#musicbrainz-devel

murdos: if you want to try that: http://wiki.musicbrainz.org/User:kuno/Search_se...

warp: lol... thanks, but I already know how to setup the search_server (http://bugs.musicbrainz.org/browser/search_serv... ;)

haha, ok

I should've known that.

sorry :)

rookie mistake!

it's just that I've not enough disk space and I don't to bother with running it myself

murdos: I think I have a framework if you want to start trying to write edit types soon :)

I'd suggest you go for the edit types that are /not/ portable to ngs - because the ones that can be ported probably need a bit of extra work

:)

aCiD2: ok :) I've pushed my works branch for review, so I should a bit more time

*have*

yay

these works don't look particularly useful, but I may be misunderstanding what we want to do with works.

my intent is to move the composition credits to work

so you split composition / performance, recordings, .... credits

ah, right. and you can have multiple recordings of a work.

exact

ok, that does make sense.

but a work can still also be for example an entire album, right?

it can... technically a work is really a standalone object in the database

it only has a name, an artist_credit

a type and some ARs

so we just need to define what we want to do with it...

song, (film?, book?), ...

[x] all of the above

:)

ok, time for some dinner. afk!

hmm actually book I can see for audiobooks, although that would a huge mess of recordings attached to one work for all the different narrators.

navap: I would consider for example an opera to be a work.

murdos: I think we will need to add better recording-work integration if we want to move all composition credits

do you have some examples?

I mean that if somebody wants to add composers for a release, they have to add also works

and link them to the recordings

which is not very easy

ok, I see...

also, if you look up ARs from picard, you currently won't get work ARs

moving the credits is probably the right goal

but I'm not sure if it can be realistically done before the release

"before the release"?

sorry, the ngs release

:)

you said a lot of interesting things, but some seem contradictory, so I'm a bit lost

when you're talking about "recording-work integration", you're talking about ui, isn't it?

yes

UI that makes some assumptions about recording-work ARs

ok... and that's why we won't be able to do that before ngs server release

for example, if there is a recording-work AR, fetch the work's ARs

"fetch": to display it, not to move/copy data?

yep

another important thing is that it must be possible to very easily add works linked to existing recordings

ok I understand your concerns

the problem is that once ngs will be released, some works are going to be created by editors

and thus the automatic move will be more difficult

one option is to have a weak UI at the beginning that will be improved with later server releases

another one would be to disable works for the NGS release

I think automatic move wouldn't be that difficult

you just eliminate duplicates

but I'm really not sure what's the best approach

I just feel we are loosing some functionality if we move the data and leave AR support as it is

but what would be the purpose of allowing creation of work if the integration work is not done?

I think collection works is important even on its own

*collecting

right

there are also problems like composer credits on release level

what to do with them?

(but identifying dupes work can be pernickety...)

good question...

oh, I was thinking about creating the works like you did

but not move over the ARs

but the works without their credits would be a bit... empty

but I see, that could be a middle ground solution

maybe copy, or is that too ugly?

create work, link them to recordings. and move composition credits later

copying could be done during initial migration (and reduced with the later move), but maintening new data would be a pain

would you however move URL ARs ('has {lyrics|score] at', ...)?

yeah

that could still use some better integration, but it isn't as critical

so I was thinking about working in that direction (automatically create work like I did) and later decide if composition credits are moved over

but that somehow change my initial goal of initializing works that have something to hold

I mean that we're not going to create work for all recordings

I really don't know what to do :(

me too

ideally I'd probably go for moving the credits

with functionalily like adding a new work if somebody wants to add a composition credit to a recording

and work ARs displayed on recording/release pages

but that's not a small change

murdos: An idea I had last night; a follow on to the 160,000 you get w/ http://codereview.musicbrainz.org/r/629/

luks: how would however solve on long term the "look up ARs from picard" issue?

we can somehow deal with a weak UI (we're already doing that with release-groups now), but Picard seems to be a real issue

we would need to make the WS return work ARs as well

brianfreud: shout!

I'd guess you could likely pick up another few thousand from http://wiki.musicbrainz.org/CSGStandard/JSBach /CSGStandard/Beethoven /CSGStandard/Brahms and /CSGStandard/Mozart titles

which doesn't have to be *that* problematic

which would also eliminate the need for a lot of manual work creation & track-work linking work in 2 of the top 10 biggest artists (JS bach and Mozart)

brianfreud: this is really a different way of initializing works

an identical title match between an entry on one of those 4 lists and any track under that artist should likely be, in of itself, a sufficient 1:n work:track matchup I'd think

I know you're thinking of doing it w/ARs; I figured you could pick up the compositional ARs in reverse here - get the work, then get the composition detail ARs from the tracks you merged in, rather than the other way around

many classical tracks already have ARs

that would result in many duplicates

luks: I kind of assumed those would all be merged together, not one work with 4000 composer ARs to Mozart :D

brianfreud: so, you're saying that for these artists, "recording name" + artist is enough to distinguish a work?

there's no chance that some re-arrangements occured?

murdos: That was the goal of CSG Standard - to present a unified common Work title that could be used.

Theoretically, anyone using CSGv1 correctly would get to the exact same title, but I'd guess that the chances of that (b/c of the confusingness of CSGv1) are really low.

If anyone did actually use the CSG Standard title, and shouldn't have, or if there was a rearrangement, it'd seem a lot easier to fix those manually post-NGS, rather than try to spot them while manually linking every single one of those tracks

aCiD2: in http://codereview.musicbrainz.org/r/540/ how does the artist credit editor fit in with warp's AC?

ruaok: I asked warp, but he said it wasn't ready to merge into master yet

so that review should probably stay open until it gets into master

ok, can you please make a note there? I'm trying to get some of the reviews closed so its easier to see what still needs reviewing

there are a handful of reviews that should be committed, but they're lingering.

really? I was sure I'd commited all of mine that I could

I know warp has a couple that have ship its.

he must be waiting for his shipment of toits

I'm guessing they are the round ones.

ruaok: I'm just putting the last touches to a bit of reorganizing of the edit migration script. I've now got it at one branch which sets the foundations, and a branch for each edit type of there. I've merged back in my "easy" edit types that i've done before

i'll be throwing away ones that try to map pre-NGS to NGS, and we'll use a read only edit type for them.

Hopefully this cleaner layout will allow murdos and luks to hope on board

what do you think about utilising our crazy dutch employee for this too?

he /is/ pretty crazy

utilizing the crazy to do what?

ask him.

i wonder if warp (poke poke) is still up

I don't want you to burn out on it, but not sure how he feels about it.

zzzzzzzzz

him and carols seem to like getting up at the butt crack of dawn.

i'm full steam ahead at the moment - I *really* want to get this pretty much done by the end of next week

because then I'm going to have to start making revision a bit more of a priorty (atm it gets 1/2 hours a day if it's lucky)

i'll try and recruit him tomorrow, I wasn't sure if you wanted him to keep pushing at other things

ok, you two figure that out.

aCiD2: in lib/DBDefs.pm.default

sub MB_SERVER_ROOT { "/home/httpd/musicbrainz/musicbrainz-server" }

djce and I have been installing the codebases for production into /home/mbserver/...

is there any reason you can see for us to keep /home/htttpd?

Can anyone get to my user page in the wiki? http://wiki.musicbrainz.org/User:BrianSchweitzer

I get "Fatal error: Allowed memory size of 33554432 bytes exhausted (tried to allocate 24 bytes) in /usr/share/mediawiki/includes/Title.php on line 639"

haha

That's quite sad :p