September 18 Users Meeting

[09:00]  Fei Yeh: Ooo movie night. :)
[09:00]  Mic Bowman: or something like that
[09:00]  Mic Bowman: i was trying to get the web content on a prim working
[09:00]  Mic Bowman: you need to turn on the automimetype detection in your viewer
[09:01]  Kite Hallison: hello
[09:01]  Mic Bowman: hi kite
[09:01]  Aaron Duffy: Hi Kite
[09:01]  Kite Hallison: Hell Mic
[09:01]  Kite Hallison: hello Aaron
[09:01]  Fei Yeh is Offline
[09:02]  Mic Bowman: hi aaron
[09:02]  Aaron Duffy: hello again
[09:03]  Mic Bowman: hi justin
[09:03]  Justin Clark-Casey: Hi Mic, hello folks
[09:03]  Aaron Duffy: Hi Justin
[09:03]  Kite Hallison: Hello justin
[09:04]  Justin Clark-Casey: One day I'm going to have to get a better computer so that the sound of the fire isn't drowned out by my laptop fans whirring away :)
[09:04]  Justin Clark-Casey: /breaks the fourth wall
[09:04]  Mic Bowman: which is hotter?
[09:04]  Justin Clark-Casey: ha. it's a tossup
[09:05]  Mic Bowman: there have been a bunch of instabilities recently
[09:05]  Mic Bowman: let me just go through the three things thatwe're working on...
[09:05]  Mic Bowman: first... there are a number of improvements for cable beach
[09:06]  Mic Bowman: this includes work on authorizing inventory services so you can bring your own inventory into the world
[09:06]  Mic Bowman: it also include more robust support for specifying the openid
[09:06]  Mic Bowman: so... you could set up a login using your osgrid account like this:
[09:07]  Mic Bowman: http://grid.sciencesim.com/login?openid_identifier=http://osgrid.org:8002/users/Mic_Bowman
[09:07]  Mic Bowman: you can bookmark that link and use it to login
[09:07]  Mic Bowman: eventually (in a few weeks) you'll be able to bring in your osgrid inventory securely that way
[09:08]  Justin Clark-Casey: a one time transfer or dynamic access?
[09:08]  Mic Bowman: the second big change that has been causing problems is that we are tryi8ng a bunch of patches to reduce memory
[09:08]  Mic Bowman: justin, dynamic access
[09:08]  Mic Bowman: authenticated each time you come in
[09:08]  Justin Clark-Casey: I take it this requires explicit trust from osgrid to sciencesim?
[09:08]  Mic Bowman: correct
[09:09]  Mic Bowman: on the memory side...
[09:09]  Rob Willis: morning all
[09:09]  Justin Clark-Casey: hi rob
[09:09]  Mic Bowman: there are two big patches we're testing
[09:09]  Mic Bowman: well... the patches aren't that big but they potentially change important things
[09:09]  Mic Bowman: the first... we've found that opensim does not release memory when a user logs out...
[09:10]  Mic Bowman: for reasonably small regions.. the amount of memory leaked is relatively small... but for complex regions, it can be as much as 100's of megabytes
[09:11]  Mic Bowman: the second area we're working to fix is optimizing the memory the physics engine uses
[09:11]  Mic Bowman: we've seen pretty significant reductions by using a mesh "dictionary" to compress the information that is shared with the physics engine
[09:12]  Mic Bowman: chamomile region was running about 2.6G of memory (beforeany user connects) and it is now running around 1.5G
[09:12]  Robert Adams: the collision meshes add up when you have 100K+ prims
[09:12]  Mic Bowman: so about 40% reduction for that region
[09:12]  Mic Bowman: the benefits depend on the number of meshes that are common
[09:13]  Justin Clark-Casey: that's all very nice. Have all the previous patches by Dan been incorporated?
[09:13]  Mic Bowman: 2 of 3 i think
[09:13]  Mic Bowman: and the 3rd was "optional"
[09:13]  Justin Clark-Casey: I'm hoping that now that jhurliman is in core there will be less opportunity for bottleneck
[09:13]  Mic Bowman: i think john was going to look at one more as well
[09:14]  Mic Bowman: the llclientview and ode patches will be submitted today or monday
[09:14]  Mic Bowman: there are two more memory patches coming
[09:14]  Mic Bowman: the third (and biggest) area of instability is in the scene update prioritizaiton
[09:14]  Mic Bowman: when you log into a region, the prims are sent to you in what is not quite but almost random order
[09:15]  Justin Clark-Casey: ha
[09:15]  Mic Bowman: (there is some ordering... but in practice the indeterminancy of the avatars position almost always queuest them incorrectly)
[09:15]  Rob Willis: yeah i still cant see alot of things here
[09:15]  Mic Bowman: the patch will prioritize based on current position and update the priorites when your position changes
[09:16]  Mic Bowman: the code... however (as rob noticed) is broken when too many avatars enter a region
[09:16]  Mic Bowman: packets start to get dropped at (abnormally) high rates
[09:16]  Mic Bowman: and scene updates are not reliable
[09:17]  Mic Bowman: so... the sum of it is 1) we're turning off the scene prioritization in most regions and 2) bear with us on the CB changes (please)
[09:18]  Mic Bowman: and when the memory patches are all submitted to opensim, you'll be able to double the number of prims in a region for the same memory
[09:18]  Rob Willis: an interesting thing is the two clocks here appear and then disappear randomly
[09:18]  J.H Chen is Offline
[09:18]  Justin Clark-Casey: are any of these patches going to be accompanaied by unit tests, if they are unit testable? Without them there is a high opportunity for very hard to detect and fix regrressions
[09:19]  Mic Bowman: you're getting some updates for the hands of the clock as they more
[09:19]  Mic Bowman: justin, none of the behavior should change for the memory tests...
[09:19]  Mic Bowman: but yes, we'll try to provide specific tests for the dictionary
[09:19]  Justin Clark-Casey: ok, thanks :)
[09:20]  Mic Bowman: though the ODE unit tests should catch any problems given the level where the patching is occuring
[09:21]  Justin Clark-Casey: just fyi, I'm thinknig next week of proposing a fixed release schedule, with release 0.6.7 coming fairly soon and subsequent releases maybe every 3 months
[09:22]  Mic Bowman: that would be good...
[09:22]  Justin Clark-Casey: this probably won't help much with stability of releases in the short term but it would be more predictable, esp. if people adjust their dev schedules accordingly
[09:22]  Mic Bowman: any other topics for this morning?
[09:23]  Lightswitch: lights on
[09:23]  Lightswitch: lights off
[09:23]  Justin Clark-Casey grins
[09:23]  Mic Bowman: has anyone tried the megaregion code yet?
[09:23]  Robert Adams: we're awake
[09:24]  Aaron Duffy: I tried but it didn't seem to work. I think I'm not far enough up the trunk
[09:24]  Mic Bowman: i was going to try on the yellowstone regions
[09:24]  Mic Bowman: the "load in order" is more of a problem since i don't use an ini file for region info
[09:25]  Mic Bowman: which i suppose means i need to figure out a naming scheme for my region.xml files
[09:25]  Aaron Duffy: the ini replaces the xml doesn't it?
[09:25]  Mic Bowman: yes...
[09:26]  Justin Clark-Casey: afaik they are both still valid ways of configuring
[09:26]  Mic Bowman: the xml files is how we manage all of our regions
[09:26]  Justin Clark-Casey: any deprecation schedule is unclear - I suspect both will continue to remain valid as all the work of parsing them is done by the nini library
[09:27]  Mic Bowman: much easier to change the mapping of regions to simulatiors
[09:27]  Justin Clark-Casey: I wish everything in that file wasn't an attribute though
[09:27]  Mic Bowman: i would prefer to not have the file re-written
[09:27]  Robert Adams: and since the xml was always writtten out with LFs, the ini files are easier to read
[09:27]  Aaron Duffy: I normally use xml, but I set up two new regions with an ini for the megaregion. They start up and run fine, but I still "blip" when crossiing between them so I don't think they are a megaregion
[09:28]  Justin Clark-Casey: doesn't sound like it
[09:28]  Mic Bowman: justin, do megaregions have to be square?
[09:28]  Aaron Duffy: I read they could be any shape... even have gaps
[09:28]  Justin Clark-Casey: that's a good question. I don't know for sure - I suspect not though
[09:29]  Justin Clark-Casey: yeah, I think that's right
[09:29]  Mic Bowman: the gaps would be nice...
[09:29]  Justin Clark-Casey: as long as they are contiguous I think
[09:29]  Aaron Duffy: hmmm... thinking about setting up a grid sized chess board
[09:29]  Mic Bowman: one thing we could do for scisim users is to start proviing an ini file for the full 5x5 region
[09:30]  Justin Clark-Casey: :D Monster chess pieces - I like how you think
[09:30]  Mic Bowman: set up as a megaregion
[09:30]  Mic Bowman: then as a region gets busy enough, you can move it into its own simulator
[09:30]  Mic Bowman: with the megaregion, there is very little overhead for having an all ocean region
[09:30]  Aaron Duffy: that would be nice. I saw one of the opensim dev's is providing a standalone release that defaults to a 512x512 region
[09:31]  Robert Adams: and when the performance patches are in a region can have 100K prims... makes a megaregion with stuff on it very useful
[09:31]  Mic Bowman: 5x5 plaza in one big region
[09:32]  Aaron Duffy: Justin... are oars working on megaregions?
[09:33]  Justin Clark-Casey: I heard a report that they do, though I haven't had confirmation
[09:33]  Doug Osborn is Online
[09:33]  Mic Bowman: on a per region basis or across the whole megaregion
[09:34]  Justin Clark-Casey: I can imagine that they would - there's nothing in the oar module that ties it into a fixed size region
[09:34]  Justin Clark-Casey: depends how the region is being modelled in the server I think
[09:34]  Mic Bowman: terrain?
[09:34]  Justin Clark-Casey: whether it's regarded as one giant region or still lots of 256 ones
[09:35]  Justin Clark-Casey: yes, that's the bigger question. I heard that it works though the source isn't very unreliable
[09:35]  Justin Clark-Casey: not had an opportunity yet to test myself
[09:35]  Justin Clark-Casey: the source being the person in this case
[09:35]  Justin Clark-Casey: er, isn't very reliable I mean
[09:35]  Doug Osborn: aloha all
[09:36]  Mic Bowman: hi doug
[09:36]  Mic Bowman: i thought you were out today?
[09:36]  Doug Osborn: i thopught so to
[09:36]  Justin Clark-Casey: hi doug
[09:36]  Doug Osborn: hey justin :)
[09:36]  Doug Osborn: i heard you talking about the megarregion stuff?
[09:36]  Mic Bowman: yes
[09:37]  Mic Bowman: brb
[09:39]  Mic Bowman: back...
[09:39]  Justin Clark-Casey: looking very grey today, doug
[09:39]  Mic Bowman: i'm looking for some feedback on a usage...
[09:40]  Doug Osborn: i've had some texture issues lately....
[09:40]  Doug Osborn: and my 3 browsers show mw different stuff.....
[09:40]  Mic Bowman: maybe this is a question for rob... how many of your studentys have facebook accounts?
[09:41]  Mic Bowman: do you use fb for any instructional support?
[09:42]  Aaron Duffy: Here we don't use FB academically... I'm discouraged from "friending" my students. Our students' email is all gmail-based though
[09:42]  Simple Clock: The time is 9:37
[09:42]  Mic Bowman: so google accounts would be better?
[09:42]  Lightswitch: lights on
[09:43]  Mic Bowman: i need to figure out if the google authentication can be passed in
[09:43]  Mic Bowman: and if anyone wants to help with the cable beach launcher...
[09:44]  Mic Bowman: doug? when was the last time you wrote C#?
[09:44]  Doug Osborn: I was hoping to start working with the launcher last week Mic but I got sidtracked
[09:44]  Mic Bowman: :-)
[09:44]  Rob Willis: it is similar to another launcher project
[09:44]  Mic Bowman: which one?
[09:44]  Rob Willis: let me check
[09:45]  Mic Bowman: the rezzme launcher is too hard to get working (too many dependencies caused by python)
[09:45]  Mic Bowman: rezzable supposedly has one for heritage grid, but i haven't found the code
[09:46]  Rob Willis: there is a project i saw called opensim launcher
[09:46]  Rob Willis: i tried it and it waas buggy
[09:46]  Rob Willis: uses a file for the paths
[09:46]  Mic Bowman: the two features we want for the CB launcher are 1) better support for multiple viewers and 2) the ability to pass in a location
[09:46]  Mic Bowman: the second of those is relatively easy.. the first is harder
[09:46]  Rob Willis: however everytime you use it, it rewrites over the settings file
[09:47]  Rob Willis: so you have to recreate that file
[09:47]  Mic Bowman: does the meerkat viewer have support for multiple instances? they have included "transparent" grid-to-grid teleports... but i don't know how to drive it externally
[09:48]  Rob Willis: it is an xml file and can be configured from a gui,
[09:48]  Mic Bowman: do you have alink?
[09:49]  Rob Willis: http://lab.newworldgrid.com/index.php/OpenSim_Launcher
[09:49]  Mic Bowman: thanks
[09:50]  Rob Willis: not sure if the source is available, i remember wanting to fix it, but couldnt
[09:51]  Mic Bowman: i think i tried that one also
[09:52]  Justin Clark-Casey: btw Mic, do you publish the transcript of this meeting?
[09:52]  Mic Bowman: i don't
[09:52]  Mic Bowman: we could certainly start doing that...
[09:52]  Rob Willis: that would be great
[09:52]  Justin Clark-Casey: it would be good for reference
[09:53]  Mic Bowman: i'll grab today's transcript & put it on the wiki
[09:53]  Mic Bowman: i suppose the next thing to do is to create a logger bot...
[09:54]  Mic Bowman: or maybe we can use aaron's twitter feed
[09:54]  Mic Bowman: :-)
[09:54]  Justin Clark-Casey: yes, one that auto-publishes
[09:54]  Justin Clark-Casey: I'm sure one must exist in SL already
[09:54]  Aaron Duffy: Can we name it BigBroBot?
[09:54]  Justin Clark-Casey: heh, there is a twitter opensim module out there somewhere
[09:54]  Mic Bowman: as in a module that publishes twitter feeds in world? or the other way?
[09:54]  Justin Clark-Casey: I know world -> twitter, not sure about reverse
[09:55]  Aaron Duffy: It could easily be done both ways. that is one bloated twitter client though
[09:55]  Mic Bowman: :-)
[09:55]  Justin Clark-Casey: ha ha ha ha
[09:56]  Mic Bowman: personally... i think opensim is a great IRC client
[09:56]  Justin Clark-Casey: if we each had an associated twitter account we could just spam this entire discussion to twitter - that would be scary
[09:56]  Justin Clark-Casey: really? I think there's a lack of functionality for that :)
[09:56]  Robert Adams: the commercial, youth virtual worlds should have SMS connectivity. After watching my younguns stay connected with theeir friends no matter where they are
[09:57]  Robert Adams: spam to gtalk if everyone is using google accounts
[09:58]  Mic Bowman: i got about half way into some additions to the IRC module that would let it connect to multiple channels & bridge them to different in world channels...
[09:58]  Mic Bowman: decided i had better things to do with my evenings...
[09:58]  Justin Clark-Casey: :)
[09:58]  Mic Bowman: any other random topics for today?
[09:59]  Doug Osborn: ummmm can we get out prim size unvcapped?
[09:59]  Mic Bowman: i think i have the prim size on all these regions set at 256m
[09:59]  Mic Bowman: you want bigger?
[09:59]  Doug Osborn: the cable beach regions i can build in seem to have the config file limit set back to 10M
[10:00]  Doug Osborn: 256 is fine :)
[10:00]  Mic Bowman: i'll check to see if i screwed up something in the settings
[10:00]  Aaron Duffy: with megaregions that is going to start feeling tiny at 256M
[10:01]  Mic Bowman: (or if the settings just changed the default without documentation)
[10:01]  Justin Clark-Casey: not afair
[10:01]  Mic Bowman: :-)
[10:01]  Doug Osborn: lol
[10:01]  Mic Bowman: not that is ever happened before...
[10:01]  Doug Osborn: when i see a megatregion I'll think about bigger prims....
[10:02]  Justin Clark-Casey: well true :)
[10:02]  Doug Osborn: but i think we are planning very small avatars and boats for our 9 region ocean space :)
[10:03]  Aaron Duffy: really the relative size of our avatars and the little "m" next to the measurements are the only things defining size here
[10:03]  Rob Willis: when changes are made to opensim svn are they added to the sciencesim straight away?
[10:04]  Mic Bowman: no
[10:05]  Doug Osborn: yes Mic how soon are we going to be able to test justins Iar transfers to Sciencesim?
[10:05]  Rob Willis: so llRot2Euler() is not working properly in sciencesim?
[10:05]  Mic Bowman: we have been keeping more closely in sync over the last couple weeks because the cable beach code is using some of the more recent opensim features
[10:05]  Rob Willis: ok
[10:05]  Mic Bowman: i believe we just did a merge yesterday...
[10:05]  Mic Bowman: we need to rebuild the binaries & resync the git repository
[10:06]  Mic Bowman: in general i prefer to stay off the most recent versions in order to find a version that is more stable
[10:06]  Aaron Duffy: I've been reading about remote tracking branches and cherry picking using git. I think that is a way individuals could update their cloned sciencesim repo from the opensim repo
[10:06]  Mic Bowman: stability has suffered for other reasons though
[10:06]  Rob Willis: k
[10:06]  Mic Bowman: for some of our internal regions, i've been just applying patches like what aaron describes
[10:06]  Justin Clark-Casey: Mic: The other reasons you talked about earlier or something else?
[10:07]  Mic Bowman: the reasons we discussed eaarlier
[10:07]  Mic Bowman: lots of experimental code
[10:07]  Mic Bowman: trying to fix other scalability problems
[10:07]  Justin Clark-Casey: your code?
[10:07]  Mic Bowman: yes
[10:08]  Justin Clark-Casey: ahhhh, so it's you guys fault! I knew it :)
[10:08]  Mic Bowman: well... i'll start with the assumption that its my fault
[10:08]  Mic Bowman: :-)
[10:08]  Justin Clark-Casey: that's very humble of you
[10:08]  Mic Bowman: and rather realistic
[10:08]  Mic Bowman: i'm very good at breaking things
[10:09]  Justin Clark-Casey: yes, well the whole edifice isn't the most stable of things to start off with
[10:09]  Justin Clark-Casey: I don't think we're going to be removing that alpha tag any time too soon
[10:09]  Mic Bowman: well... its not too bad either
[10:09]  Mic Bowman: considering the complexity
[10:10]  Mic Bowman: ok all... i need to head off to the next meeting
[10:10]  Justin Clark-Casey: I'll accept that judgement - I don't have all that many reference points for this complexity level
[10:11]  Mic Bowman: transcript will be posted here: http://www.sciencesim.com/wiki/doku.php/users/sep18
[10:11]  Justin Clark-Casey: cool, thanks mic
 
 
 
 
users/sep18.txt · Last modified: 2009/09/18 10:11 by cmickeyb
 
RSS - 2008 © ScienceSim