[9:46] Zha Ewry: Especially when he got interested
[9:47] Zha Ewry: The meat of the discussion circled around the RealXtend folks work Avatars, and a very nice discussion abut Interop, which was three way between Adam (Zaius/Frisbe) , Phillip and myself, with some nice cross discussion
[9:48] Zha Ewry: The Realxtend folks have done thier own Avatars, complete with totally new skeletons
[9:48] Zha Ewry: Which rocks, tho.. scales, to maybe 5-10 aves a sim
[9:48] Strife Onizuka: did they define it through the LAD interface or a new interface?
[9:49] Zha Ewry: They hacked on a totally seperate interface
[9:49] Burhop Piccard: So Sun, Qwaq and SL (business) are all intresting to me. Any interop discussion/interest/joint effort(gasp)
[9:53] Strife Onizuka: (building on windows only allows them to target directx
[9:53] Morgaine Dinova: We don't want it to be in opensim svn. That wouldn't be extensibility, that would be adding to a single monolithic system. Extensibility needs to be dynamic.
[9:53] Gareth Ellison: what was the problem with *nix builds on rex?
[10:02] Gareth Ellison: hides the 16-pack of red bull and smiles innocently
[10:02] Tess Linden: I think their approach is to get something working, but at least they are checking in with us to make sure their approach matches our thinking
[10:02] Zha Ewry: the ReX stuff really points out the extension points
[10:02] Zha Ewry: The lack fo them, and the need for them
[10:02] Gareth Ellison: here's the extension stuff in GMMP (how it works): each message type has a hash
[10:06] Saijanai Kuhn: Gareth that template was frozen months ago
[10:06] Zha Ewry: Adam, then pushed a lot on the question of how we wanted to keep all the money/permisions stuff seperable from the mainline protoocl pipe
[10:06] Zha Ewry: Which got a fair bit of pushback
[10:06] Gareth Ellison: saijanai - was it frozen with gaps for any unseen changes?
[10:07] Zha Ewry: Well, the problem is that at the protocol level
[10:08] Tess Linden: I explained that in a Gameworld, the Gameworld's agent/avatar domain would have to be responsible for the "inventory" and look of the avatar, esp since it has strict rules about what can and cant be taken out of the world
[10:08] Gareth Ellison: goes back to consult the sacred texts of circuit.py
[10:08] Zha Ewry: you need a clean way of dealing with them
[10:08] Gareth Ellison: for money you need a clean and reliable way of ensuring the transaction goes to whatever system finishes it off
[10:08] Saijanai Kuhn: which is where Which and chttp might come in
[10:08] Zha Ewry: Restrivtive grids are going to be an interestign case
[10:08] Gareth Ellison: where that system is becomes a social/legal matter
[10:11] Morgaine Dinova: The fact that two protocols share a common datum isn't an argument for making the two protocols one. SMTP and HTTP aren't one protocol just because they both feature hostnames.
[10:11] Gareth Ellison: as i see it, chttp + some kind of crypto signing
[10:11] Zha Ewry: There was a bunch of good discussion about that
[10:12] Zha Ewry: and then.. allowing people to build what they need under it
[10:12] Gareth Ellison: yeah, you need a standard for the agent domains
[10:12] Zha Ewry: *some* discusison of the implcations of plugging stff like that in the protocol
[10:12] Lulworth Beaumont: consistent? in terms of providing all the methods/messages to do transactions, but allowing a choice of whether that particular feature is supported or not?
[10:18] Zha Ewry: There are several models for how to manage money
[10:18] Gareth Ellison: and a "i accept your promise of X for the prim cock"
[10:18] Zha Ewry: So.. some good input for this group there
[10:18] Morgaine Dinova: There's no need to reinvent digital cash, it's been done, well, many times. The "bank" should be your wallet, with exchange functionality, not the VW ISP.
[10:24] Gareth Ellison: i doubt any of us here are competent on all the tiny details of the law (unless we have a secret attorney here) but it seems that at the very least any provider would need to check that exchange services don't make them subject to regulation
[10:25] Tao Takashi: and I doubt there is anything regulated now anyway
[10:25] Asterion Coen: Tao integalactic, as soon we will send some citizen out of here :)
[10:25] Zha Ewry: What we're doing.. is playing with havign a small set of regions, which are hosted behind our firewall, with a small, write-through asset server
[10:26] Zha Ewry: They are on the grid, conceptually, but only reachable, inside our shop
[10:29] Gareth Ellison: does that mean i should go and buy IBM stock on the basis that IBM is now an SL competitor or some other twisted distortion, find it isn't true and sue?
[10:29] Zha Ewry: In fact, at the moment, we're working through the final details of how the deply is goign to happen
[10:29] Dr Scofield: and not all of IBM has access to it, do we?
[10:29] Teravus Ousley: Essentially, cloak and dagger tactics hurt my trust in IBM, you as the leader of AWG Groupies, and LL... so I consider that a train wreck of a choice.
[10:29] Gareth Ellison: but seriously: would LL prohibit actively reverse engineering the protocol and trying to link up an opensim instance?
[10:30] Dr Scofield: teravus, how would you have liked this to have happened differently?
[10:30] Saijanai Kuhn: Gareth how would you do anythign?
[10:30] Lulworth Beaumont: if they're not going to prohibit it, they might as well give us the documentation (if there is any)
[10:30] Zha Ewry: We went public, about as soon as we had a formal agreement
[10:30] Vincent Nacon: you mean to steal their system?
[10:34] Zha Ewry: about doing the interop work as part of both AWG and OpenSim
[10:34] Gareth Ellison: if they're happy with people at IBM playing around and trying to get the sim up and haven't got any legal restrictions in the contract against that, what's stopping IBM l33t haxors firing up tcpdump and friends and going at it?
[10:34] Zha Ewry: Contirbuting the code, as we do it
[10:34] Dr Scofield: getting something cleared for publication within IBM is an incredible lengthy process
[10:35] Gareth Ellison: i'm not even talking about releasing the modifications yet
[10:35] Gareth Ellison: just getting a test up behind your own firewall
[10:35] Burhop Piccard: hmmmm... just seems IBM is coving something we are not (or don't have the bandwidth/interest to cover)
[10:35] Lulworth Beaumont: zha: to work well, I think that really does need to be cleared with all of core opensim first
[10:35] Gareth Ellison: once it looks sane and pending any tests for silly security issues, then possibly release
[10:35] Zha Ewry: Its a way for us to drive some understanding and some test work, for how to manage secured regions
[10:36] Strife Onizuka: Gareth, i think you have answered your own question, the licensing agreement most likely restricts what IBM can do with LL's code
[10:36] Zha Ewry: Well.. We're talkign puiblically with AWG about the changes
[10:36] Gareth Ellison: strife - i'm talking about what can be done with LL's network
[10:36] Dr Scofield: views it as an interim solution until we have OpenSim in robust enough state
[10:36] Zha Ewry: And. we're goign to push every line of code we build to the OpenSim trunk, as we do the work
[10:36] Morgaine Dinova: Zha: I hope someone in IBM is thinking about SCO. Some legal parasite could make an issue about IBM transfusing blood directly with LL, you don't wanna go there.
[10:36] Saijanai Kuhn: Gareth, but the main work in the protocols is to divvy things into agent and region domains to allow grid level hookups. Why hack into the IBM subgrid when its obsolete anyway?
[10:36] Dr Scofield: morgaine, we are *constantly* doing that
[10:36] Gareth Ellison: would LL object if opensim was linked *somehow* to the grid?
[10:37] Gareth Ellison: saijanai - i'm not talking about hacking into any grid, i'm talking about IBM putting another sim on their own subgrid
[10:37] Dr Scofield: you don't want to know the pain we go through just to be able to contribute code to projects like OpenSim
[10:37] Gareth Ellison: the only reverse engineering done would be to figure out how to get opensim talking to the LL grid
[10:38] Zha Ewry: We're going to be looking to implement part of the protocol changes Linden is propoising as parts of the OpenSim project
[10:38] Gareth Ellison: as LL aren't willing to release their sim source code yet, likely due to their own obligations on 3rd-party code
[10:38] Morgaine Dinova: Gareth, that would undermine all the efforts for interop entirely, if others could link their sims directly into LL's grid.
[10:38] Zha Ewry: openly, publically, and as part of the project
[10:38] Gareth Ellison: morgaine - if another party (IBM) got some none-LL code linked, that would be a push in the right direction
[10:38] Teravus Ousley: Isn't it true that simulators make calls to mySql.agni directly periodically? If so.. that would need to be refactored into a more purposed interface before it could be released to the public.
[10:39] Tess Linden: but Lul, there are many security questions at hand, we cant just allow anybody to connect in the beginning because we dont know what code they're running
[10:39] Gareth Ellison: Tess - you can trust IBM though?
[10:39] Zha Ewry: and.. we'll be looking at all the asset fetches as part of properly managing the asset abstraction layer
[10:39] Morgaine Dinova: Gareth: that would be a push in the appalling direction of no interop. It would force LL's current internal protocols to become the ABI.
[10:39] Zha Ewry: Well, Gareth, we have a signed legal agreement which covers the trust issues
[10:39] Gareth Ellison: morgaine - it could push further to opening up the source
[10:39] Tess Linden: well, we have an agreement, yes
[10:39] Dr Scofield: gareth, i expect that we a signed legal agreement
[10:40] Strife Onizuka: IBM is a huge multi-national, can you trust all it's employees?
[10:40] Gareth Ellison: zha - that's what i'm getting at
[10:40] Lulworth Beaumont: Tess: true, but because it isn't generally useful, there is an argument that the interop modules shouldn't be part of core opensim
[10:40] Lulworth Beaumont: unless we somehow agreed some general region protcol which would be used by opensim internally anyway
[10:40] Gareth Ellison: if that agreement allows it, i'd think IBM hacking up an opensim mod could be done fairly easily
[10:42] Gareth Ellison: it will almost certainly be a mess, but would be a massive step forward in getting interop going
[10:42] Teravus Ousley: heh, yes. fundimentally.. however for the purposes of this conversation.. connecting one region to another and to a coordinating grid.
[10:42] Zha Ewry: The behind the firewall discussion, is a sideshow to the AWG work
[10:43] Morgaine Dinova: This is taking a very nasty turn, away from general interop with any party that can work the open protocol, and towards special hidden agreements.
[10:43] Zha Ewry: It will et us, and Linden, explore some early ays to get value out of the work
[10:43] Lulworth Beaumont: Tess: However, I do agree with the 'not letting anyone hook up' point to som eextent. Not really for security, but because you don't know if the region you're connecting to runs well or at all
[10:43] Lulworth Beaumont: which is bad if you want to maintain the immersive illusion...
[10:44] Gareth Ellison: i fully agree with the not letting anyone link up too
[10:44] Morgaine Dinova: DrS: but that's how it looks. After all, this other IBM project doesn't contribute to the AWG interop work. It slides under it.
[10:44] Gareth Ellison: at least not just literally anyone
[10:44] Tess Linden: we have to control what we can work with in the beginning so that we can move torward a more trustable protocol that will work for everyone
[10:44] Zha Ewry: What we have is a way to look at how to do some specific work, early, while, we keep doing the interop work
[10:44] Dr Scofield: morgaine, how does it slide under it?
[10:44] Burhop Piccard: Morgaine - I don't see AWG covering everything - LL/IMB its a side issue.
[10:45] Gareth Ellison: Zha - give me a solid yes/no answer - is there any chance of IBM getting the tcpdump out and hacking up opensim mods without waiting on LL to give more specs?
[10:45] Gareth Ellison: AWG still does need to look at sim<>sim traffic
[10:45] Dr Scofield: it allows LL to gain some experience with having regions run by third parties
[10:45] Morgaine Dinova: DrS: it's hooking in directly to LL's grid, not though the public interop interface. That's "sliding under".
[10:45] Saijanai Kuhn: looks on LL-IBM subgrid as a special case. Some things may be learned from it and applied to AWG (and LL internally and OpenSim), but its not the direct AWG work
[10:45] Gareth Ellison: the physical boxes i presume are IBM's property, yes?
[10:46] Zha Ewry: there was a public interop inetrfgace?
[10:46] Dr Scofield: which i expect will result in valuable feedback to the AWG process
[10:46] Zha Ewry: Gaerth? I honestly don't know, and we don't plan to
[10:46] Teravus Ousley: well, there is still the clean room development that must take place.. essentially for that to happen.. someone at IBM would have to document them.. then publish the documentation.. which then an opensim developer would use to design the grid protocol.
[10:47] Burhop Piccard: traffic discussion - yes. But I don't think IBM/LL is there yet. Sure they are working with what is there now, not what AWG is prposing.
[10:47] Zha Ewry: We're planning on working with the AWG for all the sim-sim styuff
[10:47] Saijanai Kuhn: which is a lot of work for no real gain given what Tess is doing
[10:47] Gareth Ellison: if no source is given, no clean room development is needed
[10:47] Lulworth Beaumont: Teravus: yes. We really need docs, although one could argue direct patches *are* docs
[10:47] Zha Ewry: We're looking at exactly one project, at the moment
[10:47] Dr Scofield: as i mentioned on #opensim-dev, there won't be code access for any of the IBM OpenSim devs to the LL code base
[10:47] Burhop Piccard: If security creates traffic issues, then that is when it needs to dovetail into AWG.
[10:47] Saijanai Kuhn: Lulworth, I'd argue that isn't the case, doc-wise
[10:47] Teravus Ousley: right.. which is why it could only happen through documentation.
[10:48] Zha Ewry: Which is a "use" not code project
[10:48] Gareth Ellison: heh, if you could extract the packet logs and strip out all the sensitive bits, then post them in public
[10:48] Gareth Ellison: then i bet they'd end up being used to create SOMETHING compatible
[10:49] Zha Ewry: All of the work Which is doing, on c-http, and escrwo
[10:49] Morgaine Dinova: Well OK, "experience" is being gained of working with SL's innerds. But apparently that's not even coming back to AWG, since it's another team in IM it seems.
[10:49] Gareth Ellison: LL is committed but taking a while to release those sim/sim specs
[10:50] Saijanai Kuhn: Morgaine its an expediency thing. AWG work isnt' far enough along to allow what LL and IBM are doing so they fudged up what they're doing.
[10:50] Lillie Yifu: This is a capability that any open grid will need. If this doesn't flow into open source, then no one will trust the inter opr protocol. So I think that people should not be afraid of this project. Either it works, and it will be folded in,or it doesnt, and it doesn't matter.
[10:50] Zha Ewry: is that all the big corporationg, and many smaller companies, can't use public land in SL
[10:52] Morgaine Dinova: Zha: well one thing the insider team might want to find out is where the dataloss is within the current grid. LL seems to need help with that.
[10:52] Gareth Ellison: on the edge touch, the trust issues don't need as high standards with a closed and firewalled subgrid (as IBM has) as will be needed in public
[10:52] Teravus Ousley: I don't think so Scooter. I think SL wil l thrive regardless
[10:52] Dr Scofield: some people will still want to have property in prime locations
[10:52] Asterion Coen: give a Barbie doll to Morgaine
[10:52] Zha Ewry: Think of this, as a way for us to have some private regions wich our internal security people will let us use for confidentail meetings
[10:53] Gareth Ellison: if the raw basics can be started on now, and then lots of nice auditing before a public open - that would work
[10:53] Lulworth Beaumont: I think SL will provide the premium experience (crazy as that sounds atm)
[10:53] Morgaine Dinova: Asterion: the KVM inserts random characters, one seemed to make me dance :-)
[10:53] Tess Linden: we have several teams working to profile dataloss and fix dataloss
[10:54] Gareth Ellison: just about every sim<>sim message in the message template other than EdgeData i can figure out, and then there's likely secret CAPS
[10:54] Morgaine Dinova: Lulworth: and my guess is the opposite of yours: I think the current SL will become like the original AOL walled garden. And the real metagrid will be what's outside it.
[10:54] Saijanai Kuhn: Gateth, hacking islands without LL cooperation is sily. Either LL should be involved directly with IBM, or the work should be on getting OPenSim grids running, and Agent/Region domains working with OpenSim grids
[10:55] Tess Linden: and back and forth handshaking, we are trying to tease out the "right" things for interop, and we've been documenting it all in the SLGOGP
[10:55] Scooter Back: I find that analogy interesting, Morgaine
[10:55] Gareth Ellison: saijanai - i'm talking about if LL are taking their time to document, rather than wasting LL's time and waiting, get consent to start hacking within IBM's subgrid
[10:56] Tess Linden: Icehouse has been working on the Second Life side of it, i.e. rez_avatar
[10:56] Lulworth Beaumont: Morgaine: Interesting... :) I'm basing my thinking on the idea that if there is full intergrid hookup, then SL as a dedicated company will be able to concentrate on providing much more reliable and featureful regions, actually wihtout opening the software up
[10:56] Gareth Ellison: that cooperation would consist of at a minimal "we won't complain" or better "ok, let's both sit and watch what happens, and here's some more basic specs and staff who'll answer any queries"
[10:56] Dr Scofield: seriously doubts that the IBM legal nazghuls are going to let anyone hack within a mile of those racks
[10:56] Baba Yamamoto: accepted your inventory offer.
[10:56] Lulworth Beaumont: really a service provider like an amazon, google, etc. which use lots of open source but don't republish the stuff they use to run the services
[10:56] Tess Linden: I think many of you are confused because the IBM announcement for the behind the firewall test is in parallel with the work we've been doing to document the Open Grid protocols
[10:56] Gareth Ellison: dr scofield - don't use the H word then :)
[10:57] Zha Ewry: So.. to make a couple of quick points:
[10:57] Zha Ewry: 1) we have a publically annoucned Joint development agreement
[10:57] Gareth Ellison: of course, this opensim hacking could be done within LL on a test grid
[10:58] Gareth Ellison: but that wouldn't be as clean due to the same devs seeing LL's own code
[10:58] Zha Ewry: Most of which, has manifested in AWG work, and Chet Murthy's involvement in c-hhtp and scrow
[10:58] Morgaine Dinova: Lulworth: opening the server software up isn't the key to growth of the metagrid at all, since that software is not scalable nor extensible. The key to growth is the protocol, so that a million other parties can do their own thing, and yet all interoperate.
[10:58] Saijanai Kuhn: Proper place to do that would be the beta grid with the agent/region domain parts working
[10:58] Zha Ewry: We also, talk about lots of possible areas of interest
[10:58] Zha Ewry: the behind the firewall stuff, is one of them
[10:58] Gareth Ellison: man i'd love a shell account on those boxes
[10:59] Zha Ewry: in public, and we'll be talkign to the OpenSim and the greater SL community about some of the testign we want to do, and early chances to get some of the AWG protocol work
[10:59] Tess Linden: I have to head off to another meeting
[11:00] Gareth Ellison: the trust issues with edge touch and the fears of the dreaded H word with IBM's lawyers i can see only being an issue if it was without LL consent
[11:00] Lulworth Beaumont: Morgaine: Yes, I agree about the protocol, but I think SL will be fully hooked up - I can't see them being a walled garden
[11:01] Gareth Ellison: well, i'll concede my point if it's too much risk of "baking in" the legacy protocols into opensim
[11:01] Baba Yamamoto: cut the global identity mess out of the AWG protocol and it might be useable
[11:01] Dr Scofield: rather spends the time on OpenSim work and getting the AWG work forward
[11:01] Gareth Ellison: still believe that it could lead to a faster interop (i.e sooner)
[11:01] Morgaine Dinova: Lul: oh, LL doesn't want SL to be a walled garden, at all. But many of its residents do --- all those that re focussed on restricting the distribution of their beloved "content".
[11:01] Dr Scofield: sigh, morgaine has a point there
[11:01] Saijanai Kuhn: LL is working on a way for grids to interop. OPenSim is (I think) working on grid-level protocols. The trick will be to get the two types of grids workign together
[11:02] Gareth Ellison: let the silly content creators set a "keep my content inside SL, not other grids" flag
[11:02] Zha Ewry: One of the clear points which Phillip and I agreed on.. was that the models for interop need to allow people to restrict how braodly they want content to slow
[11:03] Lillie Yifu: If SL content providers restrict the movement of their content toomuch, then they will find that it is obsoleted. We can make better content with poser and maya now, and open grids will be moer able to do thinsgs like have better avatars, better uv maps and better vh animation. People will migrate to the better looking grids.
[11:03] Zha Ewry: but.. we can allow a rich markup of intent, and let people explore which models work well
[11:03] Saijanai Kuhn: well, its equivalent to the region codes on DVDs...
[11:03] Gareth Ellison: nice little book on how to run a web startup, mentions scrapping feature requests
[11:04] Gareth Ellison: zha - there does need to be a system for ACLs mentioning grids
[11:04] Saijanai Kuhn: bt if folks want region codes...
[11:04] Burhop Piccard: Zha - yes - on content flow.... and something corps will want too.
[11:04] Baba Yamamoto: Gareth, all my philosphy is from teh book of baba
[11:04] Mo Hax: content distribution and economy is destined for the real world withouth vw constraint ultimately
[11:04] Morgaine Dinova: Zha: "I want to restrict my content to X" == "I want to be unpopular"
[11:04] Gareth Ellison: if content creators are that ignorant, let them be
[11:04] Scooter Back: I guress my only concern would be the Lindens in my bank. THEY need to be able to transfer
[11:05] Gareth Ellison: since i think it's a superior license
[11:05] Gareth Ellison: but i won't force that on others
[11:05] Baba Yamamoto: We can afford to piss people off right now... they don't have any other options except to quit.. I don't see them quitting for very long.
[11:05] Gareth Ellison: neither should we force the "let your content go elsewhere" on others either
[11:05] Morgaine Dinova: Zha: what else did Philip have to say about Interop?
[11:05] Zha Ewry: I do think we'll find that various content flow models work better
[11:06] Zha Ewry: well, he's excited to see people like ReX doing things Linden isn't able to invest in today
[11:06] Mo Hax: when the TOS fly out the door because the grid is no longer centralized under linden, well that is the day when content ownership and management issues move back to the real economy and out of ll
[11:06] Teravus Ousley: I have no issues with a module to managed permissions with the understanding that we're marking the content for intent. We're doing what's technically feasable to protect the content.. however technically in-adequate it is.
[11:06] Gareth Ellison: ethically we have no right to force people to give their content away or redistribute it elsewhere, legally we need their license to do it
[11:06] Saijanai Kuhn: people will vote with their virtual wallets "this grid only" except for island builds, wont' be ppular with consumers
[11:06] Zha Ewry: (He was realy clear that the avatrar work ReX was doing, was cool)
[11:06] Baba Yamamoto: gareth, there is nothing to force.. the law of the internet says it will go other places
[11:06] Zha Ewry: No, actually, the discussion on the cost of calculatign physic s collisions between multiple haptically managed avatars, and the computationall load
[11:07] Gareth Ellison: Zha: haptically managed as in "the rig"? :O
[11:07] Saijanai Kuhn: Baba, "this grid only" can be enforced at the asset server level. YOu really canNOT take it with you if things are set up that way. You'd have to hack the textures off your own machine and rebuild things
[11:07] Zha Ewry: Rex has been lookign at using cameras to map body motion on to thier Aves
[11:07] Gareth Ellison: heh, with every DRM scheme there's flaws
[11:08] Tao Takashi: you would see me sitting most of the times ;-)
[11:08] Baba Yamamoto: gareth, one grid can attempt to download it and the other can say yes or no.. I think webservers have the same mechanisms
[11:08] Morgaine Dinova: Zha: that's good. Did he make any noises that might suggest Zero will embrace an extensible protocol that allows such avatar enhancements?
[11:12] Gareth Ellison: heh, i way too often use SL as a laggy IRC
[11:12] Morgaine Dinova: POST mainly starts and controls the object flows, it's not a request per update or that would be disastrously slow, by orders of magnitude.
[11:16] Zha Ewry: The long queue shoudl aggretate lots of the little updates into single pulls
[11:16] Morgaine Dinova: The client needs to be able to specify the maximum rate of updates though, that makes sense, since it varies by client and by circumstance.
[11:18] Baba Yamamoto: the cool thing about object updates now is you receive a single object update per packet.. and if you receive them or not is not important to the protocol
[11:18] Gareth Ellison: me and mr libortp have pestering to do
[11:19] Zha Ewry: and.. puts retry logic on the wrong end
[11:19] Morgaine Dinova: It's currently broken. You could receive an old update after a new one. The real semantics of UDP are almostnever desireable. Or perhaps never, full stop.
[11:19] Gareth Ellison: and put all the objects into one document
[11:19] Zha Ewry: Baba, that's what clients do today
[11:19] Zha Ewry: they open as many http gets as they want
[11:19] Saijanai Kuhn: yeah. If you get a message in EventQueueGet, it can have a dozen events in it, but if one of those events is 1000x as large as any of the others, it creates a bit of a bottleneck
[11:21] Gareth Ellison: if i have to receive 10 billion updates to morgaine's shoe's bling before i get the "OMG! DEATH AND BUNNY EXPLOSIONS!" IM that's bad
[11:21] Zha Ewry: But. I think 10 gets will upll lfar beter than
[11:21] Gareth Ellison: morgaine's shoe's bling can go on another channel
[11:21] Zha Ewry: and.. the caching will be much cleaner
[11:21] Morgaine Dinova: You need to apply some commonsense when trying to force everything into the HTTP straightjacket. It doesn't always fit, and religion shouldn't be a driving force.
[11:22] Baba Yamamoto: firefox is set to 24 http requests concurrently
[11:24] Gareth Ellison: any bunnies around will get my update next frame and i'll be in roughly the same place
[11:24] Zha Ewry: The big thing is to get the trafffic pulled, not pushed
[11:24] Baba Yamamoto: zha, you can't pull object updates smoothly
[11:24] Teravus Ousley: bangs head about the 'pull object updates' idea.. ack
[11:24] Gareth Ellison: but if i send those 20 over the same HTTP channel as everything else, the bunnies may miss the "BUNNY EXPLOSION!" IM or the "I just paid L$2000 for a prim cock" event
[11:27] Lillie Yifu: looks at the 6 foot tall invisible rabbit behind Garret and wonders if that has anything to do with it.
[11:27] Gareth Ellison: morgaine - can you think of where sending AgentUpdates lagged behind so that my "walk forward" arrives after 100s of bunny explosions?
[11:27] Zha Ewry: get the big data dumps off the UDP pipe
[11:27] Zha Ewry: and. then have the long poll. be a fast, priority sorted pipe fro push model stuff
[11:27] Teravus Ousley: Ultimately .. in our current scenario the client's push textures.. the client gets pushed object updates.. which have references to textures.. which the client then requests the textures. How is that different from HTTP except it's occuring on the UDP pipe?
[11:27] Gareth Ellison: i.e if 100 bunnies explode first, and then i walk forward, that walking forward will be out of place
[11:27] Baba Yamamoto: yes.. bunny explosion notices come over udp.. bunny explosion sounds are pulled
[11:28] Gareth Ellison: i press forward, but the bunny explosions get in the way
[11:28] Baba Yamamoto: if it's cacxhed then it's no problem
[11:28] Saijanai Kuhn: Teravus, with one pipe, you have an event list in XML format. Lots of little updates and then one BIG texture... all xml?
[11:28] Gareth Ellison: and so my walk forward comes b*t later (b== bunny explosion count, t== time per explosion)
[11:28] Baba Yamamoto: the server will tell the client "a bunny just exploaded"
[11:28] Teravus Ousley: Currently the client does request textures... based on the texture UUIDs referenced in the object updates that are pushed.
[11:28] Gareth Ellison: baba "bbbbbbbbbbbbbbbbbbbbbbbw" where b is bunny explosion and w is walked forward
[11:29] Morgaine Dinova: Sai: well the long poll (we need another term) mechanism is generic enough to allow it to be used for everything. After all, you can parametrize it to yield a continuous stream of updates, or just one, and the objects will nice and symmetrically enter your system at the same point.
[11:29] Zha Ewry: tess and Zero want it to be general enough
[11:29] Gareth Ellison: the other thing about UDP - i want different classes of bunny explosions
[11:29] Teravus Ousley: Yes, that's the only difference between the current method and the proposed http method.... the underlying protocol that it gets transferred across.. they're both pull methods.
[11:29] Gareth Ellison: what if the BunnyExplosion message has no field for "type of explosive" and "RabbitBreed"?
[11:29] Saijanai Kuhn: Morgaine, the current event QUeueGet sould either have a CAP for the textuture or the texture itself embedded i the xml
[11:31] Teravus Ousley: Yep, and that request prioritizes a list on the server.
[11:31] Gareth Ellison: ok, (bunny example out now, i'm starting to feel silly) - how about push everything, and anything above X size, give a link
[11:31] Lillie Yifu: from a cnetralized server, not peer to peer.
[11:31] Morgaine Dinova: Sai: the texture is just another object, just happens to be a big one. You don't want a separate mechanism for its transport, but merely a way to specify when you want it sent, or incremental updates to it even.
[11:32] Saijanai Kuhn: you don't want textures to come in that format
[11:32] Gareth Ellison: you want a mechanism for big objects
[11:33] Gareth Ellison: please think of giant kitten explosions in seperate short-term CAPS
[11:33] Teravus Ousley: Baba, the client could just as easily only request 1 texture per thread.. and the sim would send them in exactly that order.. because it's the only texture being requested.
[11:33] Morgaine Dinova: Gareth: it doesn't achieve anything to send down a link, unless it's a parameterto future partial requests. But if you can't partition the object, then you might as well just send down the whole thing without a link.
[11:33] Scooter Back: thinks Gareth has something against bunnies and kittens
[11:33] Saijanai Kuhn: that spam was incendentally parsed into standard python format. IT was originaly XML-LLSD
[11:35] Zha Ewry: I think that there may need to be some cool, clever throellign code in the mix
[11:35] Gareth Ellison: that's why notification of their explosion is important
[11:36] Gareth Ellison: links are good to offload big events from the main pipe
[11:36] Teravus Ousley: Unfortunately.. I see far more computational resources spent on http textures then we currently have being used on UDP textures.
[11:36] Morgaine Dinova: Well, to keep Gareth happy (and it's a valid point), the event queue stream really needs to be a bundle of streams, with different priorities attached.
[11:40] Teravus Ousley: I had all of the textures downloaded a minute ago.. I turned my view.. and there is a whole list of texture downloads that popped up.
[11:40] Morgaine Dinova: Tera: don't think "HTTP", think "TCP" :-) The downstream link doesn't close ever, in normal operation, and in principle that's more efficient than UDP today since you can omit state over TCP.
[11:41] Teravus Ousley: There's a really good chance that some of them are the same
[11:41] Zha Ewry: That's got to be horrible caching
[11:41] Teravus Ousley: we're talking about HTTP.. Have you made a simple browser that downloads images off the web?
[11:41] Scooter Back: just think what the'll do when SL converts to DiX 10
[11:42] Saijanai Kuhn: reagredless of client-side caching (don't understand http server side caching), we jus need to make sure its not all coming in through the eventqueueget long poll
[11:42] JayR Cela: has anyone notice that Nicholaz Beresford has give up on SL ?
[11:42] Morgaine Dinova: Objects sent down through the event queue stream can come out of the sim cache just like any other objects, that's no problem. After all, it's all cacheable world state.
[11:45] Saijanai Kuhn: if 6 of those events are tiny update packets and 2 of them are textures mixed i with the updates, things are likely to get messy
[11:45] Gareth Ellison: let's just say entity for clarity when we're talking about "a thing" in general :)
[11:45] Teravus Ousley: My point.. is it remains to be seen if http will actually speed things up or slow them down further.. in this case. If the network connection is bad.. it'll likely speed things up.. If it goes over http, it'll work better on Some ISPS that lower the priority of UDP packets.
[11:45] Gareth Ellison: saijanai - hence why big entities need links
[11:46] Saijanai Kuhn: right, but I'm not sure they're planning on that or even FOR that possibility
[11:47] Gareth Ellison: an event link's body looks like, surprise surprise, a link
[11:47] Asterion Coen: begin a new book 'the wiki AWG: lexic - All barbarian and weird terms used at AWG meeting' beginer's to advanced user levels
[11:47] Morgaine Dinova: Teravus: an even greater optimization then would be the client not having to do a request at all, but merely setting up the continuous downstream :-)
[11:47] Teravus Ousley: Just taking someone's word for it isn't very skeptical
[11:47] Zha Ewry: I'll mention that to Tess and Donovan
[11:47] Saijanai Kuhn: well, int eh eventqueueget, its all events in an array, links can be embedded in the array. But the model they use is that its all hashed out into the UDP packet handlers intenrally in the client
[11:47] Zha Ewry: See if we can get some proper emprical data
[11:48] Gareth Ellison: saijanai - so hack it and have big events hashed into the message handlers
[11:49] Gareth Ellison: lillie - yeah, it's for "anything important"
[11:49] Teravus Ousley: I do think a general TCP connection would run faster then a UDP connection.. however.. I don't think a HTTP request for a texture will run faster. Maybe 30 textures at a time.. parhaps.
[11:49] Morgaine Dinova: Sai: Zero said that it's planned, but that the current implementation is a true poll. But they seem to want to go beyond that
[11:49] Gareth Ellison: there's a lot of important events
[11:49] Gareth Ellison: i'm using a silly example to avoid listing them all
[11:49] Lillie Yifu: no I am sayin gtaht many of the cases that are "anything important" aren't
[11:49] Saijanai Kuhn: if you do a slproxy dump or use my python script and look at what comes in, I can't see parsing a texture as being fast at all.
[11:49] Lillie Yifu: I don't care if you can drop a bunny bomb in my face, it does not enhance my user experience
[11:50] Baba Yamamoto: i love educational institutions
[11:50] Gareth Ellison: lillie - you're taking the bunny too literally
[11:50] Baba Yamamoto: this is a performance analysis from 1996 :)
[11:50] Lillie Yifu: no I am saying that the abilitty ot gnerate interuptive events is inherently not important to many of us
[11:50] Gareth Ellison: things like "you just got kicked from this region for insulting bunnies" are important
[11:50] Saijanai Kuhn: the xml event list can eaily be several KB of text. then add in embedded textures that need to be parsed out of the embedded text...
[11:50] Gareth Ellison: if that notification comes too late, it's annoying at best
[11:51] Lillie Yifu: I don't see why the protocol should be designed aroudn the abilitty ot grief people any tiem anywhere
[11:51] Morgaine Dinova: Although humor is nice, you need to know when to stop. :-)
[11:51] Saijanai Kuhn: is beating a dead bunny he thinks
[11:51] Gareth Ellison: lillie - i take blame for the silly analogy
[11:51] Lillie Yifu: if an area wants those kind of events, let them load when the avatar drops in, and then they will experience them without having to choke on a large raft of textures, physics and therest that is your bunny expolosion
[11:52] Gareth Ellison: but surely there are important events which shouldn't wait in the laggy event queue and bog it down
[11:52] Saijanai Kuhn: so, to my other issue du jur...
[11:52] Lillie Yifu: short important administrtative events, like getting banned or ejected, don'tneed a million flavors
[11:52] Gareth Ellison: you don't need a million flavours no
[11:52] Saijanai Kuhn: right now, we have a 2-part login planned: login and rez avatar
[11:52] Gareth Ellison: but you do need to not bog them down with other events
[11:52] Morgaine Dinova: Lillie: indeed. And the opposite is true and very important: the ability to resist griefing in the protocol is going to be extremely important.
[11:53] Lillie Yifu: that's your user experience, not mine
[11:53] Saijanai Kuhn: Gareth I think we agree on that general point. NOt sure the EQG design handles it though
[11:54] Teravus Ousley: Yeah.. kind of like the UDP's throttles now.. but for the event queue, would have to be implemented
[11:54] Lillie Yifu: the important events I care about can be listed in advance, and teh others determined by area, not by avatar sendingthe griefing event
[11:54] Teravus Ousley: .. puts a small rainbow of items into the UDP stream.
[11:54] Scooter Back: High priority events, such as ground textures, features, av, and region limitations would be high
[11:54] Scooter Back: while flowers and trees would be low
[11:55] Saijanai Kuhn: which goes against the EQG being a tcp implementation of UDP
[11:55] Saijanai Kuhn: I'm still looking at how they handle things right now, not a whole new design
[11:55] Teravus Ousley: it's purpose is to fit the important data into the available working UDP stream.
[11:55] Saijanai Kuhn: the EventQueueGet cap is meant to simulate the UDP. That way they can leverage all the UDP handlers they have already in the client code
[11:55] Teravus Ousley: .. which the client adjusts based on it's packet loss
[11:56] Zha Ewry: Lilie, keep in mind, that greifing and fireworks, look much the same
[11:56] Morgaine Dinova: Well there is no "One hat fits all". Assigning priorities to object types in the download queue needs to be customizeable. Good control over that is the key to nti-griefing anyway. You can make griefing disappear entirely with good control over updates.
[11:56] Saijanai Kuhn: but having one EQG that handles textures AND avatar position updates, means that you need to parse out the xml for textures from the updates before you can handle the high priority updates
[11:56] Scooter Back: especially to the server, Zha
[11:57] Saijanai Kuhn: so, UDP or tcp, it doesn't matter. Textures need to be handled separately. But the EQG doesn't really handle that as it exists. There's only one EQG CAP per sim
[11:57] Teravus Ousley: Zha, if the HTTP method requested 20-50 textures at a time.. I could see it being faster then the current method.. but that would go against our caching method
[11:58] Zha Ewry: Depends how your structure it, T
[11:58] Morgaine Dinova: Having a bundle of download pipes means that you can handle multiple formats too -- eg. binary textures.
[12:00] Lillie Yifu: Zha: a planned event doesn'tneed all of the handling of anunplanned one. If an object on the sim has a sound or other event, that can be preloaded. The problem only occurs if the large unexpected event has to be described fresh. It doesn't need to be in most use cases except those where one avatar is trying to suprise other users.
[12:00] Gareth Ellison: yeah, but an unplanned event is the issue
[12:00] Teravus Ousley: Yes.. HTTP(not to be confused with TCP.. even though the data is crossing TCP) vs UDP
[12:00] Asterion Coen: Saijanai is abble to predict the venue of major bugs 20 seconds b4 they come
[12:00] Gareth Ellison: if your important event is inserted into an already heavily loaded queue that's an issue
[12:00] Lillie Yifu: who has control over the unplanned event is the issue. If the unplanned event is defined by sim, then no,it doesn't ened ot have all the extensibility that you are talking about.
[12:01] Lillie Yifu: It' is only if oyu want to be abel to pop on to a sim, fire off a million flying penises and run away say "Yif in hell fur fags" that the real tiem queue needs to ahve all that descriptitve capacity.
[12:01] Zha Ewry: I have no idea what is meant by unplanned in this context
[12:01] Teravus Ousley: therefore.. when thinking about it.. we must consider the performance of HTTP, not simply TCP
[12:02] Lillie Yifu: sure fireworks exist in an object, it has textures for the particle effects, and objects in rezzers.
[12:02] Gareth Ellison: and one which can be coped with better if you have >1
[12:02] Morgaine Dinova: Tera: if the entities are coming down an unending TCP stream, then what's relevant to downstream efficiency is TCP vs UDP, not HTTP vs UDP. It's not polled you know.
[12:04] Saijanai Kuhn: well, you've just used up both your two legal http connections to the server if you go that route
[12:04] Zha Ewry: Keep in mind, griefer, is a social, not technical marker
[12:04] Morgaine Dinova: Tera: nope, it's open-ended. You might send just one HTTP request for a permanent object update, and get entities coming down the non-closing TCP pipe forever more.
[12:04] Gareth Ellison: griefing doesn't come into it really
[12:04] Zha Ewry: If you can find a way to tell intent, from the set of packets
[12:05] Gareth Ellison: now, depending on the style of music i'd call it griefing, but each to their own
[12:05] Lillie Yifu: the technical differene is that if anevent is part of a sim environment, it can be defined the way objects and textures are curretnly, and does not need a special mechanism ofr extensibility in the real time evnets list
[12:05] Teravus Ousley: gads, Objects again (you really need to be careful with your terminology)
[12:05] Saijanai Kuhn: Morgaine, I'm looking at the next incremental step in the protocols. They're still stuck with an almost unmodifiable code base in the client. We still have to leverage the existing code
[12:05] Baba Yamamoto: HTTP/1.1 Spec RFC 2068 section 8.1.2.2 Pipelining
[12:05] Asterion Coen: i remember one, i listened at some sendboxs when newbie
[12:05] Lillie Yifu: if an individual user can com eon to a sim and expect that whatever real tiem event they want to drop on other people will be handled all the time as they expet to be
[12:05] Saijanai Kuhn: so a streaming TCP connection may not be a rational expectation for a next step out of LL
[12:05] Lillie Yifu: yes that's almost always a griefer use case
[12:06] Zha Ewry: Plenty of things which aren't grief, will look like that too
[12:06] Morgaine Dinova: Sai: why is the client almost unmodifiable?
[12:06] Lillie Yifu: so sim real tiem event = should be in protocol
[12:06] Gareth Ellison: any instance where there's 1000s of events that bog down the event queue is an instance where you benefit from >1
[12:06] Baba Yamamoto: A client that supports persistent connections MAY "pipeline" its requests (i.e., send multiple requests without waiting for each response). A server MUST send its responses to those requests in the same order that the requests were received.
[12:06] Lillie Yifu: user rela time event = protocol should not worry about it
[12:07] Saijanai Kuhn: Baba teh long poll allows the server to send events in whatever order it dems appropriate, including waiting for a while before sending an event packet
[12:10] Morgaine Dinova: last time I looked, nobody had bothered to answer any of my discussion points on event queue.
[12:10] Morgaine Dinova: Does anyone have agenda items for Zero?
[12:11] Saijanai Kuhn: In the SLGOGP? Don't feel bad. I did work to update teh legacy login and no-one has merged that. Tess asked me to work on SLGOGPifying the rez_avatar and I doubt that will get merged etiher
[12:11] Saijanai Kuhn: well, I have an issue about the 2 part login
[12:11] Saijanai Kuhn: It needs to be three parts. Right now its: login => rez_avatar =>stuff
[12:12] Saijanai Kuhn: it should be: login => talk to destination grid => rez_avatar => stuff
[12:12] Gareth Ellison: to the client that step (which i presume is agent domain > region domain) isn't there
[12:13] Gareth Ellison: you're talking about things like the expect_user OGS uses?
[12:13] Asterion Coen: what about some option about "log out" or login (when wanting to change of session or disconneted for idle) ?
[12:13] Saijanai Kuhn: rez avatar is in the SLGOGP or will be
[12:13] Gareth Ellison: "talk to destination grid": elaborate saijanai
[12:13] Saijanai Kuhn: and it could be part of rez_avatar as far as the client is concerned.
[12:14] Saijanai Kuhn: well, the rez_avatar right now assumes a monolithic grid. its just grabbing the sim info that the client used to get and handing it to the client
[12:14] Tara5 Oh: featuring our own Zha Ewry in RL and Philip and Adam and more
[12:15] Saijanai Kuhn: but... in a multi-grid world, either rez avatar needs more steps or there needs to be an intermediate step between logging in and rez_avatar
[12:15] Saijanai Kuhn: I think it should be an inermediate step of some kind because the client capabilities (functionality) becomes progressively more tied to the grid
[12:16] Gareth Ellison: saijanai - does the client need to talk to the destination region domain before region region host, that what you mean?
[12:16] Saijanai Kuhn: rez_avatar gets rid of ruthing but what if there are grid-specific things that are the equivalent of ruthing?
[12:17] Saijanai Kuhn: somethign like that yeah. The region domain contains references to the sims. There needs to be a step where there is contact with the region but still no commitment to one particular sim
[12:18] Scooter Back: I've been trying to understand the login steps, but I do have one question. Can't the regions be more like cell towers in the fact that data is bounced off the local cell before being rezed?
[12:18] Saijanai Kuhn: keeping it a 2-step procedure ties it into the LL model of what should happen next
[12:18] Gareth Ellison: and the region domain controller passes you off to the region host
[12:19] Saijanai Kuhn: Gareth, yeah, something like that. In the case of the current LL model, there's no need, but there might be in the future, and we don't know what some other grid's requirements are going to be
[12:19] Gareth Ellison: the domain controller should be able to reroute you to another host
[12:19] Gareth Ellison: in fact - why the **** would the agent domain have direct knowledge of region host IP?
[12:19] Saijanai Kuhn: so, while it isn't a big issue right now, I think we need to realize that there might be more steps involved then just "establish credentials" => get on the grid
[12:20] Saijanai Kuhn: Gareth, it goes back to an LL centric view of what is needed
[12:20] Gareth Ellison: auth with agent domain controller
[12:20] Gareth Ellison: watch bunny explosions and start crying gesture
[12:21] Saijanai Kuhn: we can leave that last part out when talking to Zero...
[12:21] Gareth Ellison: heh, as the initiative started with LL, it's gonna be LL-centric
[12:21] Saijanai Kuhn: sure, I'm just trying to remind folks about the long-term view. In the short term, its not needed to be implemented, but should be kept in mind as thigns are firmed up
[12:22] Gareth Ellison: would it be appropriate to mention what i'm doing with litesim now?
[12:22] Gareth Ellison: is trying to avoid the spam thing
[12:22] Saijanai Kuhn: and it can always be considered part of "rez avatar" my only worry is that there might be grid-specific CAPs available that won't fit with the 2-stage model
[12:23] Gareth Ellison: would there be any objection to me discussing briefly the litesim "supergrid"?
[12:23] Saijanai Kuhn: you can login and start doing meta-grid-wide inventory and group IM stuff without rutthing, but what about grid-specific inventory, money, IM, etc?
[12:27] Morgaine Dinova: Well in due course the agent domain will just be a proxy to the authoritative av server in your home client. After all, it's YOU who should be authoritative over your av, not some 3rd party.
[12:33] Saijanai Kuhn: Morgaine, I think Agent Domain will be the interface to your pirmary asset server. Your home computer could have assets of its own, but the grid-level assets are going to live in the grid server, as far as teh microeconomy model goes
[12:34] Morgaine Dinova: Well caching aside, the most natural model is that residence reflects ownership.
[12:35] A group: member named Clive Pro gave you A Library, Reading Room and Book, Daydream SE Islands (222, 77,.
[12:35] Saijanai Kuhn: hmmm. what do you mean by residence?
[12:36] Morgaine Dinova: Of course you can contract out residence/placement, just like you can hire storage at S3.
[12:36] Morgaine Dinova: Sai: where something resides, is located
[12:37] Saijanai Kuhn: dumb or not, it has driven SL's expansion over the past umpteen months. They didn't have 12 milllion logins just to play CSI:NY
[12:37] Morgaine Dinova: Their graph is tailing off badly, nearly a plateau.
[12:38] Saijanai Kuhn: they also don't advertise AT ALL, and their cncurrency is maxed out at less than one percent of the total account database
[12:38] Scooter Back: that means they should be forced to release to open source, so we can colaberatively expand the SL capabilities
[12:38] Morgaine Dinova: After 5 years, only 50 thousand concurrent .... that's a disastrous statistic
[12:38] Saijanai Kuhn: but concurrency is a design issue, not a popularity issue
[12:39] Saijanai Kuhn: if you look at how the concurrency graph goes, it hits teh top and the behavior of the grid goes wild and the concurrency plunges. The pattern is there 100% of the time
[12:40] Morgaine Dinova: They're intertwined. It's not just music that's logjammed. I was talking to an art gallery SL "entrepreneur" last week, and he was bottlenecked by region concurrency limits, wanted 500 minimum
[12:41] Saijanai Kuhn: sure, but the grid concurrency is the only chart I have access to. Im sure the behavior is consistent on sims as well
[12:42] Morgaine Dinova: Aye, but nobody seems to want to tackle the quest of, how come that when the alleged number of subscribers rose from 100k to 10m, the max online concurrency only rose from 10k to 50k.
[12:42] Saijanai Kuhn: you don't *ever* see a graph of concurrency in SL where it hits the top and fluctuates. Its hits the top and the system crashes (or almost) and concurrency drops 20-40%
[12:44] Morgaine Dinova: It's only hard if you refuse to think about it
[12:44] Saijanai Kuhn: Also, its being adressed gradually. The Agent/Region domain is one partial solution, at least In Zero's mind
[12:44] Saijanai Kuhn: It's Hard in terms of LL implementing something that keeps backwards compatibility
[12:44] Scooter Back: a tank can mover over tundra, snow, and swamp because of displacement. Shouldn't LL do the same thing? Decrease their foot pounds, per say?
[12:45] Saijanai Kuhn: not sure how that anology/metaphor applies
[12:45] Morgaine Dinova: Sai: well since the subject is never tackled in public, we don't really know what LL's view is.
[12:45] Gareth Ellison: apologies for my absence again......
[12:45] Scooter Back: by spreadnig out the weight of all the servers into seperate "cluster" style servers, teh load should be minimized by one individual node, but still part of the overall system
[12:46] Saijanai Kuhn: I think Zero has stateed publicly that part of the solution is the Agent/Region split
[12:46] Gareth Ellison: ok, here's the deal with the litesim supergrid - scalability is in terms of infinite regions, not infinite users per region in my current design
[12:46] Gareth Ellison: the only reason for the hybrid approach i'm using is due to simplicity
[12:46] Scooter Back: the agent/region split would essentially perform this "load balance" analagy
[12:46] Morgaine Dinova: Sai: but he hasn't stated what part of the problem is addressed by that split. The trouble is, it's not a large part of the problem.
[12:46] Saijanai Kuhn: Scooter, there are bottlenecks with the current design that need to be adressed before you can apply the multiple-node thing
[12:47] Scooter Back: if one server can only handle 50k concurent, and you have 100 servers, I believe your concurent connections would increase
[12:47] Saijanai Kuhn: There's no point with using multiple computers for a single sim if the design of the sim interface can take use of it
[12:47] Morgaine Dinova: Scooter: the sim code isn't parallelizable like that.
[12:47] Scooter Back: but Sai, wouldn't multi-node actually remove the bottlenecks?
[12:48] Saijanai Kuhn: one thing that OPenSim is doing that Zero is looking into, is splitting the script engine off from the simulator, at least for bling (attachements).
[12:49] Gareth Ellison: heh, xeon - reminds me i gotta order one.......
[12:49] Morgaine Dinova: Gareth: it's good that land acreage has been made flexible rather than 256x256m per sim, but that doesn't address even scalability, which is a matter of population density
[12:49] Saijanai Kuhn: There are likely other thigns that are on the horizon, but the big deal right now is getting agent/region domain working.
[12:50] Gareth Ellison: 256x256 is still tied into the viewer
[12:50] Saijanai Kuhn: its tied into the viewer, the implementation of the physics, the coordinate system...
[12:50] Scooter Back: so you sugest that each region be responsible for authenticating agents? or am I missing something again?
[12:50] Gareth Ellison: i'm speaking right now (after meeting is over, so as to avoid any inappropriate pseudo-spam) from my experience in starting to build a commercial grid
[12:50] Morgaine Dinova: Gareth: has anyone requested to SL that the viewer get its extent info from the sim?