User:Zero Linden/Office Hours/2008 Dec 04

[8:36] Saijanai Kuhn: g'day all
[8:36] Rex Cronon: hi sai
[8:36] Zha Ewry: Serious lack of lins
[8:36] Zha Ewry: *lindens
[8:37] Tree Kyomoon: the metaphore holds, but it doesnt make business sense in terms of making this a platform instead of just a dollhouse.
[8:38] Saijanai Kuhn: from realxtend channel:
[8:38] Saijanai Kuhn: AdamZaius: If you want to help with SOG/SOP, I need a simple document which has the explanations of the new RexProperty generic messages
[09:36] AdamZaius: (used on meshes/etc)
[8:39] Saijanai Kuhn: we really ned to get these things ccordinated with OGP
[8:39] Zha Ewry: Well, that requires either Linden gettign engaged, or opening up OGP as a less Linden Centric structure
[8:40] Patnad Babii: hello everyone nice seeing you
[8:40] Rex Cronon: hi
[8:40] Saijanai Kuhn: well, I know that the lindens are monitoring OpenSim more closely, but they aren't dealing directly with realxtend at all that I can see
[8:41] Zha Ewry nods
[8:42] Zha Ewry: But, at the moment, the whole OGP process, is pretty much a captive effort of Linden, so they either need to get engaged in the process, more seriously, or open it up more. The current path is not really wondrous.
[8:42] Saijanai Kuhn: right. That's why I keep pushing that tired old thing about outgoing cap and service discovery blah blah
[8:43] Zha Ewry: Sure, but.. the problem is without someone editing and pushing new specs, that's whistling in the dark
[8:44] Saijanai Kuhn: that's what Adam is talking about with the realxtend folk at this very sec
[8:44] Zha Ewry nods
[8:45] Zha Ewry: Which starts to tear apart the center of mass at Linden, but that is not suprrising
[8:46] Zha Ewry notices that gets people'
[8:46] Morgaine Dinova: Lemme guess ... I'm ressed on Sai's head, right? ;-)
[8:46] Animation Overide - Modified: Could not find animation '---*MM* Model pose 6'.
[8:46] Zha Ewry: s attentin
[8:46] Zha Ewry: Of course, Morgaine
[8:47] Saijanai Kuhn: AdamZaius: Heh, good luck.
[09:46] AdamZaius: That standard is kinda dead without Linden Participation, due to the large gaping flaws in it's design. (hi eventqueue!)
[8:47] Zha Ewry: Actually, more like on his belly
[8:47] Morgaine Dinova chuckles
[8:47] Saijanai Kuhn: urp
[8:47] Tree Kyomoon: anyone here ever done any coding in MQL4?
[8:47] Morgaine Dinova: So Lindens are no show again?
[8:47] Zha Ewry: So it seems
[8:47] Saijanai Kuhn: Lindens are in economic meltdown mode, I gather
[8:48] Tree Kyomoon: economic meltdown?
[8:48] Zha Ewry blinks.. "Is that based on anything other than rumor?"
[8:48] Saijanai Kuhn: recession, financial crisis, etc etc
[8:48] Tree Kyomoon: you mean due to the general crisis or something LL specific?
[8:49] Saijanai Kuhn: just from the comments whump made and the general tone (or lack) from the LIndens. They're all suddenly wearing 2-3 new hats
[8:49] Morgaine Dinova: Not surprised. Even Google would be in economic meltdown if 95% of their machines were idling while 5% are a melted slag heap.
[8:49] Saijanai Kuhn: AdamZaius: Saijanai, my input would be, just about everything between the client and the region server needs to be scrapped and rewritten.
[8:50] Morgaine Dinova: I'd agree with that
[8:50] Morgaine Dinova: But I'd go further and include the region server in it too
[8:51] Zha Ewry: sigh
[8:51] Zha Ewry: Okies. I'm off to do something hopefully more productive
[8:51] Morgaine Dinova: The only thing we have that's scalable are the viewers
[8:51] Zha Ewry: Nope, Morgaine, They blow up on NxN rendering issues too
[8:51] Morgaine Dinova: Oh sure, but that's noddy to fix
[8:52] Saijanai Kuhn: AdamZaius: Well yeah.
[09:51] AdamZaius: Therein lies the problem, LL cant really afford to scrap the current protocol
[09:51] AdamZaius: since it'd nessecitate rewriting very large chunks of the viewer due to how ingrained it is
[8:52] Tree Kyomoon: isnt that approach like fixing an engine knock with a sledgehammer?
[8:52] Morgaine Dinova: The point about the viewers is that for each extra user, you get an extra chunk of CPU. That's scaling
[8:52] Tree Kyomoon: there has to be a clever solution
[8:53] Zha Ewry: But, I don't get an extra mip for each user in my scenegrapch
[8:53] Morgaine Dinova: Just turn down what you render to your locale, easy.
[8:53] Saijanai Kuhn: Kane_Hart: whats that mean Saijanai
[09:51] Saijanai: is there anything that can be done to coordinate all the directions people are going with OpenSIm, realXtend, libmv, etc, given the LIndens are busy elsewhere?
[8:53] Morgaine Dinova: And FFS tell the server not to send you stuff you aren't going to render
[8:53] Saijanai Kuhn: Biskit: I have wondered that same thought too - I'd hate to see this get so splintered up that nothing is compatible with anything else.
[09:53] SachaMagne: ReX are forking back
[8:54] Morgaine Dinova: Biskit++
[8:54] Tree Kyomoon: that seems to be the zeitgeist of social networking in general
[8:55] Tree Kyomoon: a big splinterfest
[8:55] Goldie Katsu: we could be unconventional and do it differently
[8:55] Morgaine Dinova: It is a problem. The fact that LL is slow isn't a problem ... at least they maintained cohesion. But their ABSENCE is a major problem.
[8:56] Saijanai Kuhn: SachaMagne: compatible with what ?
[09:55] Saijanai: realxtend vs gpl viewer vs libmv vs opensim vs SL
[09:55] sdague: then get in and write some code meetings don't get you anywhere
[8:56] Tree Kyomoon: LL web services arent in XML are they?
[8:57] Latha Serevi One thing that has happened with the Web standards congeal around popular clients. Perhaps that's something we should expect / work with.
[8:57] Saijanai Kuhn: udp is binary ish. CAP stuff sets it in XML-LLSD
[8:57] Patnad Babii: the best would be to use the better technic, don't worry to much about what linden have at the moment and set it as a standard no ?
[8:57] Morgaine Dinova: Yeah
[8:58] Saijanai Kuhn: 'well, the LIndens want control so they can have backwards compatibility
[8:58] Tree Kyomoon: beta was better than vhs...I must point out
[8:58] Morgaine Dinova: Well LL isn't going to have control, because they're absent landlords
[8:58] Zha Ewry nods
[8:58] Zha Ewry: that is the thing that will burn them
[8:58] Zha Ewry: If they don't get involved, the community will go elseere
[8:58] Saijanai Kuhn: SachaMagne: ReX are working on better avatars/inworld contents
[09:57] Biskit: I foresee at some point a new release won't be fully compatible with existing code - i.e. forced SL viewer updates
[09:58] SachaMagne: opensimulator goal is not cloning secondlife
[8:59] Zha Ewry: there will be a bunch of breakage and costs
[8:59] Saijanai Kuhn: Biskit: and at that point a decision will be made to go in a different direction
[09:58] SachaMagne: opensimulator is a framework
[8:59] Zha Ewry: but.. also aooprtunities
[8:59] Zha Ewry: The biggest breakage,is the client
[8:59] Morgaine Dinova: Who is the nearest thing we have to a Zero in the OpenSim + forks area?
[9:00] Saijanai Kuhn: SachaMagne: some contributors are working on better graphical environment
[9:00] Zha Ewry: My non casual, measure of it is 25 PY to do a base and another 25 PY to get a 20FPS on mid-range hardware
[9:00] Saijanai Kuhn: Adam and jhuurliman and...?
[9:00] Tree Kyomoon loves forks
[9:00] Zha Ewry: Adam has *some* architectural judgement
[9:00] Zha Ewry: jh, hasn't shown much yet, btu he has some people at Intel to beat him on it
[9:01] Latha Serevi: What client shall we use going forward? Non-LL-hacky yet can visit SL? One idea is to use the concept of an "adapter" proxy or client layer to insulate a client from changes to a particular server.
[9:01] Saijanai Kuhn: SachaMagne: and what pple want ?
[10:00] SachaMagne: crappy avatar or 2009's avatars ?
[9:01] Zha Ewry: That's the biggest messy bit
[9:01] Zha Ewry: and, since content and users (with network effect) is a big part of the puzzle, the one I htink people constantly underesitimate
[9:02] Tree Kyomoon thinks people will want to play WOW, then teleport to nike island and buy a shoe they can wear in WOW
[9:02] Tree Kyomoon: and then teleport to a rolling stones concert
[9:02] Morgaine Dinova: There are lots of people banging on clients, I don't think it's too big a problem, except that they don't have a server spec to hack to.
[9:02] Latha Serevi: I propose an explicit adaptor/translation mechanism to recognize that not all servers will be identical.
[9:02] Zha Ewry: 10 puddles of cool function with 10 viewers is much less effective than 1 or 2 .
[9:04] Zha Ewry: Nobody has a ground up render pipeline that is tied directly to the SL viewer, and that's where most of the hard work ends up, the REX guys come closest, and they give up a lot, to get what they do (and still use a bunch of Ll client code)
[9:04] Zha Ewry: that *isn't.*(
[9:04] Morgaine Dinova: Well we can't virtualize different server systems through a proxy, because the proxy wouldn't scale. It's got to to be the client that adapts, because it has the grunt.
[9:04] Goldie Katsu: Not all servers will be identical but you do need some core minimum functionality that (or reasonable fakage) for the server.
[9:05] Latha Serevi: Morgaine, I think that since there'll likely never be a server spec that gains much acceptaqnce, it might be better strategy to have a couple of dominant clients instead.
[9:05] Zha Ewry: You also need some core consenus on how to render the basic scenegraph.
[9:05] BlueWall Slade: could both not be built to a standard?
[9:06] Saijanai Kuhn: that was the AWG goal t create a 0.9 standard as a starting point
[9:06] Morgaine Dinova: Well given a common plugin arch, all the legion of clients could quite easily incorporate server adapters to suit.
[9:07] BlueWall Slade: where does the current OpenSim fit into this?
[9:07] Zha Ewry: right now, OpenSim, is a 90% echo of SL with aspriatoins to be a more general platform
[9:07] BlueWall Slade: so, is it a good platform to build on?
[9:08] Zha Ewry: Its a platform. Good, is in the eyes of the beholder. I've seen worse code in lots of other places
[9:08] Morgaine Dinova: And it will be a generic platform I think, because it's not suffering the logjam that SL is.
[9:08] Zha Ewry: It has some spots I'm not fond of
[9:08] BlueWall Slade: as a testbed
[9:08] Zha Ewry: (it loves to use .net remoting where it should not)
[9:09] BlueWall Slade: is that somethng that can be fixed?
[9:09] Zha Ewry: Anythign can be fixed, if people want to badly enough
[9:09] BlueWall Slade: it seems that there is a lot of interest to get it right
[9:09] Zha Ewry: Part of the problem is that the current guts replicate the Linden UUID single world model far more deeply tna is desirable
[9:10] Zha Ewry: And, yes, I totally agree
[9:10] Morgaine Dinova: The Imprudence viewer peeps (Jacek + McCabe) said it pretty clear, and correctly I think. LL can't *afford* to move quickly, even if it could, which it can't.
[9:10] Latha Serevi: Morgaine makes an interesting point about a plugin architecture; I wonder if OGP could end up being the plugin API, with server hacks below the API and client hacks above it? As opposed to OGP being a spec for the over-the-wire traffic.
[9:10] BlueWall Slade: so, could we use it to develop the protocols for open VW's
[9:10] Zha Ewry: It has to be an over the wire spec, plugin specs are architecture of programs, bnot interoperablity spec. Both are useful
[9:10] Morgaine Dinova: OGP is just a protocol. Plugins are kind of orthogonal to it.
[9:10] Morgaine Dinova: Aye, as Zha says
[9:10] Zha Ewry: But they server very different purposes
[9:11] Saijanai Kuhn: well service discovery and connecting plugins to plugin servers, can be part of the AD function or not but how it works needs to be set out for all to see
[9:11] Morgaine Dinova: But I'm not even sure that OGP is going to take hold, if LL are abandoning us as they are. People will just use something simpler off the top of their head.
[9:12] Goldie Katsu: Can OGP be unhitched from Linden.
[9:12] Latha Serevi: Why are people saying LL is abandoning us? Did I miss a news item?
[9:12] Zha Ewry: The total lack of effort, It hink Latah
[9:12] Morgaine Dinova: Just no show, Latha.
[9:13] Zha Ewry: Whump pretty much did a "Ok, winding down on the gridnauts effort" e-mail last week
[9:13] BlueWall Slade: I think the new management wants to take SL the Compuserve route
[9:13] Morgaine Dinova: They have troubles. But from our perspective is that they simply don't show up even for 5 mins.
[9:13] Tree Kyomoon: is it a lack of effort, or is the revenue model dropping and they are focussing ?
[9:13] Cane Janick: yeah I saw that
[9:13] Morgaine Dinova: Who knows
[9:13] Latha Serevi: I've always thought they inherently have the wrong motivation to be the primary driver for an OGP effort anyway.
[9:13] Zha Ewry: and Zero's been AWOL, and.. nobody is saying "yeah, thisi where it goes next" while
[9:13] Patnad Babii: i think if we come up with something working then they will want to adopt it
[9:13] Goldie Katsu: I think we need an evolving OGP now, whether LL is there or not. I suspect there will be an upswing in VW work soon beyond what's going on now.
[9:13] Zha Ewry: I agree Goldie
[9:13] Morgaine Dinova: Ditto
[9:13] Zha Ewry: And, at some point, the core work needs to go
[9:13] Zha Ewry: away from a Linden editing style
[9:14] Morgaine Dinova: That's why I asked who OpenSim's "Zero" is .... someone to take the reigns
[9:14] Zha Ewry: to a more traditional "core group of ownsers"
[9:14] Goldie Katsu: And the buy in portion needs to be there.
[9:14] Zha Ewry: Realistically
[9:14] BlueWall Slade: I have suspected that LL would end up being a follower in all this , just not this soon
[9:14] Goldie Katsu: Or there will be the OGP of Linden and the one more commercial users use.
[9:15] Goldie Katsu:or rather - the one most vw providers use whether commercial or open
[9:15] Zha Ewry: you want to grab IETF, or OASIS, or Open Group, or similar to hold the IPR and structure stick. Inventing a new standards body, is not a lto of fun
[9:15] Morgaine Dinova: But "follower" is OK, BlueWall. But they're not following, they're simply absent
[9:15] Zha Ewry: Yes, exactly
[9:15] Goldie Katsu: yeah
[9:15] BlueWall Slade: Diva Canto has done some great work in OpenSim recently with the Hypergrid
[9:15] Zha Ewry: if Linden said "Well, ok. OGP and AWG and OpenSim fols, give us a design and implementatoin for X, and we'll help out and critqiue"
[9:16] Saijanai Kuhn: crash and sent me bck to last sim ick
[9:16] Morgaine Dinova: Yeah
[9:16] Cane Janick: WB
[9:16] Morgaine Dinova: Comes back to what we were saying yesterday. They're not good at harnessing the community.
[9:16] BlueWall Slade: I think she would like to get it in the AD too
[9:17] BlueWall Slade: well, they feel like they will loose
[9:17] Zha Ewry: I think that is a good discussion
[9:17] Saijanai Kuhn: I'll need the last 5 minutes for chat logging purposes if anyone can help
[9:17] BlueWall Slade: they have the wrong business view
[9:17] BlueWall Slade: they think it will harm instead of expanding
[9:17] Latha Serevi: Zha, to get back to the topic of OGP as plugin vs over-the-wire spec ... I feel like there's more I want to figure out. Surely over-the-wire is not the only place for the OGP spec to live? Some wire traffic ain't going to be OGP, but could be adapter-ed, can't we provide some kind of framework for that?
[9:17] BlueWall Slade: did anyone logon to Compuserver this week?
[9:18] BlueWall Slade: k
[9:18] Zha Ewry: afk one minute
[9:19] Saijanai Kuhn: thanks Patnad
[9:19] Morgaine Dinova: Latha: it's more of a case of OGP being specific to client-server operations, whereas plugins would perform duties for the client ... it's not too likely that they'd directly talk OGP to the server, as that gets a bit out of control.
[9:20] Goldie Katsu: ah Patnad beat me to it.
[9:20] Patnad Babii: np my pleasure ;)
[9:20] Saijanai Kuhn: Morgaine, plugins MIGHT need to inform the AD that they are handy (or not) at the least, we need a plugin architecture defined that won't stomp on other stuff
[9:21] Saijanai Kuhn: SachaMagne: i feel ogp is doomed to end. too much lobbies
[10:20] Biskit: ok - as I stated before, I don't keep up with things, I tend to try and find bugs
[10:20] BlueWall: maybe Diva will be interested in merging Hypergrid into the OGP ideas?
[9:22] Goldie Katsu: So...what is hypergrid?
[9:22] BlueWall Slade: it is the linking of grids in OpenSim
[9:22] Goldie Katsu: (I feel like i'm reading messages from the future.)
[9:23] Goldie Katsu: is it written down?
[9:23] BlueWall Slade: we can already jump from grid to grid
[9:23] BlueWall Slade: it is working fairly good
[9:23] Goldie Katsu: Rather is it written down beyond code?
[9:23] Morgaine Dinova: http://opensimulator.org/wiki/Hypergrid
[9:23] Goldie Katsu: ah thank you.
[9:23] Tree Kyomoon: nice
[9:23] BlueWall Slade: a lot of work is being done - and a lot of it is trying to figure out what is happening in the viewer
[9:23] dogtow Hand: thx :o)
[9:24] BlueWall Slade: thanks Morgaine
[9:24] [[User: Saijanai Kuhn: [10:24] SachaMagne|Saijanai Kuhn: [10:24] SachaMagne]]: i guess OGP was serious at the begining, but now it's more a joke
[9:25] Saijanai Kuhn: BlueWall: I think they got new management in M/Q, etc. and they don't get it
[9:25] Morgaine Dinova: It's that kind kind of thing that the Opensim project manages to chuck out in just a few months, which would take LL years. The future is clear.
[9:25] Goldie Katsu: Yeah, they don't have enough people and spare bandwidth to do what community collaboration can do.
[9:26] Patnad Babii: well i think the need for OGP is real, the more VW use it the more wide the grid can be
[9:26] BlueWall Slade: It would be great to see everyone who is interested take a look at what is happening there
[9:26] BlueWall Slade: it is open, and good developers are welcome
[9:26] Morgaine Dinova: Well certainly the need for interop is real. The alternative is just too horrific to contemplate
[9:26] Goldie Katsu: When/where are the meetings?
[9:26] Patnad Babii: if we want to replace the web at some point, all VW have to speak the same line
[9:26] BlueWall Slade: and good code finds it way to core
[9:27] BlueWall Slade: and there is a site for module development as well
[9:28] BlueWall Slade: http://opensimulator.org/wiki/Forge_Related
[9:29] Morgaine Dinova: Do the Opensim peeps ever gather in-world, or is it all mail and IRC?
[9:29] BlueWall Slade: We have meetings on tuesdays in-world
[9:29] Zha Ewry: The gather in OpenSim
[9:29] BlueWall Slade: on OSGreid
[9:29] Morgaine Dinova: Aha, neat
[9:29] Saijanai Kuhn: http://wiki.secondlife.com/wiki/AW_Groupies#In-World_Meetings
[9:30] Patnad Babii: maybe we should be more close with them and bring our idea at their meeting
[9:30] BlueWall Slade: ther are some really god content creators working there as well
[9:30] Patnad Babii: because it seem it have to be done without the linden at the moment
[9:30] BlueWall Slade: over 1,000 regions on OSGrid now
[9:30] Goldie Katsu: Yeah I agree.
[9:31] Patnad Babii: opensim peeps seem more "open"
[9:31] Morgaine Dinova: And "present" ;-)
[9:31] BlueWall Slade: well, you can connect regions to test
[9:31] Goldie Katsu: present I think being essential
[9:31] BlueWall Slade: we run clise to svn head on OSGrid
[9:31] BlueWall Slade: close
[9:32] BlueWall Slade: and have a mantis system
[9:32] Saijanai Kuhn: Well, once Belxjander and/or Zha get independent ADs working, we can keep OGP going with or without LL participation. Did you ever chat with Bel, Zha?
[9:32] BlueWall Slade: we have IRC integration in the region module, so we can conduct meetings in-world and bridge them to IRC
[9:32] Goldie Katsu: very nice
[9:32] Goldie Katsu: which IRC?
[9:33] BlueWall Slade: freenode
[9:33] BlueWall Slade: #osgrid-wp is the meeting place
[9:33] Goldie Katsu: tyvm
[9:33] BlueWall Slade: #opensim, #opensim-dev for OpenSim related
[9:33] Morgaine Dinova: Aha, Wright Plaza
[9:33] BlueWall Slade: #osgrid for people connecting to OSGrid
[9:34] Rex Cronon: bye everybody. rl calls:)
[9:34] Morgaine Dinova: Cya Rex
[9:34] Tree Kyomoon: bye rex
[9:34] Cane Janick: kk rex
[9:34] Zha Ewry: At this point, doing a C#, clean Agent Domain is going up my list, but.. to be honest, the real challanges, are more client side, and sorting out nice ways of ducking the single grid assupmtions which are pervasive
[9:34] Goldie Katsu: cya Rex
[9:35] Morgaine Dinova: Meerkat seemed to be heading in that direction Zha, decoupling from the single-grid model. But they're curiously quiet ... where's Gareth?
[9:35] BlueWall Slade: Zha, some people are working on viewers - so it will evolve
[9:35] Saijanai Kuhn: well, the protocol part of an AD is trivial--can be done in LSL or flash. It's the details of tracking avatars and whatnot that are complicated, IMHO
[9:36] Goldie Katsu: yeah the client is going to need a lot of changes - identifying the assumptions at least
[9:36] Goldie Katsu: and yes, it definitely will evolve and we see a lot of work on the clients.
[9:36] dogtow Hand: c# would work with newer faster machines & less resources ?
[9:36] Eddy Stryker: i'm curious to hear what problems the agent domain is attacking?
[9:37] BlueWall Slade: hopefuly this takes off, and LL will want to stay involved
[9:37] Zha Ewry: I want a clean C# AD, to keep ins ycn with the openSim code base
[9:37] Eddy Stryker: between the distributed grid services i'm working on and the hypergrid project that has been merged into opensim trunk, what else is left?
[9:37] Zha Ewry: The AD, gives you a single point of contact, seperate from regoins and grids
[9:37] j3rry Paine: lol have it on my desk next tuesday at 11 sltime
[9:37] Eddy Stryker: couldn't a well written user server do the same thing?
[9:38] BlueWall Slade: ++ on the distributed assets
[9:38] Zha Ewry: Somwwhat, but at that point, its an AD
[9:38] Morgaine Dinova: There's no need for AD to be part of Opensim, it's a separate service.
[9:38] Eddy Stryker: ok
[9:38] Zha Ewry: and at the moemnbt, the User Server pretty couple to a single grid notion
[9:38] Eddy Stryker: i will rename my user server to agent domain server then :)
[9:38] Zha Ewry: which isn't quite right
[9:38] BlueWall Slade: yes
[9:38] Eddy Stryker: not the one i'm writing in my distributed grid services
[9:38] Zha Ewry nods
[9:38] Zha Ewry: then, possibly renaming it does the trick
[9:39] Zha Ewry: What we're seing at the moment
[9:39] Zha Ewry: is the total fragmentation of the effort, tho, which doesn't serve anyone
[9:39] Eddy Stryker: i don't have strong design docs up for the user server yet, unfortunately. you can kind of see where it fits in reading through the distributed asset server documentation, but i'll have more clear documentation up in a week or two
[9:39] Morgaine Dinova: But moving ahead in parallel is pretty powerful too Zha
[9:40] Goldie Katsu: parallel is good, knowing where it fits together is good too
[9:40] Goldie Katsu: otherwise you can have tracks that don't match in the middle.
[9:40] Saijanai Kuhn: jhurliman I've gotten Whump and Infinity to commit to work on dragging the current group IM into OGP/AD as a way of keeping us moving, providing a generic outgoing endpoint for communciations, and a discovery mechanism for services accessed through the AD
[9:40] Morgaine Dinova: I think the only danger to parallel effort is when they end up requiring different client hacks.
[9:40] Zha Ewry: Which is one of the conerns
[9:41] Saijanai Kuhn: But Infinity's been unavailable, and of course, the whole economic thing has everyone tightening their belts at LL and rethinking office hour schedules and so on
[9:41] Tree Kyomoon: that seems pretty significant Saij/Jhurliman
[9:41] Eddy Stryker: saijanai: i want to make sure my user server work captures all of the latest and greatest features being thrown about. can you talk about the "discovery mechanism"?
[9:41] Saijanai Kuhn: not yet :-(
[9:41] Eddy Stryker: (i'm not touching the messaging server this year if i have any luck at all)
[9:41] Zha Ewry: Well, Infinity is back from dealng with family crisis, so. .that shoudl get better
[9:42] Saijanai Kuhn: that was what Infinity was workign on before she had her family emergency
[9:42] Eddy Stryker: what is the discovery mechanism supposed to do?
[9:42] Morgaine Dinova: Sai: I don't think it's belt-tightening that we're seeing. No way are they telling Lindens to "work harder". They're not that daft.
[9:42] Tree Kyomoon wonders if eddy still sees iridium from time to time :)
[9:42] Saijanai Kuhn: let the client ask the server for contact points with gorup IM, and whatever else we can think of
[9:42] Eddy Stryker: tree: she just left for work
[9:42] Eddy Stryker: (we
[9:42] Zha Ewry: Lucky Eddy. ;-)
[9:42] Eddy Stryker: 're engaged)
[9:42] BlueWall Slade: they are getting new watermarks on the avatars online
[9:43] Zha Ewry: Gratz
[9:43] Tree Kyomoon congratulates eddy...and please give a fond hello to her
[9:43] Saijanai Kuhn: gratz
[9:43] Goldie Katsu: gratz
[9:43] Eddy Stryker: saijanai: ah ok. so a client only has to know about an agent domain, and the AD tells the client where to find everything else?
[9:43] Eddy Stryker: that sounds reasonable
[9:43] Eddy Stryker: ty
[9:43] Zha Ewry: Well, more generically, you'd love to get to a structure where for all the services, you can get a well known starting point, and find what is implemented at that servicep oint
[9:43] Saijanai Kuhn: I'd like to see work on things like collaborative music plugins and the like but without a defined way for the client to ask about services and to send outgoing messages, its all pretty much up in the air
[9:44] Eddy Stryker: hmm
[9:45] Zha Ewry: One of the major challanages, if we do parallel innovatoin, is konwing which features are present in each chunk of the world
[9:45] Saijanai Kuhn: a plugin might use its own custom connection and protocols but we need a defined way for the client to start talking, I think
[9:45] Morgaine Dinova: Sai: well all that was originally going to be pretty easy, just REST services. But somehow it got bogged down in caps without taking us to the stage of giving us service links.
[9:45] Zha Ewry: REST, pretty well sucks at that, as a pattern, btw, but for good reasons
[9:45] Zha Ewry: Thing, is Morgaine, REST doesn't tell you how to enumerate resources
[9:46] Zha Ewry: Its one of the more amusing blind spots
[9:46] Zha Ewry: (where amusing == want to rip your hair out)
[9:46] Morgaine Dinova: Zha: sure it does ... one service simply returns a bunch of links.
[9:46] Saijanai Kuhn: well that is where service discovery mechanism comes in
[9:47] Zha Ewry: The challange is knowingwhat the resources mean
[9:47] Saijanai Kuhn: how to ask what links are available and is that different than asking fo ra specific link?
[9:47] Zha Ewry: REST tends towards a "the user/programmer knows" approach
[9:47] dogtow Hand: cpu & mem
[9:47] Morgaine Dinova: Which is the correct one, since the client knows what it's trying to do.
[9:48] Zha Ewry: There are some yes buts there
[9:48] Morgaine Dinova: That's what allows services to be stateless ... the client has a goal, and the service doesn't need to know what it is.
[9:49] Zha Ewry: The problem comes about at the matching possible goals with possible mechanisms level
[9:49] Zha Ewry: it bog easy, if the client knows to look for named resource X
[9:49] Eddy Stryker: let me throw out one design pattern we used in the asset server... when you want an asset, first you have to fetch it's metadata at a defined location. defined location is assethost.com/assetid/metadata (the assethost.com information is always contextual or explicit). metadata is returned that contains a list of "methods", which are pre-defined keywords that a client may or may not understand that map to urls
[9:49] Morgaine Dinova: When Google returns an answer to a query, it doesn't know whether your next request will be for page 2 or page 200.
[9:49] Wrapp Seiling: This is just gettining very interesting folks.... but rl calls....ttyl/
[9:50] Zha Ewry: and, it shouldn't have to
[9:50] Morgaine Dinova: Cya Wrapp
[9:51] Zha Ewry: but. when you're trying to match 10 possible resources with10 possible plugins (content negotiation style) that's not much help
[9:51] Zha Ewry: and, a bunch of the services we want to expose
[9:51] Zha Ewry: are more peer to peer, than driven by the client. Not hard, but some structure would help quite a bit
[9:51] Zha Ewry: At a minium
[9:52] Zha Ewry: you need to actually have some well known startin points
[9:52] Zha Ewry: Not at all clear I need more than
[9:52] Zha Ewry: "region_server/offered_services"
[9:52] Morgaine Dinova: Zha: just keep it client-driven: If client says "Gimme a list of X", then the client's X handler will handle the result. There is no types problem when it's client-driven.
[9:52] Zha Ewry: to do a get on
[9:53] Zha Ewry: One second...
[9:53] Zha Ewry: I want to toss a link at people
[9:53] Saijanai Kuhn: Morgaine, sure the client can talk to any server it wants, but for authentication purposes within the metaverse, seems the AD should be in the loop at the start THEN bow out
[9:54] Goldie Katsu: and what are the x's that can be requested, from where?
[9:54] Zha Ewry: http://duncan-cragg.org/blog/post/getting-data-rest-dialogues/
[9:55] Goldie Katsu kicks herself again for not cleaning up her default browser of all flash windows before the meeting.
[9:55] Zha Ewry: Look in particular, at Duncan's discussion on the way that you cross manage gbay and ebay in there
[9:55] Morgaine Dinova: Goldie: the client knows the X's, since it's programmed with their handlers. As to "where" ... each grid has an entry point for that, and it doesn't even need to be the AD, since many services could be public, needing no auth.
[9:56] Zha Ewry: That breaks down, a bit, Morgaine, when it's the moment that Region Simulator one wants to pass a request on to regoin simulator two.
[9:56] Zha Ewry: The client is one step removed fromt hat picture
[9:56] Morgaine Dinova: Sure, it'll have to talk to the ADs before then. However, I'm just pointing out that service discovery is not a stumbling block.
[9:56] Zha Ewry: Its not a deep one
[9:57] Zha Ewry: Duncan's patterns in there, offer a lot of food for thought
[9:57] Morgaine Dinova: Thanks for that Zha
[9:57] Zha Ewry: But, they general reality, is that REST (and 90% of web services) get into some degree of messyiness when you have versioning and varients
[9:58] Zha Ewry: (for that matter, I'd argue, that 90% of programming has issues with versioning and varients)
[9:59] Zha Ewry: Discovery (or if you think og it as introspectoin and reflectoin, in non distribuetd contexts) are ways of coping with that
[10:00] Freemason Magic: a new collaborative programming platform i heard can help this problem is JAZZ.NET
[10:00] Morgaine Dinova: Zha: does that interview get more serious beyond the first page? "eBay: we don't care about scalability and interoperability|Morgaine Dinova: Zha: does that interview get more serious beyond the first page? "eBay: we don't care about scalability and interoperability]]: we have quite low traffic on our APIs - certainly compared to port 80; plus we're eBay so we don't need to interoperate with anyone."
[10:01] Zha Ewry: it does
[10:01] Morgaine Dinova: kk
[10:01] Zha Ewry: the writing style is light
[10:01] Zha Ewry: but, look at the sectoins where he walks through the SOA vs REST tradeoffs in subscribing to interest lists
[10:02] Zha Ewry: And mind you, he does use the e-bay architect to setup and knock down a few strawmen, watch for that
[10:02] Zha Ewry: But.. he's not far off in putting those words in e-bay's mouth
[10:03] Morgaine Dinova: Yeah, it's getting better. And it was just a knockdown approach
[10:05] Saijanai Kuhn: in an hour, we have Which's office horus, if he's available. I'll be there recording the .log but will need to leave after a few minutes.
[10:06] Saijanai Kuhn: might be good to continue this at hsi hours since he's a real live linden
[10:06] Saijanai Kuhn: and he's sorta into this stuff at the code level
[10:07] Morgaine Dinova: And at 3pm, there's the User Experience OH --- don't miss it, as it's the 1st of the month, so Benjamin and other Lindens will be there (in theory).
[10:07] Zha Ewry: OK. I need to deal with annoying things like Lunch and next phone calls
[10:07] Morgaine Dinova: And since Jacek's a leader there, the Imprudence viewer is close by
[10:07] Saijanai Kuhn: KK I'll merge my notecards and log and post thi. This was an important meeting, I think
[10:08] Goldie Katsu: So Which is at 11 LST?
[10:08] Saijanai Kuhn: right, Goldie
[10:08] Goldie Katsu: I agree, I look forward to the transcript.
[10:08] Morgaine Dinova: Which is a particularly good viewer project to get attached to, because it's run under Git so easy branches and merges.
[10:08] BlueWall Slade: thanks everybody
[10:09] Zha Ewry: I think my metaphor for the holiday season is going to be surfing
[10:09] Morgaine Dinova: Hehehe
[10:09] Morgaine Dinova: Just hope it's a manageable wave
[10:10] Zha Ewry: Well, yes
[10:10] Goldie Katsu: Maybe I should be learning to snowboard *looks out at the snow covered trees*
[10:10] Zha Ewry: Disruptive change has that shape
[10:10] Saijanai Kuhn: jhurliman: for example, i think the opensim team is going to resist OGP teleport as they just finished implementing the (IMHO, superior) hypergrid teleport model
[10:10] Saijanai Kuhn: saijanai_: seems to me that hypgrid could fit behind the scenes with the client requesting a generic TP/Logoutofsim/logintosim call and the grid handles it behind the scenes as it likes
[11:03] Saijanai Kuhn: saijanai_: the client call is to the ad and the ad does its thing and passes the new seedcap back to the client. lots f magic goes on behind the scenes the client doesn't know about
[11:04] Saijanai Kuhn: saijanai_: http://wiki.secondlife.com/wiki/OGP_Explained#OGP_Draft_5_Teleport
[11:05] Saijanai Kuhn: jhurliman: i'm just saying, i don't think it will gain a lot of traction because it doesn't provide anything new
[11:05] Saijanai Kuhn: jhurliman: teleport between grids already works fine
[10:10] Saijanai Kuhn: saijanai_: OK. I think what the AD adds is the authentication thing, which would be important for a commercial grid like SL
[10:11] Zha Ewry: hypergrid, teleport, also does funny things to the way you knti things togteher
[10:11] Morgaine Dinova: Ask jhurliman if there are any showstoppers in his TP mechanism that would prevent a "walking through a mirror" type of teleport, ie. TP portal in the middle of a region
[10:12] Saijanai Kuhn: he's sitting right there (waves at Eddy)
[10:12] Morgaine Dinova: Oh, lol :-)
[10:13] Morgaine Dinova: I thought you were cross-posting from IRC because he wasn't here ;-)
[10:13] Goldie Katsu: lol
[10:13] Saijanai Kuhn: well, he made a comment in irc so I passed it back here
[10:13] dogtow Hand: :o)
[10:13] Zha Ewry: I think, at the moment (and this may be a client driven hack issue)
[10:13] Saijanai Kuhn: I did passyour comments back to him since he's not paying attention here
[10:13] Saijanai Kuhn: jhurliman: no nothing, i think that's being added soon through a new scripting function
[10:14] Morgaine Dinova: Coolness
[10:14] Zha Ewry: that the only major difference, is that hypergrid is sort of anchored at the moemnt on the physical token to tp
[10:14] Eddy Stryker: sorry, SL doesn't beep at me like IRC nick-completion does
[10:14] Zha Ewry laughs
[10:14] Zha Ewry: usefl tool that, if one has sound up
[10:14] Saijanai Kuhn: is there any reason why the AD can't do a call to hypergrid protocols?
[10:15] Zha Ewry: not clear you'd bother doing it that way
[10:15] Zha Ewry: Look, the AD is going to wither away, unless Linden has a reason for it
[10:15] Morgaine Dinova: Aye, I have IRC feeding into Festival, even nicer. IM needs decoupling so we can do things like that too.
[10:15] Saijanai Kuhn: seems to me that the AD handles authentication in a way acceptable to SL. If HyperGrid can talk to the AD and do its thing via the AD, there should be no diff
[10:16] Zha Ewry: Actually.. at the moment, the AD is basically "trust me"
[10:16] Morgaine Dinova: Then it offers you a bridge ... :P
[10:16] Eddy Stryker: the only real thing the hypergrid protocol does is send the target grid a list of home services for an avatar. so the inventory server and asset server where the user's assets are gets accessed from the foreign grid
[10:16] Saijanai Kuhn: sure, but trust can be firmed up. Is there an equivalent or better trust mechanism in hypgrid?
[10:16] Zha Ewry shrugs
[10:17] Zha Ewry: OpenSim doesn't have a huge focus on trust and authentication
[10:17] Eddy Stryker: saijanai: hypergrid isn't a security mechanism, it's a way of teleporting. the asset server, inventory server, and user server (ahem, AD) have to worry about trust. not the teleport mechanism
[10:17] Eddy Stryker: unles you define "untrusted teleport"
[10:17] Zha Ewry: But there is *no* good reason to do it Linden's way, if they aren't playing
[10:17] Saijanai Kuhn: right. I don't see there being a real conflict here. WoW has its own TP mechaism. but if you want an interface with Wow, provide an interface to the AD's request for login/TP
[10:18] Saijanai Kuhn: Zha I don't think they have said "not playing," but rather "super busy in immediate coding mode"
[10:18] Zha Ewry shrugh
[10:18] Zha Ewry: that is internet for "lettiing the wave wash over you"
[10:19] Morgaine Dinova: You know, the analogy with game worlds isn't innacurate at all. A lot of them provide access to their database services, so that third party websites can display weapons and armor sets etc.
[10:20] Zha Ewry: OK. this time lunch calls for real. 10 minutes before food goes away for the day.
[10:20] Morgaine Dinova: Cya Zha :-)
[10:20] Saijanai Kuhn: KK transcript ends here. Merge merge

User:Zero Linden/Office Hours/2008 Dec 04

Navigation menu

Search