Chat log from the meeting on 2007-10-16

From OpenSimulator

Revision as of 12:10, 16 October 2007 by Tleiades (Talk | contribs)

Jump to: navigation, search

[10:52] Neas Bade: hey Tleiades
[10:52] You: brb
[10:52] nebadonb izumi: thats wierd daedius
[10:53] Neas Bade: hey ckrinke
[10:53] Charles Krinkeb: Hello all
[10:53] Daedius Moskvitch: I'll try to get a more detailed log for a bug later
[10:53] Neas Bade: do we want to go upstairs for the meeting? Give people a chance to experience physics by wandering up the stairs?
[10:53] Daedius Moskvitch: Hehe
[10:54] Charles Krinkeb: Coitainly
[10:54] Neas Bade: hey all come joing up here
[10:54] nebadonb izumi: nice
[10:54] Charles Krinkeb: Sorry about the lack chairs.
[10:55] Neas Bade: well, we still don't have sitting working yet do we?
[10:55] nebadonb izumi: yea the chair's are being delivered next week
[10:55] nebadonb izumi: :D
[10:55] Charles Krinkeb: no. That was a sideways comment.
[10:55] Neas Bade: hehehe
[10:55] Daedius Moskvitch: ;)
[10:55] Neas Bade: backordered eh?
[10:55] nebadonb izumi: lol
[10:55] Charles Krinkeb: But, hey, the flags wave.
[10:56] You: back
[10:56] Charles Krinkeb: neas. One of our thoughts was to build out from the island towards a corner with both a welcome area with prims explaining OpenSim and a place for sim owners to describe their sims.
[10:56] Neas Bade: come up stairs, it's pretty nice up here
[10:57] Charles Krinkeb: Darok: Come up the stairs, please.
[10:58] Daedius Moskvitch: I like the idea of making this a hub of information =)
[10:58] Charles Krinkeb: He may be chatting with DanXor on #opensim-dev
[10:58] Neas Bade: ckrinke, Tedd needs some help being able to log in
[10:59] Neas Bade: can you help him with that?
[11:00] You: I like this place, kind of cool
[11:00] Charles Krinkeb: I tried. The only two choices are resetting home location or creating a new user with "Wright Plaza" as a home location.
[11:00] You: I like this place, kind of cool
[11:01] Charles Krinkeb: The libopenjpeg-libsl is throwing off an exception every 5 seconds, am I am holding my breath about the physics right now worrying over the console on another screen.
[11:01] nebadonb izumi: thats not physics related
[11:01] You: the sim can run pretty stable with that exception
[11:02] nebadonb izumi: someone probably tried to look at the map
[11:02] You: I see that quite a lot on my dev machine
[11:02] Charles Krinkeb: It appears to be a glibc_2.4 vs glibs_2.3 issue which may be related to the ode not found issue on some other servers.
[11:04] You: I never really understood why LL decided to use j2k
[11:04] Charles Krinkeb: Daedius: Even though we are just avatars in here, I feel like I am meeting you and creating a sense of community.
[11:04] Daedius Moskvitch: Oh yes, I completely agree Charles =)
[11:04] You: Tedd... great to see you here
[11:04] Daedius Moskvitch: Its very nice to finally meet some of you all here
[11:05] You: that is one of the magical things about SL
[11:05] Tedd Maa: where is my SL inventory?
[11:05] You: lol ...
[11:05] Charles Krinkeb: Hey. Tedd is here. Great.
[11:05] Tedd Maa: *n00b PSYKE*
[11:05] Tedd Maa: ;)
[11:05] You: easy .. I'm working on that :)
[11:05] Tedd Maa: it worked creating a new user
[11:06] Charles Krinkeb: Just to set a tone for the meeting, I think one of our key issues in the weeks to come is going to be balancing finishing things partly done versus adding new features.
[11:06] You: Tedd, I read your wiki entry on scripting for the architecture group
[11:06] You: interesting reading
[11:07] Tedd Maa: I need to make another version of it, a few things thats missing and a few things people misunderstood
[11:07] You: I think it is important that we talk about the mailing list
[11:07] Tedd Maa: thats what happens when you write document and watch exciting serial killer on TV at the same time
[11:08] You: it is pretty unstable, and it is the only reliable lasting communication we have
[11:08] Charles Krinkeb: To put it all in perspective: Even though this place looks very advanced, it is prone to crashing with recent changes and is just barely hanging together.
[11:09] Tedd Maa: run it in Debug mode so we know what line that crashed and what brought us there ... debug crash error messages are very useful
[11:09] Tedd Maa: release compiles are not so much
[11:09] Neas Bade: Tleiades have you had mail not got through?
[11:09] You: most of this week the mailing list was dead, atleast for me
[11:09] Neas Bade: there was a hang up for a couple of days, but that all seemed to flush
[11:09] nebadonb izumi: yea i didnt see many mails this week
[11:09] Tedd Maa: but why is mailinglist difficult now?
[11:09] Tedd Maa: should we consider google?
[11:10] Neas Bade: I don't think people are posting to the list that much
[11:10] Charles Krinkeb: We should encourage more posting to the list as one-line ideas are tough to express.
[11:10] You: no, and an unreliable list makes it even more unlikely that people will use it
[11:10] You: but IRC isn't good for everything
[11:11] Neas Bade: right, I want to know if there is some issue that happened other than the delay last week
[11:11] Tedd Maa: lately some people has been a bit busy, my self included, so the general ml/commit has gone down a bit... from experience this is just a relatively short period
[11:11] You: it is the second time, I have had delays in the flow
[11:12] Neas Bade: well, adam did fix the dns blacklisting for the box
[11:12] Neas Bade: which means we can move mail over to there instead of berlios
[11:12] Neas Bade: if people believe that is still an issue
[11:12] Tedd Maa: dns blacklisting/mail server I have some experience with
[11:13] Neas Bade: I didn't do it before when we setup the box because it was in a relay black list, which meant lots of people wouldn't get the email
[11:13] Neas Bade: but that is better now
[11:14] You: Well, the important thing is that the list works, I think that from now on, I will shoot of a test mail every day
[11:14] nebadonb izumi: hello everyone
[11:14] You: so, if it doesn't show up, you'll know it is missing
[11:14] You: hello nebadonb
[11:15] Charles Krinkeb: hello, Mw
[11:15] Neas Bade: tleiades is this the Nwc.XmlRpc post, or is there one after that?
[11:16] You: the Nwc.XmlRpc post, got stuck in the flow
[11:16] Neas Bade: it came through fine for me
[11:16] Wright Juran: hi
[11:16] You: a few hours ago, I recevided a couple of mails from the list, but I knew from the web-page on berlios that they were there
[11:16] Charles Krinkeb: Hi, Chill. Welcome.
[11:17] chill ken: hello
[11:18] Neas Bade: it looks like the hotmail -> berlios handoff saw some delay
[11:19] Neas Bade: but it's hard to figure out which side muffed that
[11:19] Neas Bade: my email from yesterday got through in about 15 minutes
[11:19] You: ok.. I'll try an switch to gmail, maybe that will make tings go better
[11:19] Tedd Maa: Looks like a gym club in here
[11:19] Neas Bade: Inventory -> Create -> Body :)
[11:20] Neas Bade: Tleiades, if this is still a problem in a week, I'd suggest we move the mailing lists to the box
[11:20] You: while I agree with Charles that we need to focus on fixing bug, I also think that we need to push forward on adding more functionality
[11:20] Neas Bade: that will let us control grey listing and spam measures there, and hopefully debug any other issues that happen
[11:21] You: Ok.. I'll setup a gmail account
[11:21] Charles Krinkeb: neas: Are you setting the agenda?
[11:21] Neas Bade: so, what's our top bugs list?
[11:21] Neas Bade: I didn't think so, but I can if no one else wants to :)
[11:21] nebadonb izumi: Well i thik its vital that Grid mode be brought inline with Standalone
[11:22] nebadonb izumi: let the gap widen too much more could cause even more bugs down the line
[11:22] Neas Bade: nebadonb, agree, but that is a larger scale issue
[11:22] Neas Bade: hard to call that a single bug :)
[11:22] Neas Bade: so what are the discreet things we need to get there?
[11:22] You: I'm quite a good ways along with inventory and assets
[11:23] Neas Bade: Tleiades: great!
[11:23] You: those are two major differences
[11:23] nebadonb izumi: yea thats the biggest gap right now i think that will go along way Tleides
[11:23] nebadonb izumi: well one Grid mode bug
[11:23] nebadonb izumi: is unstable region crossings
[11:24] nebadonb izumi: and also if a Region or Sim goes offline in an unfriendly fashinon, it adversly effects the neighboring sim/regions
[11:24] You: and problems logging into a sim, if the neighbors are down
[11:24] nebadonb izumi: yea
[11:24] Charles Krinkeb: neas: I think the remoting error issue is pretty high. I put the ones I thought were the top few (at least the ones I worry about the most) on the Office_Hours page.
[11:25] Neas Bade: ckrinke, great, let me bring that up
[11:25] Tedd Maa: I'm sort of designing a binary remote procedure call protocol for script engine
[11:25] nebadonb izumi: yea i think overall that is the biggest Grid mode bug at the moment
[11:25] Tedd Maa: not sure what RPC things are needed to rhe rest, but ...
[11:25] You: we need to talk about comms protocols
[11:25] Charles Krinkeb: I also was unable to update the grid from r2060 to r2090 as the GridServer locked up twice when run (no exception, but the process had to be killed [twice])
[11:25] Tedd Maa: *needed for the rest
[11:26] Neas Bade: so on the grid crossing issues it seems like we've got at least 2 bugs to dig into
[11:27] nebadonb izumi: yea
[11:27] Neas Bade: 1) gracefully handling the downing of adjacent regions
[11:27] nebadonb izumi: i think thats top priority
[11:27] You: and some serious latency issues in the protocol
[11:27] Neas Bade: which means figuring out a sane way of invalidating the regionhandles I think
[11:27] nebadonb izumi: yea if a sim goes offline gracefully its not an issue
[11:28] Charles Krinkeb: Yes. The regionHandles are the issue in the database. Invalidating their use would help immensenly
[11:28] nebadonb izumi: but if it crashes its big trouble
[11:28] Neas Bade: 2) a more generic "figure out what is borking the crossing" issue
[11:28] You: I have some code, almost ready for that
[11:28] Neas Bade: which probably requires some debug tracing put into place
[11:28] You: charles got me sidetracked, on the inventory asset thing
[11:28] Neas Bade: tleiades, great! (again)
[11:28] nebadonb izumi: oh and ALso neas
[11:28] nebadonb izumi: in Windows
[11:29] nebadonb izumi: in grid mode sims running Windows
[11:29] nebadonb izumi: there is no border crossing what so ever
[11:29] nebadonb izumi: it only works in Linux at this point
[11:29] nebadonb izumi: Linux to Linux
[11:29] Neas Bade: pshaw, who needs windows anyway ;)
[11:29] Neas Bade ducks
[11:29] nebadonb izumi: and variation of throwing a windows server in the mix it doenst work
[11:29] nebadonb izumi: LOL
[11:29] Daedius Moskvitch: =)
[11:29] Neas Bade: do we have bug numbers for these bugs?
[11:30] Neas Bade: i.e. have they been reported yet?
[11:30] Charles Krinkeb: GridServer has a method to check for live sims called "CheckSims" that is unused. Perhaps we should enable it before crossing or throwing a remoting exception?
[11:30] nebadonb izumi: I beleive there are Mantis #'s off the top fo my head im not sure what they are
[11:30] Charles Krinkeb: Yes, they are all in Mantis.
[11:30] You: charles.. that method is dirt old, and must be completely rewritten
[11:30] You: hence my mail on the list regarding xmlrpc
[11:31] Neas Bade: can someone find them and give them to me? I think I'll create a Top 10 bugs list on the wiki, so we can get people focussed on some of the big bugs
[11:31] You: and my item on the agenda regarding comms protocols
[11:31] Charles Krinkeb: Mantis#449 and #461 are pertinent.
[11:31] Neas Bade: I need to get mantis to start sending me email on new bugs, I can't figure out why it won't do that... :(
[11:31] Charles Krinkeb: Mantis#493 deals with the new grid server stuff not working.
[11:33] Neas Bade: MW, do you want to weigh in on protocols?
[11:34] You: today we use a mixture of rest and xmlrpc and a tad of binary .net remoting
[11:34] Neas Bade: per IRC, I think MW is tied up at at the moent
[11:34] You: I would suggest that we move towards rest
[11:34] Neas Bade: Tleiades +1 on that
[11:35] You: and that the message format be XML, *not* LLSD
[11:35] Neas Bade: XML over JSON?
[11:36] Neas Bade: agreeing that LLSD isn't what we want here
[11:36] You: hmm .. JSON is a definate maybe
[11:36] Tedd Maa: how much traffic is there?
[11:36] You: XML is just so darn easy to convert to objects in .NET
[11:36] You: XML is just so darn easy to convert to objects in .NET
[11:37] Neas Bade: xml is fine as well
[11:38] Neas Bade: Tleiades, I don't think you'll get any resistance on moving some of the coms protocols to REST
[11:38] Tedd Maa: what is the amount of communication?
[11:38] You: I think it is important that we are prepared to see server components like grid, asset and inventory implemented in something else than mono
[11:38] Neas Bade punts to Tleiades on Tedd's question
[11:39] You: Tedd.. well need a lot of comms to get all the funtionality in place
[11:39] Neas Bade: as I assume there is a good specific instance you've already got in mind for this
[11:39] You: groups, profiles
[11:39] Tedd Maa: we need a high volume binary RPC for scripting
[11:39] Neas Bade: right, and we'll probably also need that for physics if we decide to remote that
[11:40] You: the first place I'll use this is for grid server, and maintaining a list of live sims/regions
[11:40] You: next I'll use it for assets
[11:40] You: then inventory
[11:40] Charles Krinkeb: Question: Architecturally, are then any perceived issues of concurrent execution in our user and grid server code when we have 20 users logging in at the same moment with our comms protocols ideas?
[11:40] Neas Bade: which isn't a 100hz sort of operation like scripting and physics need to be
[11:41] Charles Krinkeb: Hello Chris. Welcome
[11:42] Neas Bade: actually, ckrinke, it would be good to see how your box running this is doing right now, as we've got ODE running and a reasonable amount of folks here (maybe time for another crash-a-thon)
[11:42] You: no, I think the physics and scripting needs a faster protocol, compared to what I am doing
[11:42] Neas Bade: yeh, definitely agreed
[11:42] Chris Down: Hello Charles
[11:42] Neas Bade: I consider physics and scripting the exceptions here vs. the rule
[11:42] Tedd Maa: the library will be simple enough, using interfaces on each side ...
[11:43] Tedd Maa: just thinking it might be good to use a single protocol all over?
[11:43] Charles Krinkeb: neas: top says 50% of the cpu and 211meg of memory used for mono. This is region is running on the server where osgrid, svn and most everything else is running on.
[11:44] You: why couldn't we use the same?
[11:44] Neas Bade: if we move to a binary protocol for our user <-> grid <-> asset services we're going to have to gut it again down the road if we want to be able to be compatible with the agent / region domain model that is coming from Linden.
[11:44] Tedd Maa: Charles; does the box have HyperThreading or dualcore or something?
[11:44] Neas Bade: if we start using REST in that space now, it will be an easier transition
[11:45] You: but why not use HTTP for everything?
[11:45] Charles Krinkeb: I doubt it. It is an AMI running on EC2 (Amazons Elastic Computing Cloud) and is a virtual machine.
[11:45] Tedd Maa: HTTP is only well suited for humans (just like plain XML and HTML), not well suited for computers
[11:45] Neas Bade: Tleiades for something like scripting where you really will need closer to 10ms turn around, http is going to be an issue I think
[11:45] You: have you guys done any measurements on the difference between binary tcp and an HTPP keep alive connection?
[11:45] Tedd Maa: scripting requires high speed, low overhead, etc
[11:46] nebadonb izumi: its a 1.6ghz Processor with ike 2 gig of ram
[11:46] nebadonb izumi: or something
[11:46] You: http keep alive?
[11:46] nebadonb izumi: One EC2 Compute Unit provides the equivalent CPU capacity of a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor. This is also the equivalent to an early-2006 1.7 GHz Xeon processor referenced in our original documentation. See Measuring Compute Resources for a complete description of an EC2 Compute Unit.
[11:47] You: well, it is just a suggestion
[11:47] Tedd Maa: Imagine script server running 10.000 scripts... The overhead for all those requests between server and region with HTTP REST would be something like 10-100 times more traffic and ditto CPU cycles to decode the traffic
[11:47] Neas Bade: honestly, I think your real overhead isn't even the http, but the xml parsing
[11:47] nebadonb izumi: according to Amazon Website its 1 Virtual Core
[11:48] Neas Bade: it can always be revisited later
[11:48] nebadonb izumi: 1.7 GB memory 1 EC2 Compute Unit (1 virtual core with 1 EC2 Compute Unit) 160 GB instance storage (150 GB plus 10 GB root partition) 32-bit platform I/O Performance: Moderate
[11:48] Neas Bade: it seems that both Tedd and Tleiades are making progress in their domains with sane approaches. We can always go back later and look to see if one solution works better for the other once we've got more working code. :)
[11:49] Tedd Maa: yeah, my point was simply that the binary rpc for scripting has to work 100% and will be very efficient ... but if arguments like compatibility comes into the picture then REST is probably better
[11:49] Neas Bade: yep
[11:50] Tedd Maa: when LL opens their grid we need to be compatible, so that is the direction we need to go in
[11:50] You: I'd like to do some measurments
[11:50] Tedd Maa: I'm hoping to get Babbage onboard on the remote script engine
[11:50] Neas Bade: that would be nifty :)
[11:50] Neas Bade: Tleiades agreed
[11:50] Neas Bade: though we'll need to have more bits working first to do that, I think
[11:51] Charles Krinkeb: What would it take to be able to edit a prim directly and expect the script editor to come up? That would expedite script testing.
[11:52] Neas Bade: Ok, I'm going to have to run quite soon. Can I get a volunteer or two help me on sorting out the Top 10 Bugs and getting them up on the wiki (and mailed to the list)?
[11:52] You: well... grid mode inventory and assets
[11:52] Neas Bade: also has anyone had a chance to look at the roadmap:
[11:53] Charles Krinkeb: I'll take care of that.
[11:53] Neas Bade: Tleiades, how close is that code to being ready for the tree?
[11:53] You: assets are very close
[11:53] Charles Krinkeb: neas. shall I post the log also or is Tl doing that this week?
[11:53] Darok Kaminski: I have to go, bye
[11:53] You: bye Darok
[11:53] nebadonb izumi: bye Darok
[11:53] Neas Bade: Also, based on MW's comment on IRC, can we move this meeting an hour earlier?
[11:53] Neas Bade: byr Darok
[11:54] Charles Krinkeb: Yes
[11:54] nebadonb izumi: or have 2 meetings a week
[11:54] Neas Bade: as that would make it easier for MW to attend
[11:54] You: I won't have any issues by moving this one hour back
[11:54] nebadonb izumi: for differnt hemispheres
[11:54] Daedius Moskvitch: I am all for an hour back =)
[11:54] nebadonb izumi: yea anythings good for me
[11:54] Neas Bade: great
[11:55] Neas Bade: I'll change that on the wiki
[11:55] Neas Bade: ckrinke, you also have the todo to get IRC briding working for this region by next week :)
[11:55] Daedius Moskvitch: w00h00! =)
[11:56] Charles Krinkeb: neas. I need this to be a configuration option or I loose the ability to update this region as I update all the others.
[11:56] Charles Krinkeb: We can leave it frozen in time if you wish, revision speaking.
[11:57] Neas Bade: ok, I'll commit to making this something you can put into the .ini by the end of the week
[11:57] Neas Bade: beat me up if i haven't done it by friday
[11:57] You: the code is pretty easy, it is all in there, it just needs to be dusted off, and some config file gymnastics
[11:57] Tedd Maa: my tummy hurts... I ate too much tacos
[11:57] nebadonb izumi: lol
[11:57] Charles Krinkeb: No way will I beat you up. You do as good a job of it as I do beating myself up.
[11:58] Neas Bade: mmmmm tacos
[11:58] Tedd Maa: I need tacos! I need them or I will explode! That happens to me sometimes...
[11:58] Charles Krinkeb: Kudos to nebabon for building this beautiful corporate park.
[11:58] Daedius Moskvitch: Indeed!
[11:58] Neas Bade: Yeh, I am quite impressed
[11:59] Charles Krinkeb: Too bad the water doesnt run <heh>
[11:59] You: yes, lets hope it survives moving to a new inventory/asset storage
[11:59] nebadonb izumi: thanks guys
[11:59] Charles Krinkeb: Not if the gridserver doesnt run. And it doesnt run now.
[11:59] nebadonb izumi: no big deal if it doesnt
[11:59] nebadonb izumi: i can rebuild it
[11:59] Neas Bade: we can do some hand migration on that front
[12:00] nebadonb izumi: this was only about 4 hours of work
[12:00] Neas Bade: that's still more than would be polite to loose
[12:00] Charles Krinkeb: Our hope is to add a welcome center for newcomers and a place to allow sim owners to describe their sims.
[12:01] Neas Bade: ok, I'm running off. Thanks for everyone coming. I'm really happy we are getting such attendance already :)
[12:01] Neas Bade: and it's really nice to see the environment evolve around us
[12:01] nebadonb izumi: yea good stuff
[12:01] You: hmmm... it should be possible to write some code, which will export from the SQLite databases, and import it into the gridmode stuff
[12:01] Charles Krinkeb: See you later all. Thanks for cominig.
[12:01] Daedius Moskvitch: Take care =)
[12:01] You: bye
[12:01] nebadonb izumi: yea it shjoudlnt be hard
[12:02] nebadonb izumi: although i noticed your using Binary for the UUID in mysqql
[12:02] nebadonb izumi: while Sqlite is using varchar
[12:02] nebadonb izumi: just needs some converting
[12:02] nebadonb izumi: wont just be a simple export/import
[12:03] You: I'm worried about the size of the database, that will cut the size down by 56 per asset
[12:03] nebadonb izumi: yea
[12:03] nebadonb izumi: its not a huge deal
[12:03] Chris Down: Bye everyone. I hope to be able to get more involved soon.
[12:03] nebadonb izumi: all we would have to do is save-xml
[12:03] nebadonb izumi: load-xml
[12:03] Tedd Maa: perl regex should be able to do that
[12:03] nebadonb izumi: and i retexture everything
[12:04] nebadonb izumi: i assume we will still be storing the assets the same way
[12:04] nebadonb izumi: or will we actually begin using the assets folder?
[12:04] You: the assets folder will only be used by the asset server
[12:05] nebadonb izumi: yea so initially the assets will still be stored on the sims in the same fashion it is now/
[12:05] nebadonb izumi: ?
[12:05] nebadonb izumi: in the .yap file
[12:05] nebadonb izumi: ?
[12:05] You: no
[12:05] nebadonb izumi: yea so just transfering the asset data in the database wont work
[12:06] You: the assets will be stored using mysql on the asset server
[12:06] nebadonb izumi: is that how it is now?
[12:06] You: and all the sims will use http/rest to retrieve the assets
[12:06] nebadonb izumi: yea
[12:06] nebadonb izumi: right now its being stored in the .YAP fie right?
[12:06] nebadonb izumi: or am i wrong about that
[12:06] You: well, yes and no
[12:06] You: we have two implementations
[12:07] You: depending on the setting in OpenSim.ini
[12:07] nebadonb izumi: ah ok
[12:07] nebadonb izumi: what setting uses the .yap?
[12:07] Tedd Maa: there are plans to move physics engine to remote?
[12:07] You:
[12:08] Tedd Maa: if so, should we look at some kind of tree structure?
[12:08] You: asset_database = db4o
[12:08] You: makes it use the yap file
[12:08] You: anything else will make the assets go to SQLite
[12:08] Tedd Maa: example; level 1 frontend server operating on low resolution, level 2 medium resolution, level 3 high resolution
[12:08] nebadonb izumi: ok cool
[12:08] nebadonb izumi: thats good
[12:09] You: but that will change soon
[12:09] nebadonb izumi: that makes it easier
[12:09] nebadonb izumi: yea
[12:09] nebadonb izumi: cant wait
[12:09] Tedd Maa: level 1; 1 server, level 2: 2 servers, level 3: 5 servers <- support many more users than a single server could
[12:09] You: ok, capturing the log now, and saving it to the wiki

Personal tools
About This Wiki