[06:10] <lifeless> wgrant: btw http://wiki.postgresql.org/wiki/Serializable is likely to fuck us over. Excuse the direct language.
[06:10] <lifeless> wgrant: if we don't already drop our serialization level :)
[07:12] <wgrant> lifeless: Yeah, I'd seen that. IIRC most of our stuff is READ COMMITTED.
[07:25] <lifeless> wgrant: btw
[07:25] <lifeless> wgrant: arrays and indexing, what did you learn.
[07:29] <lifeless> bigjools: so tell me about these subscription services
[07:29] <bigjools> lifeless: in a bit, busy
[07:32] <wgrant> lifeless: For bug access checks, array indexes work but aren't really useful (since most bugs are public it's faster to just read them all and filter inline). For tag searches I haven't really looked yet, but I suspect they won't work well unless the planner can use stats on the columns, which AIUI is only implemented for tsvector so far.
[07:35] <lifeless> wgrant: thanks
[07:35] <lifeless> wgrant: I'm thinking subscriptions you see
[07:35] <lifeless> wgrant: +tag and -tag arrays, f'instance
[07:39] <wgrant> lifeless: GIN-indexed string or int array columns should work very well for that. http://www.postgresql.org/docs/9.1/static/intarray.html provides some additional support for integer arrays.
[07:39] <wgrant> But most stuff is doable with GIN natively now.
[07:41] <lifeless> wgrant: is that in trunk or we'd need the addon ?
[07:41] <lifeless> wgrant: in core I mean
[07:42] <wgrant> lifeless: Still contrib, even in 9.1, AFAIK
[07:42] <lifeless> so, we could map tags to ints via a data table
[07:42] <lifeless> and ditto queries
[07:42] <lifeless> and use that
[07:42] <wgrant> That would probably be optimal, yes.
[07:43] <lifeless> we've run contrib before
[07:43] <lifeless> this seems safer than the online defrag contrib module
[07:45] <lifeless> bigjools: would maas be a reasonable project to cargo cult buildout for django w/system packaged django use ?
[07:46] <bigjools> lifeless: I *think* so, flacoste just fixed it to do that properly
[07:46] <lifeless> lp:mass right?
[07:46] <lifeless> bah
[07:46] <lifeless> maas. thingy.
[07:46] <bigjools> I have not tried it myself yet having been mostly on a plane or in a jetlagged blubbery mess
[07:46] <bigjools> maas, yes :)
[07:46] <lifeless> indeed
[07:46] <lifeless> moving + tzs -> nightmare
[07:47] <wgrant> Jetlagged blubbery messes are the best situations to rework critical infrastructure :)
[07:47] <bigjools> wheeee
[07:47] <lifeless> actually there is some research supporting that
[07:47] <wgrant> Forces you to accidentally think outside the box :)
[07:47] <lifeless> drunk works too
[07:47] <wgrant> Because you're delirious.
[07:47] <bigjools> currently trying to book inspections for rentals next week. Lazy Aussie agents don't work saturday afternoons it seems
[07:47] <lifeless> http://www.wired.com/wiredscience/2012/02/why-being-sleepy-and-drunk-are-great-for-creativity/
[07:48] <lifeless> bigjools: ^
[07:48] <bigjools> wow, I should be one of the most creative people in the world then
[07:48] <lifeless> right now, yeah, you probably would be
[07:48] <bigjools> or does it require simultaneous sleepy and drunk?
[07:48] <lifeless> nah
[07:48] <bigjools> :)
[07:49] <lifeless> basically, get the supervisor module turned off -> more options considered when thinking about something
[07:57] <lifeless> http://tools.ietf.org/id/draft-briscoe-tsvwg-re-ecn-tcp-09.txt is brilliant
[07:59] <lifeless> last paragraph of section 5.1
[08:00] <bigjools> !
[08:07] <lifeless> indeedy
[08:37] <nigelb> woah, bigjools, you moved?
[08:39] <nigelb> I like how LP is now having more people in the East.