#musicbrainz-devel

were we hoping to do that for /ws/2?

well, it's going to take centuries for clients to start using them, so we should implement them in some fashion ASAP so clients who want to can already start :)

heh, yeah

even if they're mostly dummy values on our end at first.

I guess for /ws/2 we'd have to have an upper limit of 1/s (i.e., the headers can only give you better than that)

yeah

anyway, point stands that we should discuss! :)

I'll create a ticket if we don't have one yet.

looks like no, everything on jira that mentions rate limiting is pretty much stuff for 3scale :)

* mwtuea: Add back "has no relationships" notice

* mwtuea: refactor relationship grouping code a bit

ianmcorvidae: http://tickets.musicbrainz.org/browse/MBS-5874

ianmcorvidae, I just need to change that to a detach_404, right?

Leftmost: not sure -- it may just be 'not_found'

I see a bunch of /error_404s in the code.

you should make sure I'm right first, too, but :P

It is a privacy leak, I think, but it's one that's copied over from an existing one in the code.

hm, I think $c->detach('/error_404') would do it, yes

yeah

I'll file a separate ticket for the other one and correct this one.

anyway, try that and make sure it returns the same page for a nonexistent versus a private collection, I guess, that should be all

It doesn't, because non-existant collections break. Whoops.

And existant is a non-existent word.

Is codereview.mb.o having some issues?

Or is it me?

seems fine here.

Bits of diffs keep wanting to not load, then it doesn't want to respond at all.

ianmcorvidae, all fixed up.

* mwtuea: MBS-5591: pressing enter with <select> open submits dialog

* mwtuea: warpify var statements

¬_¬ warp

mwtuea: MBS-5836: validate entity types against link types in relationship editor

Project musicbrainz-server_beta build #324: SUCCESS in 29 min: http://ci.musicbrainz.org/job/musicbrainz-serve...

* Ian McEwen: MBS-4947: generatesql update

Now grooving to: ZZ Top ~ Rough Boy (4:52) (from Greatest Hits) (1986)

erh wromng srry

It's my fault, isn't it? :-P

no, code review just has some weird quirks which I never seem to get used to.

It's okay, you can blame that on me.

for example: http://codereview.musicbrainz.org/r/2526/

why does Oliver get one big block with all his comments in it, and I get three separate blocks. it does not make sense.

luks: did you create libdiscid milestones manually or did launchpad pick that information up from my settings automatically (after some time)?

I mean I set everything up and was hoping it would catch that information somehow, but I am suprised why that happens now and not 1-2 hours after I added that information.

JonnyJD: sorry, I don't remember

luks: No, I mean like 13 hours ago.

what are you using launchpad for anyway?

luks: libdiscid went from no releases/tags/milestones to all releases 13 hours ago.

Just reminding ubuntu packaging guys when there are new releases

ah

I'm not sure that is of any help

the package is primarily maintained in debian, so will be just copied from there

would be nice when they just catch downloads and tags automatically and it certainly looks like it now

yes, no clue what the actual synchronizing workflow is. In debian 0.3.0 is in experimental and it looks like they have a new maintainer now

and having an outdated project in launchpad doesn't help. Deleting projects in launchpad is impossible as far as i see

luks: anyways. Thanks for the information. I hope that launchpad project is working "by itself" now.

ocharles: how/where do I review NES ?

please can anyone direct me to documentation for running musicbrainz with something other than the lightweight development server described in the INSTALL.md. I understand how nginx and fastCGI work, but never done it with perl before. Thanks.

mhanney: see admin/nginx/001-musicbrainz

mhanney: and we run the server itself using daemontools, the daemontools run file contains something like this:

mhanney: plackup -Ilib -s FCGI -E deployment --port 55901 --nproc 100 -pid fcgi.pid --keep-stderr=1

I don't think our daemontools run scripts are in a git repository, they probably should be.

thank you derr84, awesome

derr84, looks like a daemontools run file is in repo here admin/nginx/service/run

mhanney: oh indeed :)

Thank you warp, sorry for the case of mistaken identity, Your help is very much appreciated.

kepstin-work: ping

hi

I'm working on making a LATEST files or something for the change-mbids feed.

what should the file contain?

a number?

a filename?

full path?

full path isn't quite future proof if it gets mirrored.

either the number or the filename would be best, I think.

I'm leaning towards filename.

since the application has to parse out the number anyways to sequence the file downloads, I think I'd prefer just the number. But either works.

ok, I'll do that.

is having this be a http service off the plan atm?

this latest stuff is much simpler if it's just a 3xx redirect

ocharles: the benefit of the current solution is that it's trivial to mirror without special server support

sure, but at the cost of users

and given we have none of them right now... we don't exactly need to worry about scaling :)

well, the point of the latest file is that an application wants to find out "what is the sequence number of the most recent update packet?"

just reading that from the body of a request is probably easier than parsing the number from the url returned in a redirect

ocharles: I'm hoping someone will adopt this code.

you wouldn't parse from the URL, you'd fetch the latest which is a json file in its own right, with a sequence number inside it

ruaok: cool

uff, still no robots.txt to let me crawl

ha. foiled.

want one?

what should it say?

permit everything?

for my needs, yes :)

i ended up scripting `seq` and `curl` to do the same

that would have to be at the top level of http://ftp.musicbrainz.org tho, which is hosted by osusl, right?

kepstin-work: i'm downloading from http://changed-mbids.musicbrainz.org

huh, is that gonna be the final location for them, or is this still subject to change?

that isn't the final location.

its mapped into ftp

most people should be fetching from there, but ocharles should probably fetch from ch.mb.org

whats-changed or whats-new is quite a fun name

ftr, a 2 week catchup (downloading all those json files) takes around 6 minutes

while that's quite a big delay, i'm not so sure that's extreme

that's what it is

i imagine Keep-Alive will do you some good, but you're gonna roundtrip on every json packet