest thing that we're doing right now is an open source
music codec, as part of the OGG Project, the name of the codec is OGG
Vorbis.  And what we're trying to do is basically give the world and the
'Net a completely open alternative to the closed formats that currently
exist.

DVORAK:  Now, I've heard of Vorbis.  Is this different than regular Vorbis
that's been...?  I just heard it, I don't know too much about it.  I know
it's a competitor with MP3.  And I'll just catch people up here.  MP3, as
popular as it is, still is a licensable technology, it costs people money.

MONTGOMERY:  That Vorbis is ours, yes.

DVORAK:  So this Vorbis is the Vorbis that I've been reading about?

MONTGOMERY:  That's correct.

DVORAK:  Okay, and you just added this word, 'OGG,' to the beginning of it.
Which stands for?

MONTGOMERY:  It doesn't stand for anything, it's the name of an umbrella
project that Vorbis is part of.  Right now, Vorbis is most of the OGG
project.

DVORAK:  Okay, it's O-G-G for you OGG freaks out there.  You have a thing
called OGG Squish.

MONTGOMERY:  OGG Squish was the name of the original codec that we were
working on, that's for lossless audio.  That codec is not currently
available, but we plan to be updating that pretty soon after Vorbis comes
out.

DVORAK:  Now is Vorbis going to be lossie, as it were, audio, you do some
information, and the lossless one you don't.

MONTGOMERY:  That's correct.  Vorbis is lossie, like MP3 is lossie.  And
Squish is a lossless codec.  Lossie codecs are fine for music distribution,
but generally in the production process you don't want to be losing
information...

DVORAK:  Right.  Why don't you explain lossie-lossless to people out there
that don't know what you're talking about?

MONTGOMERY:  Lossie audio compression, like Vorbis, MP3, TwinVQ, other
compressions you may be familiar with, generally try to lose information
that is not audible to the human ear anyway, in order to make the files
smaller.  Most of the lossie compressions that are out there right now
don't completely succeed, they still do compromise sound quality,
especially if you use them on the same file more than once.

DVORAK:  Right. Okay, that would make sense.  People out there who want to
see another example, in the imaging world you have a couple of technologies
that are lossie, J-PEG being the most important.  In other words, there's
some information that's lost and they just assume that you're not going to
notice, even though it's noticeable to me.  And then you have another file
format, such as TIFF, which can actually be compressed, but it doesn't lose
any information under any circumstances.  But it's also huge by comparison.

MONTGOMERY:  That's correct.  And the same comparisons apply to audio.  The
lossie codecs are popular because it compresses the information a great
deal more.  The lossless codecs are better quality, but, of course, the
files are comparatively huge.

DVORAK:  Why do this?

MONTGOMERY:  Up until recently, and even now, the amount of space that raw
music takes up on a computer is just huge.  And even though hard drives and
storage space on your PC, now DVD ROM's, and so on, remove most of the
limitations of storage on your PC.  Getting it onto the PC or downloading
it through the 'Net, the straw that sucks the information into the computer
with, is still pretty narrow.  And for that reason, you want to be
transferring as little information as possible.

DVORAK:  Why do you want to do this when MP3 has already taken pretty much
the market by storm?

MONTGOMERY:  Well, two reasons.  The first being that MP3 is aging, and
most people are fairly satisfied with the quality of it, but audiophiles
generally are not.  It is possible now to construct formats that are higher
quality and use less space.  And the big reason, really, is the fact that
MP3 is patented, several pieces of it are patented, Fraunhoffer and Thomson
claim that you cannot make an encoder that does not infringe on their
patents.  And for that reason, the format is not as freely available as
consumers perceive.

DVORAK:  Yeah, I perceived it as being freely available.  The other thing
is I noticed that the quality does improve with the bit rate, but then the
file size obviously increases too, to the point where it's pretty big.  For
people out there who look at MP3's, the typical so-called CD quality, if
you have the player like WinAMP, it tells you the bit rate, which is the
rate at which it was encoded.  And typically, people believe that CD
quality is about 128, although most people like to see 160 or higher, and
then there's lower rates for just compressing the heck out of stuff, which
sounds terrible, it's noticeable.  Let's say a file size of about 5
megabytes for a music file encoded on MP3 with 128K bit rate, what file
size do you think you could get down to using the newer technology that you
guys are working on?

MONTGOMERY:  Vorbis offers the same trade-offs that you have available with
MP3.  People who want more quality obviously can encode at a higher bit
rate.  Because you can encode at any bit rate you want, you can choose 128K
with Vorbis as well, but it will sound better.

DVORAK:  So, in other words, you'll have about the same file size, but it
will sound better?

MONTGOMERY:  You have the same trade-offs available.  You can make it
arbitrarily smaller, and, again, you lose quality precipitously as it gets
smaller.  At every bit rate, the goal, and I think that we've already
achieved it, is that Vorbis beats MPEG on all technical merits.

DVORAK:  Now, I suppose the market will judge that eventually.  Did you
ever decide on a file extension yet?  VOB?  Or do you know anything?

MONTGOMERY:  Yeah, Brian asked me about that.

DVORAK:  That's the first thing that came to my mind.

MONTGOMERY:  The official extension of Vorbis because it is an OGG codec is
.ogg.

DVORAK:  Oh, so you're going to use the OGG thing?

MONTGOMERY:  Yeah.  Vorbis using the OGG bit stream format.  It's a
container around the Vorbis information the same way that Quicktime is a
container around the many codecs that it uses.  And Vorbis will not be the
only OGG codec coming.

DVORAK:  What do you guys use for a revenue model, or do you even care?

MONTGOMERY:  I think you hit the nail on the head there.  We really don't
care.  This is stuff that we would be doing anyway.  We do it for any
number of reasons--the same reasons that open source hackers and
programmers use all over the world.  We are funded, and we do individually,
we are individually compensated for our work on this, but I don't think
anyone in the project is looking to become wealthy off of what we're doing.
We'll well paid to do we love, and that's enough.

DVORAK:  Idealism dies over time, I'll have to tell you right now, but
that's okay.  Who is funding you?

MONTGOMERY:  At the moment, well, we were being funded by the Greenwich
Internet Radio, out of San Francisco here, before they were acquired by
ICAST, and ICAST is continuing that funding.

DVORAK:  I always imagine something like this to be a buy-out candidate
like from Microsoft or someone.  You know, just come along, and the next
thing you know you're working for Microsoft.  Which if I was Microsoft, I
would look at this.  Why not?  Because they were thinking of doing some
sort of a Microsoft codec.  I don't know where they were going to get it,
but they were going to do an MP3 competitive file format.

MONTGOMERY:  Well, they have done that with Windows Media, the .WMA
architecture.

DVORAK:  Is that the music one?

MONTGOMERY:  Yeah, it's music. And it is, in fact, based on MPEG.

DVORAK:  Oh, is it?

MONTGOMERY:  Yeah.  It is not compatible, but it is based on many of the
same things that MPEG that is based on.

DVORAK:  Is it any good?

MONTGOMERY:  Personal opinion?  I think it's not as good as MP3.

DVORAK:  Oh, so in other words, it's no good.  Nobody uses it that I know
of.  Do they?  Or does something use it once in awhile?

MONTGOMERY:  There's a large push by Microsoft to get companies to use it.
How much success they're going to have in the long run, of course, you
can't tell until it happens.  I don't see a great deal of market
penetration for Windows Media.

DVORAK:  What are other competitive technologies out there for this kind of
thing?  Let's go over all of them for people who want to follow this if
they can.

MONTGOMERY:  Most of the competitive technologies that don't belong to MPEG
now are older.  The compression that Sony uses in their mini-disk player;
the Compression Digital Compact Cassette from Philips; AT&T is working on a
compression called PACK, I'm not sure where exactly that is right now, you
hear from them eventually.  But for the most part, MPEG, the MPEG
consortium has managed to assimilate the good codecs, AAC from Fraunhoffer
and TwinVQ from NTT, now Yamaha.

DVORAK:  And what about these music initiative systems that are like, you
know, these guys are putting copyright protection into the system somehow,
and they're using a codec of some sort because it's a MP3-like player?  How
does that work, and what are they up to?  What's going on there?

MONTGOMERY:  Well, basically, a lot of those initiatives are spun from the
RIAA, Recording Industry Artists Association, maybe, I'm probably
mis-expanding that acronym.

DVORAK:  Recording Industry Association of America, I think.

MONTGOMERY:  Association of America, thank you.  And they're very hot to
see all sorts of protections put into the bit stream.  Unfortunately, their
model is not extremely practical, and most consumers who are paying
attention to it have serious misgivings about what their real motives are.

DVORAK:  Are they using any special technologies for the compression at all?

MONTGOMERY:  In SDMI?  I'm not actually certain what the protection is
layered over.  Considering that the RIAA and MPEG Consortium have been
going back and forth a little bit and everyone is expecting, or at least
fearing a grand alliance, we'll probably hear eventually what the
technology underneath SDMI is.  To some extent, unfortunately, I'm
technically oriented paying attention to my own work, and for that reason
can occasionally lose track of what's going on around me.

DVORAK:  You're the lead architect, which means you have to be the guy
that's at the blackboard all the time.  How do you get into this scene that
you're in, which is the compression technology scene, how does that come
about?  What did you do before?  What makes it interesting to you?

MONTGOMERY:  I got into it when I was a college student, and the biggest
reason that I got into is because PC's were just then becoming powerful
enough to do real music work on.  OGG and the codecs that I've worked on
since then have fallen out of a larger project, six years dead, that was
simply trying to do professional quality mastering of audio on a normal PC.
And at the time, a 1-gigabyte hard drive was considered absolutely
cavernous.  And that didn't hold much audio, so you had to compress it.  I
wasn't really satisfied with what was out there already, and I started
playing with it.

DVORAK:  How do you start playing with it?  I mean, who do you hang out
with?  Because this is like a black art, I mean, this is not your business
application.  This is nothing... or even, you know, machine coding a
computer.  I mean, this is something screwy.  It reminds me, of course,
like the guys who do the image stuff, I mean, there's a lot of weird
imaging compression schemes, you know, the fractal stuff, for example.  And
I just can't even imagine what a meeting is like with the architect for
some of these designs.

MONTGOMERY:  Well, actually, that really is how I started working, I
started playing around with it.  Because I didn't know anyone who was
working on this.  I've met people since then, not very many, unfortunately.
But the first steps were grabbing a computer, writing some code, seeing it
didn't work, and then ransacking the library.

DVORAK:  Do you think there's a cross-over between audio compression and
other forms of compression?  Because the people who specialize in file
compression are not the same people who necessarily specialize in image
file compression.  And audio compression is another group altogether.  I
mean, I  don't see that much crossing over, although MPEG seems to have
stemmed from the video compression group.  Although I don't know if there
was a separate code within the MPEG standard or not.  But, anyway, is there
a cross-over that can take place here with these kinds of technologies?

MONTGOMERY:  There's plenty of classical compression in an audio
compressor.  There's also plenty of very audio-specific signal processing,
you won't find that practically anywhere else.  But, like you say, it's a
merger of the two, and picking up any beginning textbook on data
compression, you'll find a lot of those techniques in a typical audio
compressor.

DVORAK:  So you would recommend that somebody... that's where you would
start, is, you know, some kid out there is thinking, 'You know, I can't
believe you can't shrink this even more.'  I mean, that's the goal, right,
just to shrink stuff as much as you can without changing it?

MONTGOMERY:  That's pretty much the goal.  Another goal, actually, is to
teach the people out there who want to learn about it.  When I was learning
this there were people available to tell me about it and there were
references, but not nearly as many as I would have liked.  I would have
liked someone to go and ask questions of when I was stumped.  And there
have been numerous occasions in the past six years when I just didn't know
where to go, and something always occurred to me eventually, but it would
have been nice to sit down with someone and talk it, muse about the various
aspects, play with the problem, tinker with the puzzle with a friend.  And
that's something, actually, that the OGG Project would like to provide to
programmers on the 'Net today because that's something that didn't exist
when I started.

DVORAK:  Well, that's noble.  What is the Web page?

MONTGOMERY:  xiph.org.  ogg.org will also take you to the page.

DVORAK:  Okay, ogg.org, and xiph.org.  Where does that word come from?

MONTGOMERY:  It's actually the name of aquarium fish.

DVORAK:  Oh, that's why you have fishes all over the logo.  So xiph.org
works and so does ogg.org.

MONTGOMERY:  As well as vorbis.org.  xiph.org can be difficult to
remember--and too many technology companies also have X's in their names.
I was out-voted when we chose the name.  So Vorbis and ogg.org as well.

DVORAK:  Well, for all you kids out there who want to get into this, this
is the guy, Monty Montgomery.  He is the lead architect for Xiphophorus,
which is a company working on the OGG Project that will be compressing
music and hopefully people might be standardizing on it in the near future
if they're lucky.  Monty, thanks for being with us today.

MONTGOMERY:  Thank you very much.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.