#musicbrainz

Any idea why a guess case for sentence keeps all caps?

ok, figured it out, had the "Keep uppercase words uppercased" box checked.

So, in the modern digitial age....... do we really care if CD is also available as download if its the exact same thing?

Hey!

guys how I must act, http://musicbrainz.org/show/edit/?editid=14401545 my edit, later i found cover with barcode. I must canceled my edit and create new one with barcode or how?

another words if edit is not applied yes, but new information founded, how I must act?

*yet

generally yes, you can only cancel the edit and make a new one

but I can approve this one if you want

I've said yes but I really not sure with release date I am googled a lot, on official site and on all wikis and fan sites and even cd cover date is 2002, but if i search for barcode date only 21 January 2003

It's really promblem to me, beacuse it's not only one album with same situation, I want to have in my player date 2002, but add this year to musicbrainz i think mistake

oo luks your approve

is the date on the cover the copyright date?

if so, that's not necessarily the release date...

nikki: I correct understand if in cover I watch "(c)year", I must add this year to MB even, I can't find sources on net?

no

see http://wiki.musicbrainz.org/Release_Date#What_a...

the first section

thanks, but with this rule in MB years often differs with huge other sources like official sites, wikis, fan pages, etc

hmm...

http://replay.web.archive.org/20021206225748/ht... says 2003-01-21

I approved the edit because all the links mentioned 2002

dubwai: should I change it to 2003?

yes but with barcode: http://i.imgur.com/UdFzk.jpg

January 21, 2003.

thanks luks

ok, done, sorry for the confusion caused by approving the edit :(

nikki realy thank for you link archive.org. I am stupid try only Young Tree Recodrs site witch doesn't work. Now I found all original release dates :-)

cool :)

for Yong Tree releases of course :-)

I think MB must support "copiright date"

because many people use MB as database for their tags and for example when they get tags for Hebron Gate and date 2003 they may disapointed because on the net date in 90% is 2002

even offisial site have date 2002, but with news that worldwide release was in 2003

I want to enter a classical album but the track names are all in german - how do I twist them around to match the ClassicalStyleGuide?

Do I just try to follow the style guide except with german words?

or should I just enter the tracknames as is?

never mind, there do appear to be accomodations for that in the style guide...

Hi! How do I add tracks with multiple artist? I'm importing a CD from FreeDB

ok, found it in the docs

any german speakers around?

bah. not even mineo

Ubuntu is upgrading, wish me luck...

Isn't this the upgrade that makes it be based in some not-GNOME thing?

I wonder how big of a change that is

the old GNOME UI is still there

but I wouldn't upgrade that close after release anyway :)

at first it was like "uh oh... too close after the release...", now it's "uh oh... too long ago"

I'd really like to upgrade this laptop, but I'm afraid it will break everything :(

the last release was not that bad though

I told simon the last time he upgraded that it would break, and it did

luks, I see you have been working on an open source finger printing technology... think its going to replace musicip?

that's my no-so-secret goal, yes, but it has a very long way to go :)

Not to get too techie, but I assume a fingerprinting technolgy produces the same fingerprint accross wav, flac, mp3?

similar fingerprint, not necessarily the same

they are always compared to similarity and if it passes certain threshold, it considers the two fingerprints be the same track

s/to/for/

So, I guess depending on the threshold set, determines the number of muliple fingerprints per song.

correct

Any ballpark estimate of how many multiple fingerprints you are getting for the exact same song percentagewise? Your or even if you know the current technology in use on MB.

Snartal: that's hard to say, the thresholds in my implementation are not final, I want to define song as MusicBrainz recording, so I'll wait until most recordings are properly merged in MB NGS before evalulating the threshold choice

but, currently the top track in my database has 30 fingerprints

on average I have 1.16 fingerprint per track

but the database is not very big, so it doesn't say much

and I have no idea about MusicDNS, the technology is a blackbox from our point of view

oh, I was wrong, the top track has 58 fingerprints :)

most of them are useless though, it's just a matter of how you store the data

I decided to store rather more information than less

Thx, I'm alway curuious.

I know I found one track in MB with 20 PUIDs

fingerpriting technologies work very differently

Just pondering how one song can get 20 PUIDs

btw, PUIDs correspond to "tracks" in my database, not "fingerprints"

MusicDNS prefers false negatives to false positives

popular songs are usually distributed cut off, etc.

and the song length matters when looking up a PUID

Even if the song varies by a second or two?

I think it's better to have songs with 20 PUIDs than a single PUID that refers to multiple songs

I think the threshold is somewhere around 10 seconds

I would agree with a single PUID to muliple songs, as that is a false positive.

Though 20 PUID for the same song off the same album presents issues also.

why?

note that it's not necessarily from the same album, people submit PUIDs for very different tracks

that's why you get things like http://musicbrainz.org/show/puid/?puid=7223ffa5...

it's not because the fingerpriting technology doesn't know the difference between the tracks

it's because people submit garbage to the database

Well, there is one of the issues for 20 PUID for the same album/track/release. Can't sort out the garbage.

I think that the only way to sort it out is to delete anything questionable

if it was a mistake, it will get submitted again

Assuming a fingerprinting tech is highly accurate..... there shouldn't be muliple PUID per song.

that would be ideal, but it's not realistic

song/album/release..... I should say.

Not realistic from the point of fingerprinting technology, or from the point of users submissions?

technology

if you aim for one ID per song, you will necessarily get false positives

but it depends heavily on what you define "song" to be

So, in a nutshell, fingerprint technology is good, but a high level of accuracy, say 99.99% is just not there yet.

I don't think it will every be there

things get very fuzzy in corner cases, sometimes you really just can't tell what is a different song and what isn't

Yea, no matter what, you always will have Alpha and Beta risk/error.

With the large percentage of error coming from humans - lol

Thanks for the convo, it was intereting.

interesting.

Do folks think it should be a bit more clear that an artist legal name should not be an alias? It is not mentioned in this doc.... http://musicbrainz.org/doc/Artist_Alias

Snartal: it is, under "When NOT to use aliases"

I guess I'm not seeing it there. At least is doesn't appear as clear as in http://musicbrainz.org/doc/Same_Artist_With_Dif... .... where it talks about legal name, and specifically mentions ArtistAnnoation.