#metabrainz

lucifer: that seems like dodgy data in Spotify, but it should still be matched I guess

kellnerd: thanks for the updates, I'm glad things are mostly working! I'm traveling back atm but I've seen your message about the attributes issue, I just might be a bit slow to respond until I get home

will have plenty of time to look into it on the flight though

Have a good flight home!

thank you!

bitmap: the comment under CREATE TABLE artist_release suggests we sort by catnos, but AFAICT we do not? MBS-6796 is still open and find_fast doesn't seem to use them

MBS-6796: Use catalogue numbers in release group sorting https://tickets.metabrainz.org/browse/MBS-6796

Were we planning to but found some issues?

Wait, no, I'm being an idiot, that's _release, we don't store catnos for _release_group

We could consider storing the alphabetical first catno for all releases in the RG, I guess, if we find that's useful, but that'd be a schema change

Marked that ticket schema change at least

Oooh, thanks for the knockout flow

(that sounds like you're a great battle rapper now)

lol

right, we'd need to materialize first_release_catno I guess. dunno how useful that'd be outside of this case

Thanks yvanzo

reosarevok: yeah, could be dogdy data or just how they manage data internally. i think we should probably try to match if possible. fwiw, i did some counts and there are 30k such instances (having ft, feat or featuring in name) out of overall 3.3M artists.

[musicbrainz-server] 14reosarevok merged pull request #2675 (03master…more-flow-strict): Make more files flow strict or strict-local https://github.com/metabrainz/musicbrainz-serve...

[musicbrainz-server] 14reosarevok opened pull request #2678 (03master…autocomplete2-flow-strict): Make Autocomplete2 flow strict https://github.com/metabrainz/musicbrainz-serve...

moooin!

yvanzo: you've left the airbnb, I trust?

Hi outsidecontext ! If I’m not wrong, you’ve been using weblate for translations on the picard android app, right?

alastairp: hi! interesting, https://oauth.net/2.1/ is now in the works and it has some differences from the currently widespread OAuth 2.0.

morning lucifer

morning!

if it's not published, I'd be tempted to just completely ignore it

makes sense, yeah its currenlty draft.

although, I'm looking through the "Major changes" list - we are already planning on using some of those

and some of the things that are removed, we're probably not planning on using anyway

> Refresh tokens for public clients must either be sender-constrained or one-time use

that one is interesting. does that tie into our question about if we want them to expire?

"OAuth 2.1 consolidates the changes published in later specs to simplify the core document."

implicit grant has been removed but we plan on keeping it for now and yes the other difference is refresh tokens.

so it's not so much a "new" version, but more a rewriting of the spec + new features and recommendations

how do you do javascript app auth without implicit grant?

(and is there an existing RFC for the replacement?)

https://oauth.net/2/browser-based-apps/ I guess

"Among other things, it recommends using the Authorization Code flow with the PKCE extension instead of using the Implicit flow."

well, if that works, I don't see much problem with suggesting that we use it

yup agreed.

fwiw django-oauth-toolkit most recent release changes authorization code flows to have pkce to be enabled by default, we've had no problems getting various clients to adapt to it

(we = $dayjob, nothing to do with meb)

tykling: thanks. atj_mb suggested that we make it required, which we're going to do. we're using https://authlib.org/ + their flask library

it's currently optional in musicbrainz.or

right. it protects against some scenarios that are relevant when the auth server and the resource server are not the same, not relevant everywhere but a good thing to get standardized

that's great actually, because this will be the case for us (I believe - resouce server: musicbrainz.org, auth server: metabrainz.org)

oh right of course, well great :D

tykling: it sounds like you might know a bit about oauth ;) do you mind if we run some documentation/plans past you to get your feedback on it?

not at all, though I may take a few days to find time to look at it, my schedule is crazy at the moment

happy to help if I can :)

great, thanks. we finished most of our planning doc and made a start on a test implementation last week, I'll clean up the doc and send it your way

I think I caught monkey's sniffles :(

is that like monkeypox lite

Awww, don't call it that !

:p

gooood moooorniiiing

I'm still astonished that cocktail bot is a legal weapon.

legal or lethal?

lethal, yes, that too.

lucifer: have you started making troi changes yet? if not, I think I will take a stab at those later today, to get into some coding after a week of madness

mayhem: which changes?

click and how to load patches

or should I go check the PR list? heh

ah yes, i did make small changes but not all. i'll open a PR later today, you can complete the rest if you prefer.

ok, I'll wait for the PR then see.

yes getting the spotify cache PRs merged would be nice. currently, the containers are running on custom branches.

also https://github.com/metabrainz/troi-recommendati... is open, the only missing piece for getting the minimal spotify playlist export feature in LB is adding new permission scopes and frontend changes.

ok, I'll start there

that site is surprisingly useless.

it was a little overloaded when it hit HN yesterday. I wonder which APIs they use

certainly a lot easier to do if you have all of the data locally

lucifer: in LB#2198, why are the tables (e.g. spotify_cache.album) not in create_tables.py ?

Add spotify metadata cache: https://github.com/metabrainz/listenbrainz-serv...

oh right, i forgot to add it there 😅

:)

mayhem: about that PR, i am thinking to rename `mapping.spotify_metadata_cache` to `spotify_cache.raw`. thoughts?

raw is a bit short for a table name, I think.

raw_cache_data ?

sure sounds good

also in LB#2198 I see references to couchdb still -- is that still in use in this context?

Add spotify metadata cache: https://github.com/metabrainz/listenbrainz-serv...

nope, i too just saw that and am working on removing those.

k

wondering about https://github.com/metabrainz/listenbrainz-serv...

the cache is more or less stable now, should we revert to spotipy main instead of the fork?

lucifer: what is the purpose of the raw table? why can't we just insert into the normalized tables directly?

if you think that will work, then by all means, less for us to look after

mayhem: we insert directly to main tables and as well raw. raw is only there 1) in case we discover some bug in the migration script that ported the initial 13.5M albums data 2) we want to change the normalized schema.

but we should remove that at some point soonish, yes?

yes makes sense. once we are satisfied with the normalized schema and seen it work in prod for some time, sounds good to drop raw data table then.

personally, if you are already resolving playlists correctly, then I think we can nuke this, honestly.

and the data migration tools too, not need to keep that around, methinks.

*no

so the first prototype of schema didn't have a position column in rel_track_artist so when retrieving the artist name, the names got unordered and there were some mismatches. i added the column in the second prototype to improve matches. i wonder if there are more such bugs lurking which only come up after testing on a wider array of recordings.

ok, if you prefer that

i'll open a ticket to remove the raw data tables so that we don't forget about it.

k

mayhem: oh also, when you have time later, please generate these keys: https://help.apple.com/developer-account/#/devc... . i'd like to setup apple music scraper soon so that we can start building the cache which take awhile to become usuable. meantime can work on adding listening history support.

what type of development do you need this for?

Developer ID Application ?

https://developer.apple.com/documentation/apple... says nothing about this

this is the next step on developer id application.

oh, never mind, I got the wrong initial branch.

ah ok.

and then the instructions still are incomplete.

oh well... 😞

is apple music somewhere down in the list?

no, this media account is the only one listed. now trying to create the id it wants.

I created one, but it didn't show up.