[13:05] <niemeyer> Good mornings
[13:20] <fwereade_> niemeyer, heyhey; nice holiday?
[13:20] <niemeyer> fwereade_: Heya!
[13:21] <niemeyer> fwereade_: Yeah, nice overall.. went to my dad's for a nice barbecue. Ale was working over most of it, though, which removed some of the feeling of easter.
[13:21] <niemeyer> fwereade_: How about you, good stuff there?
[13:22] <fwereade_> niemeyer, aw :(
[13:22] <fwereade_> niemeyer, had a lovely easter actually; went out with cath's family for *awesome* food, and a spot of light walking in the countryside afterwards
[13:23] <niemeyer> fwereade_: Oh, sounds nice indeed
[13:24] <niemeyer> fwereade_: How's the countryside like there?
[13:24] <fwereade_> niemeyer, at this time of year, both rocky and green, once you actually get out there
[13:24] <fwereade_> niemeyer, lots of flowers
[13:24] <fwereade_> niemeyer, over summer it's a bit brown but still really rather lovely
[13:25] <niemeyer> fwereade_: Ah, colorful then
[13:25] <niemeyer> (at this time)
[13:25] <fwereade_> niemeyer, yeah :)
[14:45] <niemeyer> mthaddon: ping
[15:41] <niemeyer> Lunch, biab
[16:12] <hazmat> fwereade_, re error on private cloud with constraints.. http://pastebin.ubuntu.com/922019/
[16:22] <hazmat> ic the problem
[16:37] <hazmat> fwereade_, effectively required constraints aren't populated for private clouds.. http://pastebin.ubuntu.com/922057/
[16:45] <hazmat> az and instance type are still valid to specify even lacking the ability to define the vocabulary, the values should pass through
[18:06] <hazmat> fwereade_, i ended up just putting in bandaid to remove the using_amazon check when setting up ec2 provider constraints in r518.. else private cloud usage using the ec2 api was broken.. as the provider requires both constraint keys defined in several places, and the default selection of instance-type m1.small relies on it as well.
[18:07]  * hazmat moves on to checking out the sub agent branch
[18:15] <fwereade_> hazmat, thanks, and sorry
[18:17] <fwereade_> hazmat, but I would say that surely we should in fact be getting values rather than getiteming them, rather than exposing meaningless values through the UI?
[18:29] <hazmat> fwereade_, well typically the standard values do have meaning in private contexts, just not the ones we think they do.. i looked at doing gets.. but that's also problematic because the default-instance-type is resolved via constraint
[18:29] <hazmat> which fails when the constraint itself defined
[18:30] <hazmat> what a tangled web we weave when first we practice to abstract ;-)
[18:31] <fwereade_> hazmat, I'm confused; default-instance-type short-circuits constraints
[18:31] <hazmat> fwereade_, it does.. but say you don't have it defined
[18:31] <hazmat> fwereade_, then the default value comes from generic constraints
[18:31] <hazmat> but those are only setup upon registering generics
[18:31] <hazmat> ie. instance-types
[18:32] <fwereade_> hazmat, hmmm, yeah, got you; I'd imagined people were always using default-instance-type and default-image together :(
[18:32]  * hazmat grabs some snacks and call its lunch
[18:45] <hazmat> fwereade_, that might be the case for some, but not all.
[18:45] <fwereade_> hazmat, clearly so :(
[18:50] <fwereade_> hazmat, everyone uses default-image-id, though, right?
[18:57] <hazmat> fwereade_, on private clouds yes
[18:58] <fwereade_> hazmat, jolly good
[19:00] <fwereade_> hazmat, when you have a mo, please let me know whether you feel http://paste.ubuntu.com/922279/ is trivial enough to be a trivial
[19:03] <fwereade_> hazmat, tiny tweak to above, http://paste.ubuntu.com/922285/
[19:07] <hazmat> fwereade_, hmm.. i don't see this as particularly better.. the common vocabulary would suggest common usage even if we can't guarantee the value mapping
[19:10] <hazmat> although likely the ec2 zone is wrong given its part in value construction
[19:10] <hazmat> fwereade_, also i noticed that passing instance-type when it wasn't defined didn't trigger a user error
[19:12] <hazmat> fwereade_, ie. given this patch the user wouldn't ever be able to specify via constraint any instance type they'd have to fallback to specifying it via hot swapping in environments.yaml... even though the value they want to give would work if just passed thorough
[20:02] <fwereade_> hazmat, fair points;
[20:02] <fwereade_> hazmat, I shall think on them
[20:03] <fwereade_> hazmat, thanks
[20:05] <hazmat> fwereade_, np.. i think the underlying issues are the same as when this was implemented just that we haven't had time to put in place the right solutions.. i mean  from a different perspective.. --constraints are exposed on private clouds, if usage of them is ignored without warning that's a bad ux. the ideal scenario is of course the same, constraints work everywhere, but that requires both image mapping and instance type mapping definitions, which i thin
[20:05] <hazmat> k we agree is the long term solution.
[20:06] <fwereade_> hazmat, yeah, I think I'm converging on an implementation of that