AW Groupies/Chat Logs/AWGroupies-2009-06-02

From Second Life Wiki
Jump to: navigation, search
  • [9:31] Zha Ewry: Good morning all
  • [9:31] Morgaine Dinova: I should have added an initial question to the agenda on the Gridnuats interop: "Was everyone happy with step #1? Was it all working as intended?"
  • [9:31] Nexii Malthus: good..afternoon
  • [9:31] Morgaine Dinova: Hiya Zha!
  • [9:32] Flimsey Freenote: Hey Zha
  • [9:32] Zha Ewry: Time Zones.Sigh.
  • [9:32] Zha Ewry: I'm going to give people a couple more minutes.. and then I was planning on spending a little time talking about two way pipes, deployment patterns and how we might actually seperate out services friom servers, so we get a nice general pattern
  • [9:33] Morgaine Dinova: Nice. Is that before or after the written agenda?
  • [9:33] Zha Ewry: We have written agenda?
  • [9:33] Rex Cronon: hello everybody
  • [9:33] Zha Ewry looks for it
  • [9:33] Nexii Malthus: https://wiki.secondlife.com/wiki/AW_Groupies_In-World_Meeting_Agenda
  • [9:33] Morgaine Dinova: https://wiki.secondlife.com/wiki/AW_Groupies_In-World_Meeting_Agenda
  • [9:33] Saijanai Kuhn: https://wiki.secondlife.com/wiki/AW_Groupies_In-World_Meeting_Agenda
  • [9:34] Saijanai Kuhn: any of those links will do...
  • [9:34] Nexii Malthus: lol
  • [9:34] Morgaine Dinova: May the best link win
  • [9:34] Zha Ewry: Well you all may talk about 1) all you want. T hey are entirely Linen Lab Policy decisions
  • [9:35] Rex Cronon: now we know who is good at typing:)
  • [9:35] Sebastean Steamweaver: Beg your pardon?
  • [9:35] Zha Ewry: As such I will happily say "Policy of Linden Lab is not somehting I have an opinion on"
  • [9:35] Sebastean Steamweaver: Hey there NExii
  • [9:35] Zha Ewry: "TP to Main grid, asset Bit, all the associated stuff, is 100% policy choices of Linden Lab"
  • [9:35] Morgaine Dinova: AWG is not LL though, Zha. We have interop requirements, and they're expressed in AW Groupies. And in a sense we're a pressure group on LL.
  • [9:36] Zha Ewry: I'm just pointing out that you should feel free to pressure away. But they are policy statements and requests of Linden lab. Not architecture or programming issues.
  • [9:37] Morgaine Dinova: Given an AD, we may use nothing but LL's OpenID Provider service.
  • [9:37] Saijanai Kuhn: Philip said that 'they're working on it' in Groupies a few weeks ago. I'm guessing the most we can get is an update on the status of hte work (if we're lucky)
  • [9:37] Morgaine Dinova: Indeed.
  • [9:38] Saijanai Kuhn: Infinity won't be able to make it to Groupies today, but thinks she'll be at Zero's if I understood her correctly
  • [9:38] Zha Ewry: Well, you have nobody here attm who can speak for Linden Lab.. So. we can talk about it, but.. answers aren't going to be obtained here.
  • [9:38] Morgaine Dinova: The topic is really a discussion, not a proposal. What do WE want of LL? It's been a year since Gridnauts interop happened. Assuming that interop isn't dead, what do we want for step #2?
  • [9:38] Zha Ewry: That was my understanding of Infinity's schedule as well
  • [9:39] Rex Cronon: jumping to wow and back?
  • [9:39] Zha Ewry: I think, realisically, you aren't going to see anything to main grid until:
  • [9:40] Zha Ewry: a) you have the inventory bit, b) you have some form of X.509 implemented so that Linden has reason to trust other regoins and ADs.
  • [9:40] Morgaine Dinova: Don't forget that we have Zero OH at 1pm today. So if perchance there is a general view on what we'd like for Gridnauts interop step #2, there's our chance to get it noted and feedback on it from LL.
  • [9:40] Zha Ewry: But.. That's a personal speculation, not anything beyond that
  • [9:40] Lillie Yifu: First time I have had attachments stripped on tp in a while
  • [9:40] Morgaine Dinova: Don't need trust for any of those 4 points in the agenda
  • [9:40] Zha Ewry: You are nicely ruthed, Lillie
  • [9:41] Morgaine Dinova: Trust is being used as an excuse for doing nothing.l
  • [9:41] Morgaine Dinova: If trust is a showstopper, do those things that don't require it. Simple.
  • [9:41] Zha Ewry shrugs
  • [9:41] Saijanai Kuhn: political pressure from content designers probably prevents them from discussing things in more detail
  • [9:42] Zha Ewry: If there's no interest in a technical discussion, I think I'll go write some code
  • [9:42] Morgaine Dinova: So let's do the easy case: interop for those assets wholly owned by the TP'ing party. Is there a problem there?
  • [9:42] Saijanai Kuhn: I'm quite willing to move on to your agenda Zha
  • [9:42] Rex Cronon: i have a suggestion for step #2! how about setting up a dedicatd grid(not more than 4 sims, maybe not even that) for testing
  • [9:43] Morgaine Dinova: Zha: I'm happy to move to your non-agenda item after you've looked at the agenda item.
  • [9:43] Zha Ewry: I have nothing to say, Morgaine. Telling Linden Lab how to run thier company is not my job
  • [9:44] Morgaine Dinova: Zha: stop that. AWG is not an LL rubber-stamping exercise. Have some spione.
  • [9:44] Morgaine Dinova: And spine too.
  • [9:44] Dahlia Trimble: why would LL listen to zha?
  • [9:44] Morgaine Dinova: The reason we discuss things with LL is because we DONT just say yes.
  • [9:45] Morgaine Dinova: There is value in second opinions.
  • [9:45] Zha Ewry: I think its safe to say that Linden is not unaware of the desires of some people to do more on interop sooner
  • [9:46] Zha Ewry: Feel free to give thier business development and technical planning people a business case which supports thier adding scarce resources to the task.
  • [9:46] Dahlia Trimble: it doesnt appear to me that LL is interested in interop at all anymore since the upper management changes
  • [9:46] Morgaine Dinova: Well interop is just like bug triage of Jiras. If we express an opinion on what is most desired, and it's viable, then it would get more attention that items which we do not want. It think. Isn't that a reasonable assumption?
  • [9:46] Morgaine Dinova: Dahlia: uh oh
  • [9:46] Morgaine Dinova: Is Dahlia right?
  • [9:48] Morgaine Dinova: If LL doesn't want interop anymore (assumption), then probably nor does the LL/IBM duo, which may explain why IBM has put a hold on Zha's AD (but Zha's won't be able to comment on that of course, no worries).
  • [9:48] Zha Ewry: I think Infinity is pretty supportive of the next steps, and has been talkingt o people who want to work on X.509, on the policy language issues and basic plumbing
  • [9:48] Saijanai Kuhn: well, seeing how MMOX was announced after M took charge, I'm not sure that that is true
  • [9:49] Zha Ewry: And, OGPX is on the agenda for the next IETF, with supoprting work
  • [9:49] Zha Ewry: At the same time, I can count developers
  • [9:49] Saijanai Kuhn: anyway, this DOES seem to be turning into a rehash of our problems with LL's feet-dragging on certain things
  • [9:49] Morgaine Dinova: I don't want to rehash old discussions. I'd simply like to get a feeling for what the natural step #2 is for Gridnauts.
  • [9:50] Zha Ewry: The current hold up on getting the AD out, is the matching client. Itd oesn't work with any publically available code at the moment.
  • [9:50] Saijanai Kuhn: Zha, what is needed for it to work?
  • [9:50] Zha Ewry: Well
  • [9:50] Zha Ewry: Its painful
  • [9:50] Morgaine Dinova: Can't you get it to work with PyOGP?
  • [9:50] Zha Ewry: The OGP chanegs are on a 1.19 base for the most part
  • [9:51] Rex Cronon: what about what sai has done? it could be moded to work with whatever comes out? isn't that a possibility?
  • [9:51] Zha Ewry: On top of which we've added the code needed to properly handle inventory over caps and some changes
  • [9:51] Zha Ewry: We're looking at several sane paths for getting the client code out as reasonable patches
  • [9:52] Zha Ewry: (I've never looked at it directly, mind you)
  • [9:52] Zha Ewry mutters about GPL and issues beyond my control
  • [9:52] Saijanai Kuhn: the only possible way to work it with pyogp right now would be to have a regular client sitting in the same sim as pyogp's to monitor what the pyogp client is doing.
  • [9:52] Zha Ewry: On the source side
  • [9:52] Morgaine Dinova: Jeez, the mindless legal separation (not blaming Zha)
  • [9:52] Zha Ewry: OpenSim
  • [9:52] Morgaine Dinova: So what's the solution?
  • [9:52] Saijanai Kuhn: is the sim code backwards compatible with non GOP clients?
  • [9:52] Saijanai Kuhn: OGP*
  • [9:53] Zha Ewry: I'm currently waiting for the dust to settle on Mellanie and Diva'smassive basing ont he services model in OpenSim
  • [9:53] Zha Ewry: at which point, I'll re-base onto the post dust settled code
  • [9:53] Zha Ewry: My personal preference, would be to have the OGP changes go into snowglobe
  • [9:53] Saijanai Kuhn: i.e., can you log a regular client into a sim while bypassing the AD login?
  • [9:54] Zha Ewry whinges about multiple UUIDs and pain
  • [9:54] Rex Cronon: why not rewrite both the client and the server. as one set of fucntions classes are added to server, functions/classses to test the communication with server should be added to the client
  • [9:54] Dahlia Trimble: what happened to the prior OGP client?
  • [9:54] Zha Ewry: practically, I hope to have answers in the next few weeks
  • [9:54] Zha Ewry: the current OGP client isn't maintained in terms of forward marges with mainline
  • [9:55] Morgaine Dinova: Are there any showstoppers in getting together with Enus and Sai (I'll help too) and getting the AD to work with OGP first, instead of waiting for Diva and co?
  • [9:55] Saijanai Kuhn: ZHa, does the current client work, even half-assed, with the sims that can handle your AD login?
  • [9:55] Zha Ewry: You can login normally, that's no big hassle. The problem is.. that.. you're not going to get your inventory on the cap path
  • [9:55] Dahlia Trimble: last I tried the current OGP client it crashed every time I tried to log into an OGP sim
  • [9:56] Zha Ewry: Sigh.
  • [9:56] Saijanai Kuhn: Zha, so you have a visual of your pyogp client and do your protocol testing with pyogp
  • [9:56] Rex Cronon: to me the amount of time spent on "fixing" the server and client seems to be the same as if u were to do a complete rewrite
  • [9:56] Zha Ewry: I don't think.. that the current Vaak deployment is valid, but someone shoudl ask Enus
  • [9:56] Dahlia Trimble: I still have an OGP region running
  • [9:56] Zha Ewry: The server code on OpenSim, is circa 3,000 lines, II think. Not huge
  • [9:57] Zha Ewry: The client changes are messier
  • [9:57] Saijanai Kuhn: which is why you don't do them with the visual client during testing.
  • [9:57] Morgaine Dinova: Dahlia: I don't suppose it's possible to put pyogp in front of Idealist as a proxy, handle the interop in pyogp, and pass everything else to your back end?
  • [9:57] Zha Ewry: essentially a gridder approach
  • [9:57] Dahlia Trimble: Morgaine, it's open source, go for it ;)
  • [9:57] Morgaine Dinova: Haha, well said :-)))
  • [9:58] Saijanai Kuhn: the libraries would handle it, but there's no proxy shim written to handle that yet
  • [9:58] Dale Innis: ( Do we switch to Zero's office at 1? )
  • [9:59] Zha Ewry: At the moment, doing the AD work and such, has been very motivating for understanding what's totally undocumented in the OGP base stuff Linden has pushed out
  • [9:59] Rex Cronon: this meetin usually lasts untill 11, so plenty of time to go to zero:)
  • [9:59] Morgaine Dinova: Dale: yeah, but don't forget about Opensim office hours at 1100
  • [10:00] Zha Ewry: In particular, the whole specs are only *just* starting to accomodate the notion that they define sets of services, which might be deployed differently than Zero's current visino of how to partition them.
  • [10:00] Ahuva Heliosense: hep
  • [10:00] Zha Ewry: There are some *lovely* fifty foot wide assumptions at the moment jammed in the middle of them.
  • [10:01] Ahuva Heliosense: yep. but dont see anyone or anything
  • [10:01] Zha Ewry kicks the sim
  • [10:01] Zha Ewry: Its hould not belaggy
  • [10:01] Morgaine Dinova: Realistically, we're not going to get Zero to add additional services to his blessed AD+RD model. He's not flexible.
  • [10:01] Sebastean Steamweaver isn't suffering lag.
  • [10:01] Techwolf Lupindo: zero is not at his office yet.
  • [10:02] Saijanai Kuhn: not for 3 hours I would expect
  • [10:02] Zha Ewry: Actually, Infinity has been very responsive to specific feedback
  • [10:03] Zha Ewry: Zero.. I think, hasn't been really forced tot hink through the patterns yet
  • [10:03] Zha Ewry: The moment you ask "What happens if I redeploy a "service" off the "agent domain" server the client is talking to.. things get interesting
  • [10:03] Morgaine Dinova: Zero's clearly engaged in something else, probably SL2 (if LL has any sense). Since he's been AWOL for 8 months now.
  • [10:04] Kopilo Hallard: sigh
  • [10:05] Zha Ewry: I don't get special insights into Linden's tasking of it's technical people
  • [10:05] Zha Ewry: Well, beyond being sure that Propsero isn't working on any OGP material ;-(
  • [10:05] Zha Ewry: (sigh, joke, bad one)
  • [10:05] Morgaine Dinova: Well we know Zero's preferences from 18 months of talking. And we know that he's not working on any of them atm. :-)
  • [10:05] Saijanai Kuhn: enus: saijanai: last i checked it does, and last i checked vaak was ok
  • [10:06] Morgaine Dinova: Prospero's doing good stuff in astrophysics meets though, last one was great.
  • [10:06] Zha Ewry nods
  • [10:07] Zha Ewry: Just not going to be handing out any more restart bears
  • [10:07] Morgaine Dinova: Dahlia: you're still a cloud after all this time. Which client is that?
  • [10:07] Dahlia Trimble: I am? I'm using snowglobe
  • [10:07] Zha Ewry: Yeah, and popping groups isn't forcine you to rez
  • [10:07] Dale Innis: ( Cloudy for me also )
  • [10:08] Dahlia Trimble rebakes...
  • [10:08] Morgaine Dinova: LOL
  • [10:08] Zha Ewry: And..Crashes?
  • [10:08] Nexii Malthus: lol
  • [10:08] Morgaine Dinova: Poor Dahlia
  • [10:08] Twisted Laws: baked clouds?
  • [10:08] Zha Ewry: Ouch
  • [10:08] Flimsey Freenote: mm crunchy
  • [10:08] Kopilo Hallard: lol
  • [10:08] Zha Ewry: Ah.
  • [10:08] Zha Ewry: There you are
  • [10:09] Zha Ewry: Either that was a quicky crash, or tp away and back
  • [10:09] Morgaine Dinova: Wb Dahlia . You're now visible too :-)
  • [10:09] Dahlia Trimble: oops I hit ctrl-shift-H instead of ctrl-shift-R :/
  • [10:09] Dale Innis: indeed :)
  • [10:09] Kopilo Hallard: xD
  • [10:09] Kopilo Hallard: >_> <_< >_>
  • [10:09] Zha Ewry: snowglobe's been remarkably more stable since the three big thread crashes werr sorted last week
  • [10:10] Twisted Laws: was working great for me till i teleported here and then it kept locking and eventually made me relog, then all was well again
  • [10:10] Zha Ewry: Whast I've observed, is I get lots of good discussion from Infinity (and sometimes even Zero) when its focused on plumbing and security for OGP
  • [10:11] Zha Ewry: and.. very litttle interest, I think, in the short term, on deployment
  • [10:11] Morgaine Dinova: Zha: you mentione Snowglobe in the AD context. Do you reckon that patches for interop will be accepted?
  • [10:11] Zha Ewry: Well, a discussion with Philip shoudl prove interesting ;-)
  • [10:11] Morgaine Dinova nods
  • [10:11] Zha Ewry: We're on his calendar to talk about that
  • [10:12] Zha Ewry: There are mildly horrific bits involved in doing that so it's sanely multi-pathed, of course.
  • [10:12] Morgaine Dinova: Actually, that would be very illuminating. If they reject them (good code assumed), then their interest in interop becomes questionable.
  • [10:13] Saijanai Kuhn: so Zha, Tentatively ENus thinks pyogp will work for logging in, though he'll check to be sure. We'd need a description/definition of the new protocols to add to pyogp for testin
  • [10:13] Zha Ewry: Easily enough done
  • [10:13] Morgaine Dinova: Sai++
  • [10:13] Morgaine Dinova: Zha++
  • [10:14] Morgaine Dinova: Specs on wiki?
  • [10:14] Zha Ewry: No new changes on login, we pass along a few more items on TP, and we've two bits of devious cleverness under the hood, which aren't wildly different than what Diva and Hypergrid do to pass assets around under the radar of the OpenSim regions
  • [10:15] Saijanai Kuhn: not that I know of. caps-based inventory goes beyond the official OGP as far as I know
  • [10:15] Zha Ewry: And. we use a varient of the 1.20 "fetch inventory descendents" cap
  • [10:15] Zha Ewry: That's probably the biggest oddity
  • [10:15] Saijanai Kuhn: that may actually already have a test sample in pyogp...
  • [10:16] Zha Ewry: Not 100% the same as the one that never quite worked on beta
  • [10:16] Zha Ewry observes that never quite worked on beta grid, is a hint there
  • [10:17] Morgaine Dinova: Does pyogp have an OGP-only option, switching off all legacy? (Yeah, I know it wouldn't do much :P)
  • [10:17] Dahlia Trimble: are there any efforts at all towards compatibility in a graphical client?
  • [10:17] Zha Ewry: I'll take a todo to get some of that onto the wiki, or in a similar public place
  • [10:18] Saijanai Kuhn: MOrgaine, right now its all "pull" the example client evokes whatever login/etc methods it wishes to from the same lib
  • [10:18] Saijanai Kuhn: libs(s)
  • [10:18] Zha Ewry: I'll also observe, in passing, that, there is some remarkably nasty couplling between the client, the region and the assignment of ItemIds and UUIDs in the inventory area
  • [10:18] Morgaine Dinova: Sai: understood
  • [10:19] Zha Ewry: (Take a peek at what hypergrid has to hook in there, to see what I mean)
  • [10:19] Morgaine Dinova: Wish we could brain dump Zha
  • [10:19] Morgaine Dinova: Then grep :-)
  • [10:19] Zha Ewry: You'll get there
  • [10:20] Kopilo Hallard: lol
  • [10:20] Zha Ewry: This one, tho, isn't mine. Someone else did the inventory reflecion
  • [10:20] Zha Ewry: Well, the code read
  • [10:20] Zha Ewry: we're looking at the client/region assumptions
  • [10:20] Zha Ewry: The hint is that hypergid ends up shimming things in the communicatiosn stack
  • [10:21] Zha Ewry wimpers softly
  • [10:21] Morgaine Dinova is afraid
  • [10:21] Zha Ewry: Well, keep in mind, that although it's been what, 5 years now?
  • [10:21] Zha Ewry: The Region/Client code were the same code base originally
  • [10:22] Zha Ewry: Every so often, that's really, obvious
  • [10:22] Morgaine Dinova: Hard to shrug off one's inheritance.
  • [10:22] Zha Ewry: Yep
  • [10:22] Zha Ewry: So.. To poke, briefly at one other thing..
  • [10:23] Zha Ewry: I've been banging on implicit coupling
  • [10:23] Zha Ewry: Right now, the client assumes that every service speaks back to it, on the event queue that it has from one of the regoins or the UDP stream
  • [10:23] Morgaine Dinova: Since I can't get anyone to talk about Gridnauts step #2, I'm ending that agenda point (if it wasn't already). But will re-raise it in future one day :P
  • [10:24] Zha Ewry: I think you should, Morgaine.
  • [10:24] Zha Ewry: I'd love to see Linden say something
  • [10:24] Morgaine Dinova nods
  • [10:24] Zha Ewry: So, when you get a cap
  • [10:24] Dahlia Trimble <-- afk but will check chat log on return...
  • [10:24] Zha Ewry: which is giong to cause async notice to the client
  • [10:25] Zha Ewry: So for example, (to make one up, but plausiably)
  • [10:25] Zha Ewry: "Tell me about login/logoff of my friends"
  • [10:25] Zha Ewry: You're implicitly (and that's bad, implicit is a bad thing in interface design)
  • [10:25] Zha Ewry: going to start getting either UDP or eventqueue packets saying
  • [10:25] Zha Ewry: "Joe has logged on"
  • [10:25] Zha Ewry: "Dahlia has logged off"
  • [10:26] Zha Ewry: "Joe has Logged off"
  • [10:26] Saijanai Kuhn: that already happens right?
  • [10:26] Zha Ewry: right
  • [10:26] Zha Ewry: But.. rightn ow
  • [10:26] Zha Ewry: they come in on either the UDP from the region
  • [10:26] Zha Ewry: or the event queue
  • [10:26] Zha Ewry: and ADs don't have a UDP stream
  • [10:27] Zha Ewry: OGP is modeling this as
  • [10:27] Zha Ewry: "posting caps on the client over the rhttp connection"
  • [10:27] Zha Ewry: (or faking it over the event queue)
  • [10:27] Zha Ewry: The effect, is the same
  • [10:28] Zha Ewry: You've modled caps ont he client, but they don't have public http URLs (nor should they)
  • [10:28] Zha Ewry: so, what you want, on some level is tot hink about those notifications as:
  • [10:29] Zha Ewry: (obscure)/friend_status_update
  • [10:29] Morgaine Dinova: I get all this, but don't see the problem (other than the dual source). It's always going to be async unless there's an event gatherer, which then has implications on scalability.
  • [10:29] Zha Ewry: Give me am oment
  • [10:29] Morgaine Dinova: k
  • [10:29] Zha Ewry: Becuase you will
  • [10:29] Zha Ewry: I hope
  • [10:30] Zha Ewry: So.. right now, the implicit pipe is always the eventqueue you got when you stared talking to the AD or region granteing the cap
  • [10:30] Zha Ewry: If I offload, no
  • [10:30] Zha Ewry: I want it to be a seperate pipe
  • [10:30] Zha Ewry: So, what I want is to ask for a cap:
  • [10:31] Zha Ewry: <map>

<key> “Cap requested”</key> <string> request.presence.updates.ogp.org </string> <key> “client_rhttp_endopint”</key>

  • [10:31] Zha Ewry: whoops one more line
  • [10:31] Zha Ewry: <uuid> uuid for current client rhhtp endpoint </uuid> //

</map>

  • [10:32] Zha Ewry: (Now, maybe not a UUID, but some unqiue way of idenifyin the endpoint on the rhttp stack)
  • [10:32] Zha Ewry: which lets the cap granter come back with something like:
  • [10:32] Zha Ewry: Capability response:

<map> <key> “capability returned” <string>“request.presence.updates.ogp.org” <key> “capability uri” <string> https://random.service.provider.com/uuid/presence_update_cap <key> response uri” <string> “https://presence.server.random.service.provider.com/uuid/event_queue” </map>

  • [10:32] Zha Ewry: which says
  • [10:32] Zha Ewry: "My REST resources will be on this cap"
  • [10:33] Zha Ewry: and.. I will be replying with any aysnc stuff on this event queue/rhhtp pipe
  • [10:33] Zha Ewry: The uniqueness token on the request is needed so that we can allow services to share pipes
  • [10:34] Morgaine Dinova thinks about last line
  • [10:35] Zha Ewry: It tells the cap granter: "Here is the current nickname of the host end of the client we're sharing, if you have access to it, just use it. If you don't you'll need to get one setup"
  • [10:35] Zha Ewry: "so, now, the previousl implicit cap gets a nice URI:
  • [10:35] Morgaine Dinova: Given that there will never be more than a very few services, why not keep a pipe per service?
  • [10:35] Morgaine Dinova: Sorry, leave question to later
  • [10:35] Zha Ewry: No, its a good question
  • [10:36] Zha Ewry: several reasons
  • [10:36] Zha Ewry: Most importantly, those pipes are in fact expensive, and easily shared
  • [10:37] Zha Ewry: Making them explicit allows a bunch of useful things
  • [10:37] Zha Ewry: We can, as a service deployer, use them as a nice way to allow pipes to be re-established, in various situations
  • [10:38] Zha Ewry: We can also do things like allow all the child agents to share a pipe
  • [10:38] Zha Ewry: which becomes cogent, if we want to offload some event processing
  • [10:38] Morgaine Dinova: Oh god
  • [10:38] Morgaine Dinova: (Intended as a profanity :P)
  • [10:38] Zha Ewry: hmm?
  • [10:38] Zha Ewry: Hmm?
  • [10:39] Zha Ewry: (chat lag)
  • [10:39] Kopilo Hallard: darn chat lag >_<
  • [10:39] Morgaine Dinova: It's OK, I just preferred the protocol not to be implementation-dependent ... child agents
  • [10:39] Zha Ewry: It isn['t.
  • [10:39] Zha Ewry: totally a service pattern
  • [10:40] Zha Ewry: You can gang any number of services which delievr events onto an event queue
  • [10:40] Zha Ewry: "Sim status updates from adhacent regions" is one candidate
  • [10:40] Morgaine Dinova: KK. Well several reasons for pipe sharing then, that's cool.
  • [10:41] Zha Ewry: The point is that id you're going to model the rhttp stuff correctly, we suddenly get some nice leverage out of it, but we have to actually work out the service pattern
  • [10:41] Zha Ewry: It also has the very ncie effect
  • [10:41] Dale Innis: ( Can't let children share pipes, even in Adult regions )
  • [10:41] Zha Ewry: that if we pull out rhttp and slide in xmpp
  • [10:41] Kopilo Hallard: lol bad pun
  • [10:41] Zha Ewry: Good things happen, because we have the URI of the endpoint
  • [10:42] Zha Ewry: So.. one thing I've been noodling over, is trying to fully describe what a trivial service would look like
  • [10:42] Zha Ewry: and please.. please please do not think I am suggesting we annotate our services at runtime
  • [10:42] Zha Ewry: But...
  • [10:42] Zha Ewry: So.. now, the service for status updates about login/logoff becomes:
  • [10:43] Zha Ewry: Cap name, gets you access to a cap with N REST resources, and specifies an client named endpoint rhhtp,// or eq:// or maybe xmpp://, and a set of messages which can be expected over that pipe
  • [10:45] Zha Ewry: "key update_notidication map key "agent name" string, key "agent uri", string agent uri" key "status string" string "Joe Miller has logged on using the hexagon client" /map" for example
  • [10:45] Zha Ewry: minus typos, and plus a bunch of XML markings
  • [10:45] Zha Ewry: ideally, we'd have that level of description for all the services, and then, bang them into a profile and say "This is OGPX baseline 0.1Alpha"
  • [10:46] Zha Ewry: (and *all* of it being LLSD style maps is extensible, so if you want to slap in "Sound to play" on the status message you can)
  • [10:48] Zha Ewry: <key>sound to play</key> <map> <key> type of sound reference</key> <string> <remoted URI of wave file</string> <key>uri</key><stirng> http://sounds.zha.org/random/annoying trumpet flourish.wav"</string></map></map>
  • [10:49] Morgaine Dinova: That sounds very cool Zha. Going to read it all again later to get the nuances.
  • [10:49] Nexii Malthus: lol
  • [10:50] Zha Ewry: You're getting a pre-view of what Ill be posting towards he OGPX list shortly
  • [10:50] steampunk Jetpack (murga's build - kopilos script): All Go
  • [10:50] Zha Ewry: basically, trying to fully pin down what I'm thinking of as "OGPX underpinings"
  • [10:50] Dale Innis: ( would it be beating a dead horse or raising a ghost to ask why we're not actually using XML at all, except as a content-free map-wrapper? )
  • [10:50] Zha Ewry: "This is what we mean by a "status update service"
  • [10:51] Zha Ewry: And, then, just enough service dsicovery so we can describe how "our" service packages them up
  • [10:51] Zha Ewry cringes at "service discovery"
  • [10:51] Zha Ewry: There are some halfway decent words in the OGP base docs about whaty LLSD not pure XML
  • [10:52] Dale Innis: ( k ty )
  • [10:52] Zha Ewry: the biggest of which is that it's one of several serlizations, and not static)
  • [10:53] Zha Ewry: Ill freely admit, I don't deeply care which of 10,000 or so existign scheme we use, as long as it doesn't hard bind messages onto a compiled like form
  • [10:53] Zha Ewry: (google protocol buffers, seem to come mightly close to doing that)
  • [10:54] Mystical Demina: just seems things that don't need the optimizations around efficient package distribution could use easier tools and be easier for other people to implement if you use more standard formats
  • [10:55] Morgaine Dinova: XML is an encapsulation. LLSD provides a choice of them, so you can use XML if you want --- keep an eye on negotition options. Many of us would like to add a second binary serialization of Google Protobufs.
  • [10:55] Zha Ewry: Yep
  • [10:55] Zha Ewry: And the spec is going to say:
  • [10:55] Nexii Malthus: eek, andrew linden office hour
  • [10:56] Morgaine Dinova: Opensim office hours now too -- bad clash.
  • [10:56] Zha Ewry: LLSD defines the layout, the serilizations are: XML, Binary,JSON (with specific mappings) and any tother we define
  • [10:56] Techwolf Lupindo: later all
  • [10:56] Morgaine Dinova: Cyu tech
  • [10:56] Rex Cronon: TC
  • [10:56] Zha Ewry: The problem with GPBs is I think you'll find it hard to do extensible messages
  • [10:56] Zha Ewry: Not impossible
  • [10:56] Zha Ewry: Just not elegent
  • [10:56] Dahlia Trimble arrives just in time to go :/
  • [10:56] Morgaine Dinova: lol
  • [10:56] Kopilo Hallard: danr sleep winning >_<
  • [10:56] Dahlia Trimble: bye all :)
  • [10:56] Rex Cronon: going 2 adrew 2. tc all
  • [10:57] Dale Innis: I thot the OpenSim meeting was at noon SLT. Is this the Wright Plaza one?
  • [10:57] Mystical Demina: yes
  • [10:58] Annie Obscure: ok, I'm off myself
  • [10:58] Annie Obscure: thanks all
  • [10:58] Mystical Demina: but thinking it is at 1
  • [10:58] Zha Ewry: Okies. 11:00 SL
  • [10:58] Zha Ewry: Time to call it a meeting
  • [10:58] Zha Ewry: See people at Zero's OH
  • [10:58] Mystical Demina: checking
  • [10:58] Mystical Demina: sorry looks like osgrid is now
  • [10:59] Dale Innis: I am seeing one person in Grasmere if that's still where Zero is.
  • [11:01] Morgaine Dinova: Thanks Zha, great meeting
  • [11:02] Zha Ewry: Thanks. Trying to basho nt he concepst