Re: all you MP3 people out there... a fix for lars!

[Lars Magne Ingebrigtsen <larsi@gnus.org>]

wmperry@aventail.com (William M. Perry) writes:

>   One thing I _cannot_ believe is that the cddb people require a specific
> _ordering_ to the stuff in the query.  Completely BGOUS.  That's the whole
> point of having NAMES for those damn things in the format, right?  *sigh*
> Even after 5 1/2 years of working on web stuff, people still come up with
> ways to disgust me. :)

Everything about the cddb format disgusts me.

Let me count you the ways, but first, here's a typical cddb entry:

# xmcd CD database file
#
# Track frame offsets:
#	150
#	15660
#	42115
#	70277
#	94335
#	127810
#
# Disc length: 1869 seconds
#
# Revision: 0
# Submitted via: IdiotJukebox 1.0 (Emacs 20.3.1)
#
DISCID=46074b06
DTITLE=Isotope 217 / The Unstable Molecule
TTITLE0=Kryptonite Smokes The Red Line
TTITLE1=Beneath the Undertow
TTITLE2=La Jeteé
TTITLE3=Phonometrics
TTITLE4=Prince Namor
TTITLE5=Audio Boxing
EXTD=
EXTT0=
EXTT1=
EXTT2=
EXTT3=
EXTT4=
EXTT5=
PLAYORDER=

Ok, we have something that looks like comments first, and then data.
But, no.  Everything in the "comments" section has to be there, and
has to be there in that order.  First we have the track frames, and we
see that they cleverly neglected to include the concluding track
frame, which means that there really is no way to say exactly how long
a CD is.  One can approximate by looking at the "Disc length" entry,
but that's only with second granularity.  So, they throw away info.

That DISCID thing looks really clever, doesn't it?  32 bits; that
should be four billion possibilities, right?  Nope.  The last byte is
the number of tracks, so this byte is usually in the 09-0f range.  The
two previous bytes are the track offsets added together (!), which
means that the range used here is limited as well.  That leaves the
first byte, which is actually uses a bit of %-ing to get at, which
means that it's pseudo-random.  I guesstimate that the likelihood that
two cd's generate the same ID is about one in a million, which is
quite a feat.

Next, we have the DTITLE.  We see that they cleverly put the artist
and the album title in the same string, separated by " / ", which
means that the "/" characted can't be included in an artist name or
album title.  Groovy!

Then the track titles follow, and if a title is longer than 70
characters, one gets

TTITLE0=Big Car (Remix Vocal Edit) very very very very very very very very
TTITLE0= very long title

Clever, huh?  

Then, at the end, we have the comments, and no client ever fills out
track comments, but they have to be there anyway.

I really couldn't have come up with a more stupid format if I were
stoned out on alcohol.  It's totally unflexible, loses information,
un-extendable, disallows certain inputs, and is verbose.

Ptui.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen