Perl 6 - the future is here, just unevenly distributed

IRC log for #metacpan, 2016-05-10

| Channels | #metacpan index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:04 jdv79 ranguard: nice
01:53 metacpan joined #metacpan
01:53 metacpan [cpan-api] oalders pushed 1 new commit to mi/es2: https://git.io/vrfiJ
01:53 metacpan cpan-api/mi/es2 a047d09 Olaf Alders: CPANTesters db returns an empty string rather than 0 for versions.
01:53 metacpan left #metacpan
02:20 metacpan joined #metacpan
02:20 metacpan [cpan-api] oalders pushed 1 new commit to mi/es2: https://git.io/vrfPo
02:20 metacpan cpan-api/mi/es2 a7b0856 Olaf Alders: Strip some characters from versions before looking up dists for CPANTesters data.
02:20 metacpan left #metacpan
02:32 grantm joined #metacpan
02:51 michael joined #metacpan
04:31 punytan_ joined #metacpan
04:52 mickey oalders++
05:09 oiami joined #metacpan
07:29 Relequestual joined #metacpan
07:32 mickey joined #metacpan
07:34 mickey joined #metacpan
07:58 neilb joined #metacpan
08:10 neilb joined #metacpan
08:11 Relequestual joined #metacpan
10:13 jdv79 ranguard: "GET requests with a request body will no longer work"?
10:18 haarg jdv79: what are you wondering?
10:26 jdv79 why?  ES 2.3 supports that and i think a few clients use it.
10:27 jdv79 and if its necessary whats the workaround?  can all that be serializied into the uri?
10:34 ranguard jdv79: https://github.com/elastic/elasticsearch-perl/blob/a40f6fb364005a8b3f21cb9422c05d0ca7769217/lib/Search/Elasticsearch/Transport.pm#L80
10:34 ranguard so use a POST instead of a GET
10:35 ranguard the reason is our CDN follows RFC... which basically says you _can_ send a body on GET, but it should be ignored and not used to alter the response, so they don't support it because otherwise it would cause caching issues
10:42 jdv79 cool.  i've never understood why somethings do get bodies myself.
10:43 jdv79 what are the caching rules?
10:44 ranguard jdv79: currently... there arn't any
10:45 ranguard jdv79: going forward we can set surrogate keys ( dist name + author ), and long surrogate cache (e.g. for the CDN only) and then call a purge on dist+author when we add something to the index
10:45 ranguard so that will work for most things that are not a _search
10:48 jdv79 good luck with that
10:48 jdv79 using a POST as GET is kinda ew though
10:49 jdv79 better if there was a uri serialization scheme such that all gets could get GETs still
10:49 jdv79 s/get/be/
11:22 ranguard jdv79: I complained to ES ages ago about it all, was told they have users/clients that use it so it wasn't likely to change
11:22 ranguard and this was the work around they recommended :(
11:27 jdv79 metacpan could do the adaptation, no?
11:30 ranguard CDN sits in front of us
13:01 dolmen joined #metacpan
14:24 kentnl https://github.com/CPAN-API/metacpan-web/issues/734  # perl is missing on metacpan again
14:44 oalders kentnl: for some reason after the initial re-index, the "latest" flag gets unset
14:44 oalders i'm just re-re-indexing it
14:45 kentnl I might guess it could be related to the dual-upload process. Unsure.
14:45 kentnl ( ie: the related issue where it randomly switches between .bz2 and .tar.gz releases )
14:46 alh joined #metacpan
19:16 jdv79 ranguard: ok, and?
19:19 jdv79 what i'm saying is metacpan can accept the json struct passed in a uri (somehow), deserialize it and issue the ES GET with that body.  the cdn sits on the perimeter, right?  the CPAN-API server and ES are both behind the same cdn, right?
19:26 oalders ES => CPAN-API => Fastly
19:26 oalders there's already an ES DSL for using the query string rather than the body
19:27 oalders it's just kind of ugly
19:29 jdv79 their docs say the ES provided search is hobbled
19:29 oalders yeah, i'm just saying it exists
19:30 jdv79 ok
19:30 jdv79 but using POST for GET.  gross.
19:30 oalders yeah
19:32 Grinnz_ keep in mind there's also a limit to how much you can put in the query string, Datatables API works via POST for this reason
19:33 Grinnz_ because it has so many parameters it's possible to reach that limit
19:33 Grinnz_ it's icky, but it's icky because of HTTP
19:33 oalders yep, there's that as well.
19:35 oalders the easy solution is just to hold your nose and use POST rather than GET
19:35 oalders while shaking your fist at ranguard
19:35 jdv79 except you can't (shouldn't) cache a POST
19:35 jdv79 Grinnz_: where is this limit?  iirc its rather large.
19:36 Grinnz_ 2048 chars
19:36 Grinnz_ for the url as a whole
19:36 Grinnz_ not as hard to reach as it sounds
19:36 jdv79 where is this limit?  i know i've had way larger.
19:36 oalders http://stackoverflow.com/questions/2659952/maximum-length-of-http-get-request
19:37 oalders that answer is old
19:37 Grinnz_ it's not a hard limit AFAICT
19:37 jdv79 that's also a browser limit.  i think the url size is generally ignorable.
19:38 jdv79 in practice i think i saw something like 8 or 16K in various situations.
19:38 Grinnz_ it gets limited by the server before modern browsers, usually
19:38 jdv79 but anyway.  its an option.  i've done it before.
19:54 ranguard jdv79: what the CDN gives us out weighs any technical smell
19:56 ranguard jdv79: also we arn't likely to cache _search results, just the individual document results
20:05 neilb joined #metacpan
20:14 jdv79 ok.  its not a smell.  its wrong.  sure.  sounds great.
20:32 oalders let's just call it a trade-off
20:45 mst jdv79: it's gross. but given the in-practice URL limits, ES chose to do one icky thing with HTTP
20:45 mst and we're going to do a different icky thing with HTTP to compensate
20:46 mst I concur that it's gross, but 'hold your nose and use POST' makes for the least grossness overall IMO
20:47 ether can the magic 'reindex' button be hit for https://metacpan.org/release/DROLSKY/MooseX-Getopt-0.70 please?
20:47 ether (if it automatically refreshes in a day or so, that's fine too)
20:47 * ether just fixed some perms up there.
20:54 haarg using POST for a search like this isn't wrong.  it may not be the "most right" thing to do, but it isn't incorrect.
20:54 mst ether: didn't you have a script for that now?
20:54 mst ranguard: ^^
20:56 ranguard mst: I just pm'ed ether the same thing :)
20:56 mst :D
21:10 ranguard oalders: https://github.com/CPAN-API/cpan-api/wiki/SysAdmin#how-to-reindex-a-missing-module seems wrong ( please update and tell ether when done )
21:20 _dolmen_ joined #metacpan
21:47 oalders we no longer need to change directories before running the carton wrapper, so i've updated the wiki
22:04 melezhik joined #metacpan

| Channels | #metacpan index | Today | | Search | Google Search | Plain-Text | summary