[Playlist] chatlog

Lucas Gonze lgonze at panix.com
Tue Mar 30 22:01:42 UTC 2004


IRC log relative to RDF/XML:

key88sf	ok - what kind of stuff got simplified?
lgonze	the biggest thing was getting rid of namespaces
lgonze	wiped out mm: and dc: !
lgonze	http://playlist.musicbrainz.org/playlist/moin.cgi/ 
DefaultFormatInRdfVersionFour was the result
key88sf	looks
lgonze	there's a lot in there. we should talk out the details of the  
changes so you know what the goals were
key88sf	I dont get why you wiped out namespaces?
lgonze	we weren't getting as much from them as we thought we would
key88sf	it doesn't hurt to have them, does it?
lgonze	really, there just weren't that many external tags from foaf,  
musicbrainz and dublin core that we couldn't live without
lgonze	well, the hurt was that the format was too verbose, and writing  
software would require you to look up too many things
key88sf	I thought the point was that dc:creator uses a none which is  
already publicly accepted
key88sf	*node
key88sf	er...but that just changes the nodename
key88sf	either way, you use lookup("creator") or lookup("dc:creator"),  
yes?
lgonze	yah, but danbri's email the other day made that less obvious
lgonze	for foaf they went through all kinds of trouble to get the right  
thing to happen
lgonze	with dc:creator
lgonze	but look at it as a cumulative thing. the impact of all the  
little changes is to that the current stuff is a lot more  
comprehensible than before.
lgonze	no more foaf, either!
key88sf	ok..
key88sf	so what does RDF buy us?
lgonze	that's a good question.
key88sf	heh
lgonze	I like your comment on the list about reopening that  
conversation.
key88sf	thanks - yah I'd like to as well...
key88sf	because after implementing this
key88sf	it seems kinda useless
lgonze	lol
Russss	no mm? shocking :P
lgonze	a brutal round of culling
key88sf	btw -- if no mm, maybe we should change "Track" to something  
else
lgonze	*with mb folk involved* !
key88sf	liek "media"
Russss	yeah, looks good
key88sf	since this can be a playlist of any media type
lgonze	on webjay we've had fun times mixing in video with audio
lgonze	hang on, I'll pull up an example...
lgonze	http://webjay.org/by/lucas_gonze/mydaywasworsethanyourday
lgonze	got to open that in Real, though. sorry winamp folk. :)
lgonze	so what's next?
lgonze	the RDF conversation I guess
key88sf	so what do you think about renaming trackList and Track to  
mediaList and media
lgonze	I don't really have a deep understanding of RDF.
key88sf	and also RDF convo
lgonze	about track->media I don't really have an opinion
lgonze	in my own code for handling playlists my variables for these  
things call them media rather than tracks or songs
lgonze	but in user-facing stuff I call them songs.
key88sf	mine too
key88sf	has someone compiled a list of all the known nodes for a  
<Track> ?
Russss	we're trying to standardise on "track" as opposed to "song"
Russss	track is a nicer name I think, since song implies singing :P
lgonze	heh
lgonze	I guess I like track a little better than media for the container
lgonze	because media to me implies that this is the mp3, and we put the  
mp3 into the id or cid
key88sf	track is more specific to CDs though
Russss	it becomes a lot clearer for classical music though, you  
wouldn't talk about a classical song unless it involves singing
Russss	I dunno, "track" existed before CDs :)
lgonze	intuitively, these things seem to me to be "items" or "entries"  
more than either tracks or songs.
lgonze	I like track a little better than media because it's a little  
more generic, that's all.
|<--	RJ has left irc.freenode.net (Remote closed the connection)
key88sf	ok
key88sf	so how are we going to talk about RDF ?
key88sf	also - is there a listing of the possible nodes we know about  
for <Track> ?
key88sf	artist, album, duration, etc???
lgonze	I want to talk to dan brickley for a little while about it. or  
at least somebody with a deep understanding who's not too partisan.
key88sf	will you be able to do that within the week?
lgonze	about nodes, see the first example Track at  
http://playlist.musicbrainz.org/playlist/moin.cgi/ 
DefaultFormatInRdfVersionFour
lgonze	yup.
lgonze	also, need to find out how to adapt cc: to an RDF-less world.
key88sf	so that first one is all the possible nodes that are predefined?
lgonze	yup.
key88sf	what about mimeType ?
lgonze	the one thing it lacks is that it doesn't use multiple cid's
lgonze	don't know if there's consensus on mimeType
key88sf	publisher, copyright?S
key88sf	trackNum ?
key88sf	i guess tracknum is there
lgonze	the way we reached consensus on those nodes is that they were  
useful for fuzzy search
lgonze	I objected to having song name, artist, album etc, because I  
figured it was all going to be metacrap
lgonze	my preference was to put anything like that in the annotation or  
title. just let it be loose text.
key88sf	loose text is impossible to parse
key88sf	you need a way to get the info
key88sf	even if it's crap
lgonze	and move the good song metadata to external lookups using freedb  
or mb or something like that
lgonze	well, this is a loose consensus
lgonze	I'm just explaining my own perspective.
key88sf	ok
lgonze	I was convinced to have this stuff based on it being useful for  
fuzzy search
lgonze	that's why song runtime is in there, because it's a great  
identifier.
key88sf	sure
key88sf	I agree
key88sf	I think it's good to have as much info as possible, too...so  
players can choose what to use and what not to use.
key88sf	at least about the song
lgonze	y
lgonze	some players will have really powerful metadata lookups, some  
will want to rely on what's in the playlist
key88sf	right
lgonze	....phhh.... what else?
lgonze	the todo items are are to figure out what value we're getting  
from rdf at this point
key88sf	yah
lgonze	so, next up is criteria
lgonze	my feeling on this is that it's hard and deserves a subproject  
of its own
key88sf	well the thing is, I'd like to use it pretty soon :)
key88sf	I have a working version of it already
lgonze	:)
key88sf	which is basically what I published
lgonze	that is awesome
lgonze	coffee's boiling, brb
lgonze	back
key88sf	wb
lgonze	.......thinking......
key88sf	that's another reason I think RDF is useless here
key88sf	it forces Criteria nodes to be RDF compliant too
lgonze	off topic:
lgonze	a friend of mine just emailed that he's going to work on a c  
parser/generator
lgonze	I guess what I'm thinking about Criteria nodes is where to plug  
them in
key88sf	how about after the trackList ?
lgonze	what makes criteria hard is that it's an information retrieval  
problem. it's a relevance query. there's a lot of room to try different  
approaches...
lgonze	...
lgonze	not in the tracks themselves?
key88sf	no
key88sf	it's a criteria for generating an entire playlist
lgonze	is there a usecase where a person has one criteria followed by  
another?
key88sf	from your collection
key88sf	there should only be 0 or 1 Criteria nodes in a playlist.
lgonze	hang one, I'm going to go back and re-read your Criteria notes
lgonze	a thing you need is standard metadata field names
lgonze	to query against
lgonze	that X- notation has gotta go, dude.
lgonze	use a namespace
key88sf	you're reading a way old version
lgonze	:)
key88sf	hang on
key88sf	http://playlist.musicbrainz.org/playlist/moin.cgi/ 
OpenStandardPlaylistNamespace
lgonze	thx
lgonze	here's the direction I'm leaning --
lgonze	I don't want to define a canonical criteria format right now
lgonze	but I think it's a useful thing to do, so I don't want to do  
anything against your project.
key88sf	should we have a version number for the spec?
lgonze	my feeling is that this criteria format could come from a  
different namespace, and that will lead to the right result.
key88sf	so if it gets revised later, then parsers can know what to do
lgonze	about version number, I don't know.
key88sf	or perhaps the namespace url has the version
lgonze	ruaok was for versioning using the namespace URL. yalaforge and  
myself were for adding features later by adding a second namespace for  
the new features.
key88sf	but the new version may use the old features in a different way
lgonze	yeah, we'd have to bump up the version if we broke backward  
compatibility
lgonze	n.b. I have some features that are really important to me that I  
haven't tried to get into the common spec. I've been figuring on adding  
those as an external module.
lgonze	http://playlist.musicbrainz.org/playlist/moin.cgi/ 
OpenStandardPlaylistNamespace is a lot fresher.
key88sf	hey I gotta go --back in like 1 hr
lgonze	k.
lgonze	give me a should with lgonze: msg if you want my irc client to  
let me know you're there.
lgonze	shout
key88sf	k
-->|	ruaok (~robert at adsl-66-122-79-22.dsl.snlo01.pacbell.net) has  
joined #playlist
=-=	icepick-away is now known as icepick
yalaforge	is back (gone 03:22:09)
yalaforge	reads the backscroll
lgonze	welcome back
yalaforge	thanks
ruaok	so, is it a forgone conclusion to say that RDF is getting  
tosssed? :-)
yalaforge	seems so. I'm not happy with that though
ruaok	and in thinking about that, it makes sense.
yalaforge	may I ask who key88sf is?
ruaok	If we're not going to take advantage of dc:, mm:, faof: then why  
bother with RDF?
ruaok	has no clue on key88sf
yalaforge	right, that's the thought. and if we want to support criteria  
stuff, RDF syntax makes things more difficult
ruaok	yalaforge: I trust you want to see RDF stay?>
yalaforge	I'm no longer sure about that
lgonze	key88sf is daveb
yalaforge	cool. Hi dave!
yalaforge	oh, he's away
lgonze	so, rdf
ruaok	you mean xml?
lgonze	phew
yalaforge	we don't use much of RDF at the moment. just syntax
lgonze	do we run into problems integrating with the next gen of  
rdf-oriented tools if we use xml?
yalaforge	I don't know any of those tools or know how they'll look like
ruaok	hard to say -- what are those tools? Do they exist yet?
lgonze	and if we only use the syntax, we get the drawbacks without the  
benefits
lgonze	inference is a Good Thing.
lgonze	finding nearby playlists would be awesome.
yalaforge	and the main drawback here is a more complicated syntax
yalaforge	how would an application do this?
yalaforge	read the osp:playlist attribute and fetch it?
lgonze	yup
ruaok	SWEEEET!
lgonze	it would be fucking fantastic
ruaok	I'm going to be on a panel for Postgres replication at OSCON.
ruaok	I guess I better go write some replication code soon. :-)
lgonze	sucker
lgonze	:)
yalaforge	oh, BTW, I'll have to talk to you about that
ruaok	ok, lets do that after this.
ruaok	if you're still awake.
yalaforge	probably not, let's do it next week ;-)
ruaok	ok, sorry about that tangent.
Russss	haha
lgonze	getting pretty late for you, yalaforge
Russss	we were considering moving to MySQL because postgres replication  
is still so poor
yalaforge	no problem, I've made time for tomorrow. no need to get up  
early
ruaok	apparently that will be a hot topic on OSCON. There were  
apparently so many talks about that that they consolidated them into  
one panel.
lgonze	so it seems like rdf/xml is all the drawbacks of rdf with none  
of the benefits.
ruaok	yup.
yalaforge	so far we agree
lgonze	plus we get burned on the scary vibe of RDF
yalaforge	the acceptence problem is still huge
ruaok	I think if we can focus on using XML as the 'delivery format' of  
the playlist and having the CR tap into the cool technologies then  
we're getting the best of most worlds, I think.
yalaforge	ok
lgonze	hm, cool idea
ruaok	I wonder what danbri will think of this.
yalaforge	he'll hate us, probably
lgonze	I think he'll think his time has been wasted...
yalaforge	would
Russss	are there any advantages to RDF?
yalaforge	Russss: yes, but not if we want to have an XML schema for the  
RDF playlist
yalaforge	that restricts us to some very basic stuff
Russss	true
ruaok	but, danbri is a reasonable guy -- I think he will see our point.  
Its not like we're making flippant decisions.
ruaok	Though we are flipping a bit. :-)
yalaforge	still, the vocabulary (using namespaces) can be used for both  
our XML playlist and for RDF
lgonze	ruaok: yes, I don't like flipping.
yalaforge	I did that a few times in this project
ruaok	agreed, but flipping at this stage is cheap. Wait a few more  
weeks and it gets $$$
lgonze	true.
lgonze	it's cheap for us to bite the bullet, expensive for working code  
to do it.
ruaok	And all of the pains we are going through now are *exactly* the  
same q's I had back in the day when I was hacking on the RDF WS for MB.
yalaforge	anyway, we can remove the rdf:RDF from strawman 9 and we're  
back in business
ruaok	Its not gotten any easier -- even 2-3 years later.
yalaforge	is a bit sad his jena parser is useless
yalaforge	but that'll pass
ruaok	my thoughts keep returning to the CR.
lgonze	that's the same pain that we're trying to eliminate by making  
this decision early, matthias
yalaforge	does anyone of you know some of the XML linking standards?
lgonze	not me
yalaforge	perhaps we need some of that technology here
yalaforge	rob?
ruaok	1. Its not a trivial task. 2. For maximum adoption, what language  
should it be written in?
ruaok	nope.
ruaok	like xlink?
ruaok	or is it xpath?
yalaforge	or XPointer. yes
key88sf	back
yalaforge	nope
ruaok	.
ruaok	key88sf: hey!
yalaforge	key88sf: welcome!
lgonze	hey db
lgonze	wb
yalaforge	did you read the backscroll?
ruaok	We've been bantering the benefits of XML vs RDF>
key88sf	reading....
ruaok	thinks expose' rules.
key88sf	ok
key88sf	so we're agreed to nix RDF then?
ruaok	not quite yet.
yalaforge	but almost
ruaok	no official vote has been taken.
lgonze	I'd like to sleep on it.
lgonze	jesus, talk about an echo chamber
yalaforge	we heard all the arguments pro and contra?
yalaforge	anything else?
lgonze	I think so.
lgonze	the deciding factor for me is your point that we can't do the  
powerful stuff in rdf/xml
lgonze	so we're paying the price without getting the benefit
ruaok	yalaforge: nitpick: in english 'con' is used only.
yalaforge	oh, thanks!
yalaforge	ususally, extensibility is one of the key features of RDF
ruaok	np
yalaforge	with the fixed RDF schema we don't have it
lgonze	alright, well, seems like we know where we're at
yalaforge	what now?
lgonze	next up is to make a new version of the vocab, call it  
tentative, maybe also ask around a little to see if anybody has a good  
reason for us to stick with RDF after all, and move on to code.
ruaok	I think we need to propose a re-vote onthe list.
lgonze	agreed.
ruaok	lgonze: can you do the new version and I'll put out the call for  
a revote?
lgonze	yeah.
yalaforge	I think the vocabulary is pretty good already
ruaok	ok.
lgonze	I like what we have a lot. :)
ruaok	should I wait until lgonze has something that is XMLififed?
lgonze	we could stop here and still kick ASX's ass
ruaok	lol
yalaforge	should I change  
http://playlist.musicbrainz.org/playlist/moin.cgi/ 
DefaultFormatInRdfVersionFour to plain XML?
key88sf	plz :)
ruaok	ok
yalaforge	lucas?
lgonze	the sigh of relief heard around the world. :)
lgonze	sure.
ruaok	I'll start posting.
lgonze	yalaforge: which one of us is doing the rewrite? I don't think  
either of us minds.
ruaok	do we have chatlogs to post about this?
yalaforge	I'm rewriting the wiki page right now
lgonze	k




More information about the Playlist mailing list