September 25 Users Meeting

[09:08]  Connecting to in-world Voice Chat...
[09:08]  Connected
[09:08]  Shenlei Winkler is Online
[09:09]  Voice not available at your current location
[09:09]  Justin Clark-Casey: yes. Just flying over here I realized that I need to be in world more and go look at some of the great stuff out there
[09:09]  Justin Clark-Casey: it's an irony that I don't often do that - probably a lot less than most people
[09:09]  Mic Bowman: g'morning
[09:09]  Justin Clark-Casey: hey mic
[09:09]  Robert Adams: morning Mic
[09:10]  Mic Bowman: justin i've noticed that you could probably sleep more too
[09:10]  Mic Bowman: at least given the hours i see you on IRC
[09:10]  Mic Bowman: just us this morning?
[09:11]  Justin Clark-Casey: yeah, it looks late but to be honest I don't start until a bit later in the morning. However, occasionally I end up severely running over and pay for it the next day
[09:11]  Mic Bowman: :-)
[09:11]  Justin Clark-Casey: today my brain is being propped up by large amounts of caffeine for instance :)
[09:11]  Justin Clark-Casey: yeah, seems pretty quiet today
[09:11]  Mic Bowman: that's probably good... i'm buried
[09:12]  Mic Bowman: dan finished the optimization for pinned memory last night
[09:12]  Mic Bowman: i haven't had a chance to test it... but it removes the pinned memory used by ODE and uses the unmanaged heap instead
[09:12]  Mic Bowman: that should make the mono garbage collector MUCH happier
[09:13]  Justin Clark-Casey: excellent stuff
[09:14]  Mic Bowman: and Jim Greensky has a patch he's supposed to submit to mantis that removes the refs to llClientView so it can be garbage collected
[09:14]  Mic Bowman: otherwise those hang around forever
[09:14]  Robert Adams: we've been trying to get the 130K prim sims to work
[09:14]  Justin Clark-Casey: oooh, don't think I've heard of Jim Greensky before... (?)
[09:14]  Mic Bowman: he was an intern for me last summer & is now a "rotation engineer" working for me for the next month or two
[09:15]  Justin Clark-Casey: well, I'm mentally gearing up to propose doing 0.6.7 soon, and a regular release schedule thereafter. I'm trying to work out a way that requires minimal input (since I want some chance of running it but I'm not guaranteed to have all that much time)
[09:15]  Justin Clark-Casey: it's been quite a long time since 0.6.6
[09:15]  Mic Bowman: other than the packet pool growth... i think we can get shenlei up to 200K + prims
[09:15]  Justin Clark-Casey: that's a lot of prims
[09:16]  Robert Adams: takes a while for the viewer to load
[09:16]  Aaron Duffy is Online
[09:16]  Mic Bowman: when you send the initial update for a region with 100K prims, the packet pool starts to grow quickly
[09:16]  Justin Clark-Casey: I hear John is working on a c# -> cli compiler that's better suited to OpenSim's requirements?
[09:16]  Rob Willis: wow that took a while
[09:16]  Rob Willis: hell all
[09:16]  Justin Clark-Casey: not sure how much of that I got right
[09:16]  Justin Clark-Casey: hi rob
[09:16]  Fei Yeh is Online
[09:16]  Mic Bowman: with a couple users logging in simultaneously life gets very challenging
[09:16]  Mic Bowman: g'morning rob
[09:17]  Rob Willis: morning
[09:17]  Mic Bowman: i've turned down the bw on these regions until we figure out what the networking problems are
[09:17]  Mic Bowman: it should help to make sure you get the complete initial scene
[09:17]  Mic Bowman: hi aaron
[09:18]  Aaron Duffy: Hi... just waiting for chairs to rez :>
[09:18]  Justin Clark-Casey: hi aaron
[09:18]  Aaron Duffy: Hey Justin
[09:18]  Justin Clark-Casey: everything rezzed fairly quickly for me
[09:18]  Justin Clark-Casey: hi aaron
[09:18]  Fei Yeh: hello all
[09:18]  Mic Bowman: i think i have bw limited to 500Kbps per client
[09:18]  Mic Bowman: with the throttle set at 1
[09:19]  Aaron Duffy: I entered the region with the draw distance at 512 and it was all downhill from there
[09:19]  Mic Bowman: it actually translates into about 300Kbps
[09:19]  Mic Bowman: which takes a while
[09:19]  Justin Clark-Casey: ha, my draw distance is a measly 96m :)
[09:19]  Mic Bowman: we're reverting the scene prioritizaiton patch to help the problem...
[09:19]  Justin Clark-Casey: really need to get better graphics capabilities
[09:19]  Mic Bowman: there are dragons in the networking code that are causing some big issues
[09:19]  Shenlei Winkler is Offline
[09:19]  Justin Clark-Casey: doesn't surprise me
[09:20]  Mic Bowman: there are at least three places where updates are throttled
[09:21]  Mic Bowman: any questions for this morning?
[09:21]  Rob Willis: did you get the log i sent?
[09:22]  Mic Bowman: i did... i haven't had a chance to do anything with it... it looks like an opensim problem rather than something related to our code
[09:22]  Justin Clark-Casey: did you guys do any work on a client launcher for cable beach? Or was that just prelim discussion last week? I can't remember
[09:22]  Rob Willis: i started to look at something
[09:22]  Rob Willis: in regards to the xml
[09:22]  Mic Bowman: i don't believe john has done anything else with it recently
[09:23]  Rob Willis: for loading and saving viewers and viewer paths
[09:23]  Mic Bowman: that's great rob
[09:23]  Rob Willis: im hoping to have something running this weekend
[09:24]  Rob Willis: at least something to work from
[09:25]  Mic Bowman: ok... let me know when/if you want testers
[09:25]  Rob Willis: ok
[09:25]  Robert Adams: bats out wandering the campus
[09:25]  Justin Clark-Casey: Still working on being able to come from osgrid to here with inventory intact (presumably in a non-hg manner?)
[09:26]  Mic Bowman: yes...
[09:26]  Rob Willis: i noticed i cannot teleport there anymore, is that correct?
[09:26]  Mic Bowman: the current CB work is looking at multipel authentication for an account
[09:27]  Mic Bowman: that will make it possible to link osgrid & scisim profiles
[09:27]  Mic Bowman: right now, you can log into scisim with your osgrid ID... inventory doesn't come though
[09:27]  Rob Willis: would there be some sort of master account to group other accounts?
[09:27]  Justin Clark-Casey: all this explicit trust stuff is never going to be an Internet-scalable solution
[09:28]  Mic Bowman: http://grid.sciencesim.com/login?openid_identifier=http://osgrid.org:8002/users/Mic_Bowman
[09:28]  Mic Bowman: justin: agreed
[09:28]  Mic Bowman: two things need to happen... first, we need to support anonymous identities
[09:28]  Mic Bowman: that's something we're working on... or at least on pseudo-identities
[09:29]  Mic Bowman: the second is that you need to have some sort of transitive trust
[09:29]  Justin Clark-Casey nods
[09:29]  Justin Clark-Casey: I think ultimately a lot of stuff is going to have to move client side. All this proxying through the region seems very difficult to secure
[09:29]  Mic Bowman: the anonymity assumed by today's internet works great when there is no state carried into a server
[09:30]  Mic Bowman: vw's carry too much state to have "no trust"
[09:30]  Justin Clark-Casey: yeah, state is the major thing
[09:30]  Justin Clark-Casey: hmmm
[09:30]  Aaron Duffy: virtual world cookies?
[09:30]  Justin Clark-Casey: don't you think the need for any trust between virtual environments puts a big crimp on the network?
[09:31]  Mic Bowman: if its pairwise... certainly
[09:31]  Justin Clark-Casey: would you say that the web has the same trust structures in place?
[09:31]  Mic Bowman: however... organizations like verisign have created an indirection that works
[09:31]  Mic Bowman: what i envision is that there will be federations that ensure some level of appropriateness for their members
[09:31]  Justin Clark-Casey: yes, I think I see what you mean. Pairwise trust won't scale but trust with a comon third party migh, and might be necessary
[09:32]  Justin Clark-Casey: hmm
[09:32]  Mic Bowman: and then the federations manage trust
[09:32]  Justin Clark-Casey: might
[09:32]  Mic Bowman: yeah...
[09:32]  Justin Clark-Casey: I still not a fan of such federation though - smells of balkanization
[09:32]  Mic Bowman: so you might have an organization that will rubber stamp opensim grids that adhere to certain content protection standards
[09:33]  Mic Bowman: and then each member "trusts" anyone rubberstamped by the org
[09:33]  Robert Adams: google will be glad to federate us all
[09:33]  Mic Bowman: possibly...
[09:33]  Mic Bowman: or facebook
[09:33]  Justin Clark-Casey: would that trust restrict avatar movement or content movement?
[09:34]  Mic Bowman: possibly... but without some restrictions you can't share state safely
[09:34]  Mic Bowman: the approach we're taking is to provide some basic mechanisms so that the policies can be figured out
[09:35]  Mic Bowman: but i'd rather try it with running code
[09:35]  Justin Clark-Casey: yes. It's an interesting point - the ability to carry inventory and even rez does introduce considerations that you don't really find on the web
[09:36]  Mic Bowman: clearly... at some point we need to think about (and experiment with) ways we can improve the client to help
[09:36]  Justin Clark-Casey: hmmm, though I guess one always has to deal with people uploading inappropriate stuff to facebook/youtube/flickr, etc
[09:36]  Mic Bowman: sure...
[09:36]  Mic Bowman: and its not clear how far the parallels go
[09:37]  Mic Bowman: getting DRM to work is *Very* challenging...
[09:37]  Justin Clark-Casey: :)
[09:37]  Mic Bowman: i'd prefer to work on means to audit & let the legal system work
[09:37]  Justin Clark-Casey: yes, it's a fascinating topic
[09:37]  Mic Bowman: much easier and more flexible
[09:38]  Justin Clark-Casey: I suspect any attempt to engineer in DRM would be prone to failure anyway - I don't see it as a stable energy state
[09:39]  Mic Bowman: just to conclude on the hypergrid/cable beach discussion...
[09:39]  Mic Bowman: we're planning to build a two level trust system
[09:40]  Mic Bowman: hypergrid users will be pseudo-identified and allowed in some regions
[09:40]  Mic Bowman: and then there will be a means to use your local grid's identity (or someother identity) to fully authenticate yourself to the grid
[09:41]  Justin Clark-Casey: are you only allowing inbound hg users or outbound as well?
[09:41]  Justin Clark-Casey: by which I mean, hg users that have sciencesim as their 'home' grid
[09:41]  Mic Bowman: the "gateway" regions here are set up for hypergrid
[09:41]  Mic Bowman: but i haven't checked the outbound hypergrid recently
[09:42]  Mic Bowman: so you can hypergrid into the gateway regions
[09:42]  Mic Bowman: just not from there to the rest of scisim
[09:42]  Justin Clark-Casey: oh, okay
[09:42]  Justin Clark-Casey: just curious because of the problem of exposing the asset and inventory service access needed for hg
[09:42]  Mic Bowman: what we need is to get meerkat or something to handle the transparent relog
[09:43]  Justin Clark-Casey: Meerkat seems to be very in favour at the moment
[09:43]  Mic Bowman: they are solving the hypegrid problem in a different way
[09:44]  Mic Bowman: a much cleaner way for trust...
[09:44]  Mic Bowman: you "teleport" between grids by re-authenticating
[09:44]  Justin Clark-Casey: so more than just logging out of one grid and into another?
[09:44]  Mic Bowman: (behind the scenese)
[09:44]  Robert Adams: is Diva not working on Gridder any more?
[09:44]  Justin Clark-Casey: I'm pretty sure she is actively working on it
[09:45]  Mic Bowman: we can do a very ugly version of the same thing right now
[09:45]  Mic Bowman: or at least once we get the support for multiple inventory servers
[09:45]  Justin Clark-Casey: same thing as Meerkat or as grider?
[09:45]  Mic Bowman: meerkat...
[09:45]  Mic Bowman: bascially using the -multiple
[09:45]  Justin Clark-Casey: so they are doing more than just a relog?
[09:46]  Mic Bowman: no...
[09:46]  Justin Clark-Casey: rex were doing that more than a year ago, but not with such a slick interface
[09:46]  Mic Bowman: that's all they do... but they save passwords
[09:46]  Mic Bowman: correct
[09:46]  Mic Bowman: and i expect naali will ahve the support built in
[09:46]  Justin Clark-Casey: that's such a horrible hack
[09:47]  Mic Bowman: and hypergrid isn't?
[09:47]  Mic Bowman: architecturally...
[09:47]  Justin Clark-Casey: :) I think hg is more interseting but with fatal security flaws
[09:47]  Justin Clark-Casey: hence the need for gridder, as we know
[09:47]  Mic Bowman: the plan is to let you use your osgrid (or whatever openid provider you have) to re-authenticate
[09:47]  Justin Clark-Casey: hg is a cool hack ;)
[09:48]  Mic Bowman: with client cached tokens it can be done completely transparently
[09:48]  Justin Clark-Casey: I see. So the next step to obviate the need for saving passwords
[09:48]  Mic Bowman: chat lag?
[09:48]  Robert Adams: shared identity so you don't have to log into every grid you visit. Authentication happens automaticaly and invisibily
[09:49]  Mic Bowman: correct robert
[09:49]  Justin Clark-Casey: I think that path also naturally pushes services onto the client path rather than via the vw
[09:49]  Mic Bowman: yes... the client has to manage the authentication tokens and hand them out
[09:49]  Mic Bowman: think cookies for the web
[09:49]  Mic Bowman: in fact, john's current implementation is more or less that
[09:50]  Mic Bowman: if you can close your eyes to the ugliness... i just click links on the web to teleport between grids
[09:50]  Mic Bowman: same identity every where (and eventually the same inventory)
[09:50]  Justin Clark-Casey: yes. Identity is more than just a name thoguh - it's a profile too. Can that come up via openID too?
[09:51]  Mic Bowman: yes...
[09:51]  Mic Bowman: the profile should reference a collection of services
[09:52]  Justin Clark-Casey: one eventually has to work around the limitation of the LL protocol to Guid identifiers rather than uris or similar
[09:52]  Mic Bowman: if we're willing to fix the protocol... then there's LOTS of better ways to work this
[09:53]  Justin Clark-Casey: yes
[09:53]  Justin Clark-Casey: I agree
[09:53]  Mic Bowman: back to the need for a viewer we can really hack on
[09:53]  Mic Bowman: don't suppose looking glass is close?
[09:53]  Justin Clark-Casey: indeed :)
[09:54]  Justin Clark-Casey: not the sun project I presume
[09:54]  Mic Bowman: robert? care to comment?
[09:54]  Robert Adams: no, a viewer in C# that uses Ogre
[09:55]  Robert Adams: I'm integrating it into Radegast so one has chat and inventory ,etc
[09:55]  Justin Clark-Casey: lookingglassviewer.org?
[09:55]  Robert Adams: that's the one
[09:55]  Justin Clark-Casey: hmm, it's in my list of bookmarked viewers but I must have forgotton it :)
[09:55]  Robert Adams: lots of optimizations needed, but it getting to the point that one can hack on it to do new things
[09:56]  Justin Clark-Casey: you're the lead dev, rob?
[09:56]  Robert Adams: yes, it''s my project
[09:56]  Justin Clark-Casey: ahhh, I remember your ml posts now... :)
[09:56]  Justin Clark-Casey: sorry, bit slow today
[09:56]  Robert Adams: ref the earlier discussion about getting sleep :-)
[09:57]  Justin Clark-Casey: indeed :)
[09:57]  Robert Adams: but the goal is to get enough of a viewer that one can play with protocols, etc
[09:58]  Justin Clark-Casey: scene screenshots look pretty good already
[09:58]  Robert Adams: I've been thinking of building a new IClient to talk to LG
[09:58]  Robert Adams: I'm adding SkyX and Hydrax shortly (Ogre addins) that will give realistic sky and water
[09:59]  Robert Adams: but prim generation, placement and texturing is working
[09:59]  Robert Adams: there are the usual performance problems -- LL prims make for lots of graphics batches
[10:00]  Justin Clark-Casey: any effect of using c#? Or does it not matter since all the heavy lifting is down in the ogre libs anyway?
[10:01]  Robert Adams: the design is to use C# for comm and UI and world management and let the C++ code do the rendering
[10:02]  Robert Adams: so the C# is not involved in the rendering at all other than giving the mesh info and update instructions
[10:02]  Justin Clark-Casey: do you end up writing any c++?
[10:02]  Robert Adams: yes, I have a C++ interface between my managed code and Ogre
[10:02]  Robert Adams: I also had to figure out how to dynamically load content into Ogre so a lot of code makes that happen
[10:03]  Justin Clark-Casey: cool. I hear that one of the chief Ogre guys is quite interested in OpenSim
[10:03]  Justin Clark-Casey: Guy living on Jersey?
[10:03]  Robert Adams: that's a problem at the moment since prebuild doesn't handle C++ and there are both 32 bit and 64 bit systems
[10:03]  Robert Adams: I don't know of him
[10:04]  Robert Adams: I'll put something in the Ogre showcase and see what happens
[10:04]  Justin Clark-Casey: sorry I don't know the name, I just heard that in passing. But I know at least that one guy is aware of (and I think interseted) in it
[10:04]  Justin Clark-Casey: so you could well get a good response
[10:05]  Robert Adams: and I need someone who knows graphic shaders... not an area I know
[10:05]  Justin Clark-Casey: good graphics people always seems to be a bottleneck - relatively more people available to write all the easy server code :)
[10:06]  Robert Adams: for sure, but these days one does a lot of any graphics work in the shaders
[10:07]  Justin Clark-Casey: interesting. Anyway, I need to go. Thanks for letting me witter on. This is one of the effects that caffeine has on me
[10:07]  Robert Adams: bye Justin
[10:07]  Aaron Duffy: later
[10:07]  Justin Clark-Casey: bye rob, aaron, mic, folks. I'll leave with the sinking in the floor feeling
[10:08]  Mic Bowman: bye all
[10:08]  Robert Adams: LOL
 
 
 
 
users/sep25.txt · Last modified: 2009/09/25 10:06 by cmickeyb
 
RSS - 2008 © ScienceSim