Open Grid Public Beta/Daily Huddle Transcripts/Transcript 20080818

[15:03] Tao Takashi: Hi
[15:03] Connected undefined:
[15:04] Whump Linden: Hey Tao
[15:04] Infinity Linden: okay... nice for effect 'til you try to move with 'em on
[15:04] Whump Linden: Hey Saij
[15:07] Infinity Linden: hmm.. this should be renamed "usual suspects meeting"
[15:07] Whump Linden: shall we get started?
[15:07] BlueWall Slade: Hi
[15:07] Saijanai Kuhn: Im late in spamming...
[15:07] Tao Takashi: looks like we could also have done this in the IRC channel ;-)
[15:08] Saijanai Kuhn: forgot to spam this morning for pyogp and no-one showed. People were pissed. They use me as alarm clock
[15:08] Tao Takashi: ah, some more people
[15:08] Infinity Linden: lol
[15:08] Victor Hua: Took me a moment. Was in the middle of things
[15:08] Tao Takashi: they should learn to read the clock themselves ;-)
[15:08] Whump Linden: Hey Teravus, Tylor, BlueWall
[15:08] BlueWall Slade: hello
[15:08] Teravus Ousley: Hello :D
[15:08] Infinity Linden: yes... everyone should reset their clocks to use SLT
[15:08] Flight Band: All Go
[15:08] Infinity Linden: of course... it's easy for me to say that
[15:08] BlueWall Slade: yeah
[15:09] Whump Linden: let's get started
[15:09] Infinity Linden: since SLT == Pt
[15:09] Whump Linden: https://jira.secondlife.com/browse/SVC-2632 is the current list of issues we're tracking
[15:09] Teravus Ousley: :D indeed. Though this meeting time works for me because it's at 6:00PM EST
[15:09] Whump Linden: we're at 14 open issues
[15:10] Infinity Linden: yay
[15:10] Whump Linden: some of those are duped from the OpenSim forge and mantis trackers
[15:10] jenni Briggs: here ?
[15:10] Whump Linden: so this afternoon we've been talking about https://jira.secondlife.com/browse/SVC-2735
[15:11] Whump Linden: which is related to OpenGrid regions continuing to send UDP packets after the agent has teleported away to another region
[15:12] BlueWall Slade: but, here it kills out client session
[15:12] Infinity Linden: hmm.. bad form
[15:12] Tao Takashi: reminds me of OpenSim raising an exception of a user with the same name but a different UUID comes around
[15:12] Whump Linden: and there are two related OpenSim mantis issues: [1] and 510.
[15:12] Tao Takashi: need to make this a bugreport with some example script
[15:13] Whump Linden: and I've commented on 1054 and will go to #opensim-dev after the meeting.
[15:14] Whump Linden: but if there's an OpenSim dev here who'd like to have a look, that would be great.
[15:14] Whump Linden: and yes, I'm being a broken record on this one :)
[15:14] Teravus Ousley: :D one of the things that I want to do in the next week or so (maybe this weekend) is make the OGP patch work harmoniously with OpenSimulator.. as opposed to it being a disharmonious swap.
[15:15] Whump Linden: Teravus, with respect to the conventional login?
[15:16] Teravus Ousley: nod nods. Ideally, it'll be a module that is loaded with a configuration option, that uses the regular routines to do stuff like login/logout.. etc.
[15:16] Tao Takashi: and I would like to announce our Agent Domain called syntronik.de which you can find at [2]
[15:16] Whump Linden: yes, that's what we talked about with the opensim folks a couple of weeks ago
[15:16] Tao Takashi: source will be released soon as well
[15:17] Saijanai Kuhn: "our"...
[15:17] Saijanai Kuhn: worked so hard on it
[15:17] Tao Takashi: it's based on pyogp and you also helped with background info
[15:17] Tao Takashi: actually quite a lot
[15:17] Teravus Ousley: I do understand the differences in the regionhandle that opensimulator uses normally.. that, I think will be the biggest obsticle.. Ours are X * Regionsize, OGP is just X
[15:17] Saijanai Kuhn: ok ;-)
[15:18] Teravus Ousley: :D
[15:18] Gareth Ellison: looks like that landmark was for other the other side of the river ;)
[15:19] Whump Linden: hands Gareth a towel.
[15:19] Tao Takashi: glad you made it, Gareth :)
[15:19] Teravus Ousley: :D
[15:19] Gareth Ellison: so am i
[15:19] Whump Linden: that was a quick cut and paste from the map
[15:20] Whump Linden: Infinity has a topic
[15:20] Infinity Linden: oh
[15:20] Infinity Linden: uh
[15:20] Infinity Linden: yeah
[15:20] Whump Linden: would hand floor to Infinity, but she's standing on it.
[15:20] Gareth Ellison: out with it!
[15:20] BlueWall Slade: hehe
[15:20] Infinity Linden: just wanted to bring up the issue of code being insync with spec
[15:20] Infinity Linden: and so forth
[15:21] Saijanai Kuhn: watches Gareth prick up her ear
[15:21] Infinity Linden: a new rev of the spec is forthcoming
[15:21] Infinity Linden: and wanted to get concerns from the community about how we progress
[15:21] Infinity Linden: we're eager to get code that implements the spec deployed
[15:22] Infinity Linden: and eager to turn off the non-compliant aspects that are turned on right now
[15:22] Infinity Linden: but
[15:22] Gareth Ellison: may i suggest that we spec the production system?
[15:22] Gareth Ellison: i.e change the spec, rather than the code
[15:22] Infinity Linden: yes
[15:22] Infinity Linden: that's what we're talking about
[15:22] Infinity Linden: but
[15:22] Infinity Linden: we want to make sure everyone has
[15:22] Gareth Ellison: what was the thinking behind the current implementation? since whatever it was, it seems like it should be applied back to the specs document
[15:22] Saijanai Kuhn: unless the code is flawd, of course
[15:22] Infinity Linden: time to implement the changes
[15:22] Tao Takashi: well, I would like to get rid of things like that redirect
[15:23] Infinity Linden: the current implementation was not specced
[15:23] Infinity Linden: not publically
[15:23] Infinity Linden: publicly?
[15:23] Gareth Ellison: yeah, the redirect is quite odd, i do wonder why the caps field in the first login response is empty
[15:23] Infinity Linden: yup
[15:23] Infinity Linden: it's stuff like that that needs to change
[15:23] Tao Takashi: I would say go ahead and make it clean :-)
[15:23] Infinity Linden: okay...
[15:24] Tao Takashi: At some point though we might need some way to discover which version is running
[15:24] Infinity Linden: yeah
[15:24] BlueWall Slade: do you have a url?
[15:24] Teravus Ousley: versioning ++ :D
[15:24] Saijanai Kuhn: Dont' just toss out the spec in favor of hte code unless the code makes more sense
[15:24] Infinity Linden: there's a debate internally about that
[15:24] Tao Takashi: and I also would like to propose that we discuss the spec before you implement it
[15:24] Infinity Linden: we don't want to support every revision of every resource we've ever published
[15:24] Infinity Linden: but
[15:24] Tao Takashi: well, at some point you might have a bigger install base
[15:25] Gareth Ellison: the way to support backwards compatibility is to not remove fields too often
[15:25] Gareth Ellison: and widely publicise when you do
[15:25] Tao Takashi: right now maybe it doesn't matter but I would think that people already will run stuff on it
[15:25] Infinity Linden: hmm.... Legacy Life?
[15:25] Gareth Ellison: adding new fields shouldn't break stuff
[15:25] Teravus Ousley: hehe
[15:25] Tylor Ferraris: jeratric life
[15:25] Tao Takashi: well, you don't need to support every revision done maybe
[15:25] Infinity Linden: right now what we tend to do is change the SL channel
[15:25] Tao Takashi: but maybe the one before the most recent one
[15:25] Infinity Linden: as a way to alert the client that it needs to upgrade
[15:26] Tao Takashi: and maybe there is a way to mark things as deprecated
[15:26] Gareth Ellison: not every revision, but you do want at least a few back
[15:26] Tao Takashi: and let components complain
[15:26] Tao Takashi: I am not sure if it's just the client which needs upgrading then, it might be as well region and agent domains
[15:26] Tao Takashi: and proxies
[15:27] Teravus Ousley: agreed. complaining is important.. like adding [Obsolite{"Message")
[15:27] Gareth Ellison: Tao: ;)
[15:27] Gareth Ellison: a standard obsolete field in each response
[15:27] Tao Takashi: We maybe can spec out some deprecation protocol ;-)
[15:27] Infinity Linden: yow. that's a lot of work for... what?
[15:27] Gareth Ellison: obsolete['use foo instead of bar']
[15:27] Gareth Ellison: not a lot of work
[15:27] Gareth Ellison: 1 little field
[15:27] Infinity Linden: yes
[15:27] Infinity Linden: it is a lot of work
[15:28] Tao Takashi: LL is a big thing, nothing is just a small change ;-)
[15:28] Infinity Linden: because then you have to worry about clients that implement rev 1.14 functionality in resource A, 1.13 functionality in resource B, 1.21 functionality in resource C
[15:28] Gareth Ellison: praise be to the process!
[15:28] Infinity Linden: and you get the number of supported configuratinos == #number of features !
[15:29] Tao Takashi: you can maybe package all these components to one open grid protocol revision together
[15:29] Infinity Linden: which sux
[15:29] Gareth Ellison: infinity - you don't need to worry about legacy clients other than being polite
[15:29] Tao Takashi: we can discuss later how we can work around individual components
[15:29] Gareth Ellison: i.e adding an obsolete field in something if it's not really current, but is supported
[15:29] Tao Takashi: first they need to be there I think
[15:29] Infinity Linden: really? we haev messages that send an obsolete field?
[15:30] Tao Takashi: and right now I would assume that it's also just the LL client so everything should be pretty much in sync at that point
[15:30] Gareth Ellison: you don't have them now, but you could add them
[15:30] Infinity Linden: okay... i'm sure there's gonna be more debate about this at AWG and Zero's office hours
[15:30] Whump Linden: do we need to continue the discussion at the next office hour?
[15:31] Infinity Linden: yeah...
[15:31] Tao Takashi: all this might be part of service discovery
[15:31] Whump Linden: okay, then I'll mark it as the topic
[15:31] Infinity Linden: i suspect this is a meta-issue that will be discussed ad naseum
[15:31] Teravus Ousley: thinks that some sort of method of denoting obsolite for one revision of the protocol would be helpful for people implementing the protocol.
[15:31] Whump Linden: What Would HTTP Do?
[15:31] Infinity Linden: fail miserably
[15:32] Whump Linden: then that's an option as well
[15:32] Teravus Ousley: haha.
[15:32] Tao Takashi: in the social networking world it would probably tried to be solved with XRDS
[15:32] Whump Linden: o.O
[15:32] Tao Takashi: some services of the same type but different versions with different priorities
[15:32] Tao Takashi: if it's not there at all it fails of course ;-)
[15:32] Infinity Linden: okay... as long as no one mentions WSDL
[15:32] Whump Linden: you just did
[15:32] Teravus Ousley: >.>
[15:32] Teravus Ousley: <.<
[15:32] Tylor Ferraris: oh nos
[15:32] Infinity Linden: we can probably think about a very, very basic service discovery protocol
[15:32] Infinity Linden: but
[15:33] Saijanai Kuhn: knights who say ni cannot here the word wsdl
[15:33] Tao Takashi: and this looked like a happy huddle before you mentioned it...
[15:33] Gareth Ellison: hey people - i have a service discovery protocol
[15:33] Whump Linden: notices the moon rising like the WS-deathstar.
[15:33] Gareth Ellison: [3]
[15:33] Tao Takashi: just use XRDS-Simple, it's easy enough ;-)
[15:33] Infinity Linden: okay... topic 2... shrubbery
[15:33] Gareth Ellison: works fine
[15:33] Infinity Linden: um
[15:33] Infinity Linden: i'm not tying the grid into OpenID
[15:34] Tao Takashi: XRDS-Simple is not OpenID
[15:34] Infinity Linden: openID is broken in very fundamental ways
[15:34] Tao Takashi: I know and nevertheless it will be used on the grid ;-)
[15:34] Infinity Linden: or rather...
[15:34] Infinity Linden: OpenID is a solution for a problem that doesn't cover all of our requirements
[15:34] Tao Takashi: but anyway, XRDS-Simple is not OpenID although they use XRDS
[15:34] Infinity Linden: sure
[15:35] Tao Takashi: and I would prefer this instead of inventing our own protocol
[15:35] Infinity Linden: OpenID will be used... and then we'll use other stuff to make the rest of the grid work
[15:35] Gareth Ellison: what's wrong with a nice simple XML file listing what grids exist at domain.com?
[15:35] Infinity Linden: What.. kerberos?
[15:35] Infinity Linden: saml>
[15:35] Gareth Ellison: and all the URLs for login etc
[15:35] Infinity Linden: saml?
[15:35] Tao Takashi: XRDS-Simple is a quite simple XML file
[15:36] Infinity Linden: what is the problem that XRDS solves?
[15:36] Tao Takashi: I also wrote a python lib for it ;-)
[15:36] BlueWall Slade: a universal version of the slurl would be nice in a client
[15:36] Tao Takashi: listing services with their endpoints and priorities
[15:36] Gareth Ellison: region@domain.tld:gridname
[15:36] Whump Linden: hold on, we've drifted from indicating what verson of the spec we're running to service discovery
[15:36] Infinity Linden: we know our endpoints
[15:36] Gareth Ellison: we're at Ambleside@secondlife.com:agni right now
[15:36] Tao Takashi: but not which are there and in which versions
[15:36] BlueWall Slade: then a grid/region domain name system to map names to identifying UUID's
[15:36] Gareth Ellison: works beautifully and way nicer than HTTP region URLs
[15:37] Gareth Ellison: identifying UUIDs is a big one
[15:37] Whump Linden: and we're running over time
[15:37] Infinity Linden: yup
[15:37] Infinity Linden: this one's not going to get solved
[15:37] Infinity Linden: right away
[15:37] Infinity Linden: and
[15:37] Tao Takashi: shuts up and most likely codes some demo someday :)
[15:37] Xugu Madison: Why not secondlife.com/agni/Ambleside ?
[15:37] Infinity Linden: we should probably discuss it at AWG and Zero's office hours
[15:37] Infinity Linden: but
[15:37] Gareth Ellison: Xugu - that works too
[15:37] Gareth Ellison: i just happen to have production code using region@domain.tld:grid
[15:37] Infinity Linden: for the near term
[15:38] Xugu Madison: sorry, I like keeping my separators simple is all :)
[15:38] BlueWall Slade: heeh, trust has taken over the HO's
[15:38] Whump Linden: and has the advantage of existing libs for consuming URLs
[15:38] Tao Takashi: regiondomain.com/regions/<regionname>/<position> ?
[15:38] Xugu Madison: Tao, sounds good to me
[15:38] Victor Hua: secondlife.com/agni/Ambleside?x=123 ect...
[15:38] Infinity Linden: i think we're going to announce when aditi is moving from draft 2 to draft 3 of the spec
[15:38] Infinity Linden: (or draft 1 to draft 3)
[15:39] BlueWall Slade: ogp://...../.../.../..
[15:39] Xugu Madison: Victor, I'd say secondlife.com/agni/Ambleside/223,250,25
[15:39] Tao Takashi: as said, it would be great if we can at least shortly see draft 3 before it's implemented :)
[15:39] Teravus Ousley: is happier with Tao's version.. because it adds a space for determining the type of resource you're looking for.. '/regions/'
[15:39] Infinity Linden: Tao... right!
[15:39] Whump Linden: okay, Wednesday at Office Hours, we'll continue discussing how to indicate what version of the OGP protocol you expect, and time permitting the SLURL question
[15:39] Infinity Linden: that's definitely the plan
[15:39] Xugu Madison: Actually, I take that back, Victor's ideas's more correct
[15:39] Infinity Linden: and the idea is to finish the internal review
[15:39] Infinity Linden: release it for external review
[15:40] Infinity Linden: give folk some time to implement
[15:40] Infinity Linden: and then at some magical moment
[15:40] Tao Takashi: maybe make some roadmap of when the switch will happen
[15:40] Infinity Linden: change over from being draft 1 to being draft 3
[15:40] Tao Takashi: so we can adjust accordingly
[15:40] Infinity Linden: yup
[15:40] Teravus Ousley: :D
[15:40] Gareth Ellison: question - how do we implement it if aditi's not migrated?
[15:40] Gareth Ellison: or do we all go off and code our own agent domains?
[15:40] Tao Takashi: I will also post more about how to get that agent domain running maybe so we can also test this without an LL implementation either with this or with Gareth's proxy
[15:40] Infinity Linden: umm... run your unit tests?
[15:41] Whump Linden: there's test suites
[15:41] Gareth Ellison: tao - grab me later and rant about relicensing that proxy
[15:41] Whump Linden: probably some more wiki work for me
[15:41] Teravus Ousley: during these initial versions.. I think we're going to have to 'trial and error' a bit to get it right.. like Zha
[15:41] Tao Takashi: I might also do a version which does not require Plone but a more lightweight solution
[15:41] Tao Takashi: for testing you probabyl do not really need a CMS in the background
[15:42] Tao Takashi: Gareth: later is not today anymore though.. and I am maybe not the right person to discuss licensing anyway ;-)
[15:42] Tao Takashi: actually for me it's not today anymore anyway
[15:43] Infinity Linden: okay.. i'm gonna run off and hammer on the docs some more
[15:43] Infinity Linden: so we have something to review soon
[15:43] Tao Takashi: cool, Infinity :)
[15:43] Teravus Ousley: thank you :D
[15:43] Whump Linden: okay, any other questions, bugs
[15:43] Tao Takashi: now that syntronik.de is out the door I also might concentrate on some other stuff ;-)
[15:43] Tao Takashi: like teleporting..
[15:44] Teravus Ousley: I don't have any currently.. I'm going to be running and tracing the OGP lines in opensim later..
[15:44] Teravus Ousley: probably after that.. I'll find something :D
[15:45] Tao Takashi: I more likely check on bugs after the reimplementation
[15:45] BlueWall Slade: i will visit a few OGP regions to check teh day cycle things, since it is night now
[15:45] Tao Takashi: I guess it makes more sense to file them after
[15:45] Tao Takashi: esp. those which basically say that the implementation is not according to the spec
[15:46] Whump Linden: cool, Leyla's back and I can relay info to her
[15:46] Whump Linden: I'm going to lobby for the UDP/Ghost bug in #opensim-dev
[15:47] jenni Briggs: mine goto junk
[15:47] Teravus Ousley: my guess.. is the close routine is expecting something to exist.. that doesn't. The close routine is actually quite complicated..
[15:47] jenni Briggs: yep
[15:47] Gareth Ellison: Teravus - Thread.Abort()
[15:48] Gareth Ellison: what's complex?
[15:48] Whump Linden: Teravus, if you're interested, I can point you to logs, etc.
[15:48] Teravus Ousley: haha. 500MB ram in dead clients :D
[15:48] Gareth Ellison: ping times out, kill the thread
[15:48] Teravus Ousley: thanks. :D
[15:49] Teravus Ousley: I suppose this also fell on a time when melanie was re-archtecting the way we're doing the protocol.
[15:49] Gareth Ellison: if (m_pendingPingResponses>2) Close(true);
[15:49] Tao Takashi: so here is that blog post: [4]
[15:51] Gareth Ellison: teravus - what's so complex about killing a thread?
[15:51] Whump Linden: Gareth: one of the symptoms, however, is if you kill the simulator, you get logged out even if you've left for another region.
[15:51] Teravus Ousley: there's cleanup associated with it..
[15:51] Gareth Ellison: yeah, so do the cleanup after ping timeout
[15:51] Whump Linden: don't want to have death by backhoe.
[15:53] Teravus Ousley: the process actually starts in scene(in the context of the client thread), goes to the client object, passes to the UDP server, then passes back to the scene, then back to the client object before it calls thread.abort
[15:53] BlueWall Slade: that part of it doesn't happen in opensim proper though
[15:53] BlueWall Slade: we do that all the time tro restart our regions on OSGrid
[15:54] Gareth Ellison: [5]
[15:54] Gareth Ellison: you mean like this function, which just doesn't get invoked at all on ping timeout for some reason?
[15:55] Teravus Ousley: yep, the shutdownCircuit is the flag that it's been properly shut down
[15:56] Teravus Ousley: ... well, it did.. before the rearchitecting of the client stack..
[15:56] Gareth Ellison: the shutdownCircuit flag is set here:
[15:56] Gareth Ellison: if (m_pendingPingResponses>2) Close(true);
[15:56] Gareth Ellison: it ain't rocket science
[15:57] Teravus Ousley: well, that will leave stuff in Scene, un-closed.
[15:57] Teravus Ousley: .. and if you tried to go to log-in again.. you won't see neighbor regions.. or you'd have some other weirdness
[15:58] Gareth Ellison: m_scene.RemoveClient(AgentId);
[15:58] Gareth Ellison: teravus - i have empirical evidence to the contrary
[15:58] Gareth Ellison: i login and logout all the time with this method and it works fine
[15:58] Gareth Ellison: OGP users don't leave ghosts lieing around iehter
[15:58] Gareth Ellison: *either
[15:59] Gareth Ellison: they TP out or logout, and next thing they fade away
[15:59] Teravus Ousley: cool. :D. I'm not really interested in playing 'who's right' though.
[15:59] Gareth Ellison: no ghost
[15:59] Gareth Ellison: i'm just playing the "i know it works, so why not implement it" game ;)
[16:00] Tao Takashi: I gotta go
[16:00] Tao Takashi: cya all!
[16:00] Saijanai Kuhn: I think that Gareth's work illustrates that we dont know all the obscure things about how the OGP is working right now
[16:00] BlueWall Slade: bye Tao
[16:00] Whump Linden: later, Tao, take it easy
[16:00] Tao Takashi: have a good afternoon :)
[16:00] Teravus Ousley: Take care Tau :D
[16:01] Teravus Ousley: ok, I'm going to finish some planning for my day tomorrow (work related) then I'll break into the OGP.
[16:01] BlueWall Slade: see ya
[16:01] Whump Linden: cool, thanks Teravus, let me know if you need anything
[16:02] Whump Linden: I'm going to step away for a minute and get cofffee
[16:02] Saijanai Kuhn: whump at what point do we declare this meeting over?
[16:02] Teravus Ousley: thanks :D
[16:02] Gareth Ellison: i'm thinking of abusing local grid TPs to do selective UDP proxying
[16:02] Whump Linden: it's over

Open Grid Public Beta/Daily Huddle Transcripts/Transcript 20080818

Navigation menu

Search