[10:59:43] <dcausse>	 lunch
[14:14:32] <inflatador>	 <o/
[14:21:57] <dcausse>	 o/
[14:59:32] <dcausse>	 kind of annoying that spark-submit --files /local/path does not populate the files on the driver but only the executors...
[15:56:48] <ebernhardson>	 \o
[15:57:29] <ebernhardson>	 dcausse: hmm, for some reason i thought that worked as long as the driver was running in yarn as well, but maybe it's only --archives
[16:05:29] <dcausse>	 o/
[16:06:53] <dcausse>	 I tried but no, in https://github.com/apache/spark/pull/19102 someone attempted to fix this apparently but this did not get merged
[16:12:29] <ebernhardson>	 :S
[17:31:25] <ebernhardson>	 interesting, sometimes i think wikidata makes more sense to other people than us
[17:33:40] <dcausse>	 yes :)
[17:37:01] <dr0ptp4kt>	 ebernhardson: you wanna start an etherpad or google doc for gathering 'roadmap requests'? i started a doc but then realized probably you'd be coordinating, so figured i should ask
[17:43:32] <ebernhardson>	 dr0ptp4kt: umm, sure we can use an etherpad. I'm just editing my slack message which might not be portable :P
[17:44:10] <dr0ptp4kt>	 ok, i'll share it on the email list. i'm not on that chan
[17:44:13] <ebernhardson>	 https://etherpad.wikimedia.org/p/opensearch-opportunities-for-improvement
[17:44:24] <dr0ptp4kt>	 thx
[18:08:29] * ebernhardson wonders why cirrus-mapping-dump and cirrus-settings-dump are separate apis and not just one thing
[18:08:41] <ebernhardson>	 i don't dare look at git history, it will probably point at me :P
[18:09:05] <dcausse>	 they've been always there for me so no clue :)
[18:09:25] <ebernhardson>	 oh good, it wasn't me. Nik did that in 2014
[18:10:15] <ebernhardson>	 i'm pondering...i was thinking to add `cirrus-build-schema` to return both, but then i could also add a `cirrus-schema-dump` that reports both, i think we can probably just kill the old ones?
[18:10:42] <ebernhardson>	 or there is no pain to leave them and delete next year
[18:11:07] <ebernhardson>	 or it could be `cirrus-schema-dump` for all things, with a build=true flag
[18:13:51] <dcausse>	 sounds good to me
[19:11:23] <ebernhardson>	 trying out claude code for the first time, figure this should be a very mechanical patch, it all exists it just has to be glued together...will see what happens
[19:47:39] <pfischer>	 ebernhardson: exciting! claude needs a bit of tweaking to save tokens. defining contexts helps but also MCP services like serena, context7, and sequentialthinking. Would be interested to hear how it turns out.
[19:49:06] <pfischer>	 ebernhardson: dr0ptp4kt: I forgot to share the search metrics document with the foundation, you should now be able to open it https://docs.google.com/document/d/1WOeJjG9x2QhzfQy8VC42ckP67o70iFOVHts4aYd3Qro/edit?tab=t.0
[19:49:30] <dr0ptp4kt>	 thx ebernhardson 
[19:55:15] <ebernhardson>	 lol, hit rate limit and wnts to wait 3 hours :P
[19:55:24] <ebernhardson>	 (using sonnet, which is their mid-size model)
[19:55:51] <ebernhardson>	 it has an api thats at least facially working though, it was about to start writing phpunit
[20:37:07] <ebernhardson>	 I'm a bit dubious...but on review it seems reasonable and does as i would expect
[20:37:42] <ebernhardson>	 i worry that because i didn't dig into all the bits to write it though, there could simply be things i've forgotten :S 
[20:40:26] <ebernhardson>	 i suppose there is also opportunity for deduplication it missed, and it doesn't have the checks from MappingValidator that specific plugins are required if certain settings are set. 
[22:28:46] <pfischer>	 ebernhardson: was claude allowed/able to run tests a part of implementing your request?