(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
[11:02 AM PT] Andrew Hellershanks: Hello, everyonel
[11:03 AM PT] Joe Magarac: Oh. that wasn't logged? Repeat follows. Joe Magarac: I'm working on documenting the region state change events for login, teleport, seeing a new region, etc. See https://community.secondlife.com/forums/topic/503010-obscure-question-when-does-the-simulator-send-establishagentcommunication-to-the-viewer
[11:01] Joe Magarac: for event diagrams. Comments on what's right, what's wrong, and what's different for Open Simulator would help.
[11:04 AM PT] Andrew Hellershanks: I just read that Firestorm turned 13 two days ago.
[11:04 AM PT] Gavin.Hird @grid.xmir.org:8002: already a teenager
[11:04 AM PT] Selby.Evans @grid.kitely.com:8002: Hi everyome
[11:05 AM PT] Andrew Hellershanks: Hello, Selby.
[11:05 AM PT] Gavin.Hird @grid.xmir.org:8002: hi
[11:05 AM PT] Ubit Umarov: ok sanity restored :)
[11:05 AM PT] Ubit Umarov: hi all
[11:05 AM PT] Andrew Hellershanks: Hello, Ubit.
[11:06 AM PT] Ubit Umarov: not fan of this 4k things on win
[11:07 AM PT] Ubit Umarov: dang lost chat.. what was the url joe?
[11:08 AM PT] Andrew Hellershanks: https://community.secondlife.com/forums/topic/503010-obscure-question-when-does-the-simulator-send-establishagentcommunication-to-the-viewer
[11:08 AM PT] Ubit Umarov: ty
[11:08 AM PT] Ubit Umarov: ok for sl
[11:08 AM PT] Andrew Hellershanks: Joe, once you have it worked out and documented the information can be added to the os.org wiki.
[11:11 AM PT] Ubit Umarov: seems that is missing several things
[11:11 AM PT] Joe Magarac: Please tell me.
[11:12 AM PT] Orbert.Tatham @hg.zetaworlds.com: I love Monty Linden's comment: "The Opensim version of this has got to be better than what I'm looking at. "
[11:12 AM PT] Gavin.Hird @grid.xmir.org:8002: :-)
[11:12 AM PT] Cuga.Rajal @rajal.org:9000: LOL
[11:12 AM PT] Andrew Hellershanks: :)
[11:13 AM PT] Ubit Umarov: well you do need talk with destiny region, getting its simulator features
[11:14 AM PT] Ubit Umarov: at same time, open the UDP circuit
[11:14 AM PT] Joe Magarac: Opening the UDP circuit is UseCircuitCode.
[11:14 AM PT] Ubit Umarov: well some is there on the graphics
[11:14 AM PT] Ubit Umarov: well is more
[11:15 AM PT] Ubit Umarov: but yeah graphic does show more of that
[11:15 AM PT] Ubit Umarov: region handshake is not there and is also critical
[11:16 AM PT] Ubit Umarov: server will ignore http requests until udp regionhandshake is done afet completagent movement
[11:17 AM PT] Joe Magarac: Region handshake is on the login chart but not the neighbor chart. You are right about that.
[11:18 AM PT] Ubit Umarov: TP and login are identical
[11:18 AM PT] Joe Magarac: And there are some events such as ParcelExtendedFlags not listed.
[11:18 AM PT] Ubit Umarov: just difers on who gives ther seeds and warns the destination about arrival
[11:19 AM PT] Ubit Umarov: on opensim simulator featurs now has have helping information about the target grid etc
[11:19 AM PT] Joe Magarac: OK. Although I think that TP to a nearby region that is already a neighbor region is different.
[11:20 AM PT] Ubit Umarov: regionhandshake is critrical to also tell region abotu some viewer caps like objects caching.. etc
[11:20 AM PT] Joe Magarac: Right, the cap access is not mentioned.
[11:21 AM PT] Ubit Umarov: err caps as same features. not as protocol capabilities
[11:21 AM PT] Joe Magarac: OK.
[11:21 AM PT] Ubit Umarov: like objecs cache support.
[11:21 AM PT] Joe Magarac: What about the race condition with AgentUpdate at login?
[11:22 AM PT] Joe Magarac: Viewer has to send AgentUpdate before it knows where the avatar really is.
[11:22 AM PT] Ubit Umarov: that is noise
[11:22 AM PT] Ubit Umarov: until all handshake done
[11:22 AM PT] Joe Magarac: It's not noise. It causes interest list errors with SL servers.
[11:22 AM PT] Ubit Umarov: moslty triggered on complet agent movement
[11:22 AM PT] Ubit Umarov: it is noise...
[11:23 AM PT] Ubit Umarov: only deep into completeagent movment, the region does know about you..
[11:24 AM PT] Joe Magarac: Server will not send object updates until viewer has sent first agent update.
[11:24 AM PT] Joe Magarac: Viewer finds out where avatar is via object update of avatar.
[11:24 AM PT] Ubit Umarov: yes.. but deep into complet gent movent
[11:24 AM PT] Joe Magarac: So position of camera in first agent update is bogus.
[11:24 AM PT] Ubit Umarov: some servers do send "noise" before that.. no idea why
[11:25 AM PT] Ubit Umarov: think i still send one.. to keep some viewers happy
[11:26 AM PT] Joe Magarac: It's a problem on SL. That sudden jump in agent position on the first object update confuses the SL system which sends object updates, and some object updates do not get sent. Usually the same ones each time.
[11:26 AM PT] Ubit Umarov: well that is a twisted process and my memory is bad :)
[11:26 AM PT] Joe Magarac: Agreed that it is a twisted process.
[11:27 AM PT] Ubit Umarov: opensim only starts objects and othr avatars after knowing about you
[11:27 AM PT] Andrew Hellershanks: Sensible
[11:27 AM PT] Ubit Umarov: as i said most triggers in completagentmoment, then continued under timers
[11:28 AM PT] Gavin.Hird @grid.xmir.org:8002: well, the viewer have to ask the server where the avatar actually is on login, because it can be in a completely different place than it was at logout, both within a region or in a different region
[11:28 AM PT] Ubit Umarov: as i said you need to get http seeds and simulator features
[11:28 AM PT] Joe Magarac: But the viewer needs to know the location of the avatar via an object update to compute the camera position.
[11:28 AM PT] Ubit Umarov: the the server will ignore most yr http until you finish the lludp process
[11:29 AM PT] Ubit Umarov: the viewer only needs to display that when told
[11:29 AM PT] Joe Magarac: What message finishes the "lludp" process at login?
[11:31 AM PT] Joe Magarac: @Gavin right, the viewer's initial avatar location is a guess. The server adjusts it for "Last", "Home", landing points, ban lines, and physical obstacles.
[11:31 AM PT] Gavin.Hird @grid.xmir.org:8002: I suppose via an update makes most sense as the avatar can be on an object anywhere in the region on login
[11:31 AM PT] Ubit Umarov: when region turns you into a root etc it sends you
[11:31 AM PT] Ubit Umarov: ControllingClient.SendRegionHandshake();
[11:32 AM PT] Gavin.Hird @grid.xmir.org:8002: even inside an object
[11:32 AM PT] Ubit Umarov: ( that if it is not a region cross )
[11:33 AM PT] Ubit Umarov: then also sends AgentMovementComplete
[11:33 AM PT] Joe Magarac: Region handshake does not have an avatar location.
[11:33 AM PT] Ubit Umarov: just after that it will start send data about your avatar, etc
[11:34 AM PT] Ubit Umarov: well actually tells all others abotu you
[11:34 AM PT] Gavin.Hird @grid.xmir.org:8002: ...like inside a hollow prim or a mesh house.
[11:34 AM PT] Ubit Umarov: ther info abotu you, to you ..
[11:34 AM PT] Gavin.Hird @grid.xmir.org:8002: so there is little point in starting to send the entire view distance if the avatar cannot even see it
[11:34 AM PT] Joe Magarac: CompleteAgentMovement does not have avatar location.
[11:35 AM PT] Joe Magarac: As far as I can tell, only ObjectUpdate tells viewer the avatar location.
[11:35 AM PT] Ubit Umarov: some other things will be delayed until you do answer to handshake
[11:35 AM PT] Ubit Umarov: like objects sending ...
[11:35 AM PT] Gavin.Hird @grid.xmir.org:8002: as I said, that makes most sense Joe
[11:37 AM PT] Ubit Umarov: yes you will get a full object update for your avatar just after AgentMovementComplete
[11:38 AM PT] Joe Magarac: I don't think SL will send me a full object update for the avatar until the viewer has sent AgentUpdate at least once.
[11:38 AM PT] Gavin.Hird @grid.xmir.org:8002: the avatar is just an object just like other objects. A highly specialized one, but still an object with features such as mprhp targets
[11:39 AM PT] Gavin.Hird @grid.xmir.org:8002: morph*
[11:39 AM PT] Ubit Umarov: it needs the fullobjectupdate
[11:39 AM PT] Andrew Hellershanks: Gavin, I'm glad you corrected the typo.
[11:40 AM PT] Gavin.Hird @grid.xmir.org:8002: oh - did the typo have some meaning?
[11:40 AM PT] Andrew Hellershanks: No. I just had no idea what that word was if it was a word.
[11:40 AM PT] Gavin.Hird @grid.xmir.org:8002: all right :-)
[11:41 AM PT] Joe Magarac: Will an OS server send the avatar's object update before the viewer sends the first AgentUpdate?
[11:41 AM PT] Gavin.Hird @grid.xmir.org:8002: why would it?
[11:42 AM PT] Ubit Umarov: it may yes
[11:42 AM PT] Gavin.Hird @grid.xmir.org:8002: wouldn't that just complicate processing in the viewer compared to SL?
[11:42 AM PT] Ubit Umarov: but well im answring from memory and it is bad :)
[11:43 AM PT] Joe Magarac: Right. Something to think about.
[11:43 AM PT] Ubit Umarov: os process is not that dif from sl
[11:43 AM PT] Vincent.Sylvester @hg.zetaworlds.com: *hands Ubit some ECC pills*
[11:43 AM PT] Joe Magarac: There's a circular problem here, and I've described it.
[11:43 AM PT] Ubit Umarov stores the exx pills on a safe place
[11:43 AM PT] Ubit Umarov: ecc
[11:44 AM PT] Ubit Umarov: agentupdate is irrelant for the region until it decides it did accept the avatar
[11:45 AM PT] Gavin.Hird @grid.xmir.org:8002: be glad you don't have to resolve the circularity with out of sequence tcp packets
[11:45 AM PT] Vincent.Sylvester @hg.zetaworlds.com: You have a similar issue with the terrain patches as well, which is why it's basically impossible with the current protocol to not send them and not have the viewer freak out completely despite there being some valid use cases for omitting the patches.
[11:45 AM PT] Ubit Umarov: well there is the fun par that lludp has no order :)
[11:46 AM PT] Gavin.Hird @grid.xmir.org:8002: so whatever packet arrives will be processed as they come
[11:46 AM PT] Joe Magarac: @gaving I did. A long time ago, I was one of the designers of TCP. See the two RFCs from John Nagle.
[11:46 AM PT] Ubit Umarov: but things that do depend on order, do wait
[11:46 AM PT] Ubit Umarov: yes like that option we always turn off
[11:47 AM PT] Ubit Umarov runs and hides
[11:47 AM PT] Gavin.Hird @grid.xmir.org:8002: OK - sweet!
[11:47 AM PT] Ubit Umarov: ;)
[11:47 AM PT] Andrew Hellershanks: Joe, Nice.
[11:48 AM PT] Joe Magarac: That's because Berkeley put in delayed ACKs. You never want both delayed ACKs and the Nagle algorithm on at the same time. That gets you the fixed ACK delay.
[11:48 AM PT] Joe Magarac: But that's off topic.
[11:48 AM PT] Ubit Umarov: yes i did read your coments abotu that ages ago :)
[11:49 AM PT] Joe Magarac: Anyway, I'd appreciate it if people would look at those charts and try to fix them.
[11:49 AM PT] Joe Magarac: I really want to see one for region crossings.
[11:49 AM PT] Joe Magarac: I have no idea how that works.
[11:49 AM PT] Joe Magarac: Not in detail.
[11:49 AM PT] Gavin.Hird @grid.xmir.org:8002: it is pure magic haha
[11:49 AM PT] Joe Magarac: Right.
[11:49 AM PT] Vincent.Sylvester @hg.zetaworlds.com: Assumptions and dreams basically
[11:49 AM PT] Cuga.Rajal @rajal.org:9000: PM you for updates?
[11:49 AM PT] Orbert.Tatham @hg.zetaworlds.com: When it works at all
[11:50 AM PT] Ubit Umarov: well some code paths are similar
[11:50 AM PT] Joe Magarac: and I am tryingto nail it down so it is non magic.
[11:50 AM PT] Joe Magarac: I'm "nagle@animats.com" if you need to reach me.
[11:50 AM PT] Cuga.Rajal @rajal.org:9000: thx
[11:50 AM PT] Gavin.Hird @grid.xmir.org:8002: remember the time in SL when a region crossing often resulted in your wearables ending up at places they were not supposed to be at all
[11:50 AM PT] Ubit Umarov: but those are still not cool .. work.. but not cool
[11:51 AM PT] Ubit Umarov: very sensible to network weather
[11:51 AM PT] Joe Magarac: The reason I push on this is because, if you look at that list of seven charts of region transitions. those are the things that stall.
[11:51 AM PT] Vincent.Sylvester @hg.zetaworlds.com: I'll happily walk through the OpenSim source with you to piece things together if that helps you
[11:51 AM PT] Joe Magarac: Thanks!
[11:51 AM PT] Ubit Umarov: sinmpler at sl, bc they do control the weather better
[11:52 AM PT] Cuga.Rajal @rajal.org:9000: If you do a walk-through can I observe?
[11:52 AM PT] Joe Magarac: Yes, the have Amazon's backbone.
[11:52 AM PT] Joe Magarac: I see more out of of order packets from OS sims than SL sims.
[11:52 AM PT] Ubit Umarov: yes.. weather..
[11:53 AM PT] Ubit Umarov: regions to try to send them in order
[11:53 AM PT] Joe Magarac: But I'm in western US and so is SL's data center, which is AWS-WEST-2 in Oregon.
[11:53 AM PT] Ubit Umarov: well they set the order number on sending... easy :p
[11:53 AM PT] Gavin.Hird @grid.xmir.org:8002: OS sims can be located all over, with highly variable latency and processing capabilites
[11:53 AM PT] Joe Magarac: Right. Very aware of that. I need to increase some timers. I only allow 5 secs for the login handshake, and some OS sims don't respond that fast, for example.
[11:53 AM PT] Gavin.Hird @grid.xmir.org:8002: I'd be surprised if any OS simulator are in a proper datacenter
[11:54 AM PT] Vincent.Sylvester @hg.zetaworlds.com: Ahem
[11:54 AM PT] Vincent.Sylvester @hg.zetaworlds.com: Cough cough
[11:54 AM PT] Gavin.Hird @grid.xmir.org:8002: ok
[11:54 AM PT] Gavin.Hird @grid.xmir.org:8002: how many regions?
[11:54 AM PT] Joe Magarac: This has been good. How do I get the transcript?
[11:55 AM PT] Ubit Umarov: servers on data center?? and lose all that nice heat under the bed in winter??
[11:55 AM PT] Vincent.Sylvester @hg.zetaworlds.com: It's on the OpenSim Wiki under the Office Hours page
[11:55 AM PT] Vincent.Sylvester @hg.zetaworlds.com: They are all linked there
[11:55 AM PT] Gavin.Hird @grid.xmir.org:8002: http://opensimulator.org/wiki/Office_Hours
[11:55 AM PT] Vincent.Sylvester @hg.zetaworlds.com: Gavin, 712 at the moment
[11:56 AM PT] Gavin.Hird @grid.xmir.org:8002: so 712 out of what... 300k (if you believe HGB's stats)
[11:56 AM PT] Gavin.Hird @grid.xmir.org:8002: nice Vince
[11:56 AM PT] Vincent.Sylvester @hg.zetaworlds.com: 712 simulators, 10k in normal regions
[11:56 AM PT] Joe Magarac: Thanks very much, everyone. Everybody agrees this area is a mess. Maybe we can document how it's supposed to work and fix it. I'm talking LL people around to that position.
[11:57 AM PT] Andrew Hellershanks: Thank you for taking this on, Joe.
[11:57 AM PT] Vincent.Sylvester @hg.zetaworlds.com: Not the first time we end up fixing their mess lol
[11:57 AM PT] Andrew Hellershanks: hehe
[11:58 AM PT] Orbert.Tatham @hg.zetaworlds.com: They are even admitting that, to my great surprise
[11:58 AM PT] Andrew Hellershanks: We are just coming up to the top of the hour. Any questions before people start heading back to RL?
[11:59 AM PT] Andrew Hellershanks: I don't see anyone typing. We are done for today?
[11:59 AM PT] Gavin.Hird @grid.xmir.org:8002: have anyone tested MYSQL Server 8.1 yet?
[12:00 PM PT] Andrew Hellershanks: I'm still on 8.0
[12:00 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Major changes of concern with that?
[12:00 PM PT] Selby.Evans @grid.kitely.com:8002: bye all
[12:00 PM PT] Gavin.Hird @grid.xmir.org:8002: maybe, I am not sure
[12:00 PM PT] Andrew Hellershanks: ok, Selby. Thanks for being here.
[12:00 PM PT] Ubit Umarov: cya Selby.Evans
[12:01 PM PT] Gavin.Hird @grid.xmir.org:8002: I suppose I'll have to get it installed and see how it goes
[12:01 PM PT] Kayaker Magic: I'm still on mysql 8.0
[12:01 PM PT] Vincent.Sylvester @hg.zetaworlds.com: We fairly recently made a small change to the connectors to fix a problem on MariaDB
[12:01 PM PT] Gavin.Hird @grid.xmir.org:8002: 8.033 or something for me
[12:01 PM PT] Ubit Umarov: think im on 8.0 on my win also
[12:01 PM PT] Andrew Hellershanks: yup. 8.0.33 here.
[12:01 PM PT] Kayaker Magic: There is some confusion, MariaDB is up to 10.x and that is the default on Ubuntu now.
[12:01 PM PT] Joe Magarac: Gotta go, next meetting. Thanks all.
[12:01 PM PT] Jamie.Jordan @grid.kitely.com:8002: have a great week yall
[12:01 PM PT] Lyr Lobo smiles
[12:02 PM PT] Ubit Umarov: 8.0.31
[12:02 PM PT] Andrew Hellershanks: ok, Joe. Thanks for dropping by.
[12:02 PM PT] Orbert.Tatham @hg.zetaworlds.com: mysql Ver 15.1 Distrib 10.6.12-MariaDB
[12:02 PM PT] Ubit Umarov: 8.0.31 community
[12:02 PM PT] Gavin.Hird @grid.xmir.org:8002: 8.033 community, yes
[12:02 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Only ever tested mysql when the connector broke, using MariaDB generally since that seems to be faster in some respects, but mysql is catching up
[12:02 PM PT] Kayaker Magic: There are features of Mysql 8.0 that I want, and it is unclear if MariaDB has implimented them
[12:03 PM PT] Cuga.Rajal @rajal.org:9000: 8.0.27
[12:03 PM PT] Cuga.Rajal @rajal.org:9000: I need to upgrade
[12:03 PM PT] Lyr Lobo: I thought that MariaDB 10 is not working well with OpenSim - did that change?
[12:03 PM PT] Kayaker Magic: Features like doing binary operations on blobs
[12:03 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Both now have a weird guid type that we could use for some things, but I am not sure it's actually any faster or better
[12:03 PM PT] Ubit Umarov: (update == if not broken, break it)
[12:03 PM PT] Ubit Umarov: :)
[12:03 PM PT] Lyr Lobo chuckles
[12:04 PM PT] Andrew Hellershanks: hehe
[12:04 PM PT] Ubit Umarov: oops except in opensim case, ofc, coff coff coff
[12:04 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Added a typo, just for fun
[12:06 PM PT] Ubit Umarov: AHh think you noticed that there where no new bugs during last week..
[12:06 PM PT] Jagga Meredith: I use mariadb with no problems
[12:06 PM PT] Ubit Umarov: err i mean code changes..
[12:06 PM PT] Kayaker Magic: LOL
[12:07 PM PT] Ubit Umarov: mariadb may became a problem since they are diverging from mysql on the connector
[12:07 PM PT] Ubit Umarov: we may need to change to one that seems to handle both..
[12:07 PM PT] Ubit Umarov: ofc more external code..
[12:07 PM PT] Vincent.Sylvester @hg.zetaworlds.com: I still have that patch somewhere so that would be not much trouble
[12:08 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Probably a good opportunity to evaluate which offers better performance for OpenSim needs and focus on that
[12:08 PM PT] Gavin.Hird @grid.xmir.org:8002: how is Microsoft dotnet code keeping up with MySQL?
[12:09 PM PT] Vincent.Sylvester @hg.zetaworlds.com: something something nuget, probably
[12:09 PM PT] Ubit Umarov: think i did update the connector to a updated one from mysql
[12:09 PM PT] Gavin.Hird @grid.xmir.org:8002: ok
[12:09 PM PT] Vincent.Sylvester @hg.zetaworlds.com: That's technically where our connectors are now as well, just our version is built from source with minor fixes
[12:10 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Otherwise we'd have to use all sorts of dependencies as well, which are nasty
[12:10 PM PT] Ubit Umarov: mysql.data.dll version 8.0.31
[12:10 PM PT] Vincent.Sylvester @hg.zetaworlds.com: There is a mantis ticket about that whole ordeal somewhere
[12:11 PM PT] Ubit Umarov: think this orginal from mysql without changes we made on 6.x ??
[12:11 PM PT] Tess JL: I think it's time for me to head out. Have a great week everydboy!
[12:12 PM PT] Ubit Umarov: cya
[12:12 PM PT] Gavin.Hird @grid.xmir.org:8002: you too Tess
[12:12 PM PT] Vincent.Sylvester @hg.zetaworlds.com: *inhales some ECC memory* I don't recall the specifics anymore
[12:12 PM PT] Andrew Hellershanks: Bye, Tess. See you next week.
[12:12 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Something like ancient times version though yeah
[12:12 PM PT] Vincent.Sylvester @hg.zetaworlds.com: Newer stuff has all unsafe and other speed hacks
[12:13 PM PT] Vincent.Sylvester @hg.zetaworlds.com: type security go brrrr
[12:13 PM PT] Vincent.Sylvester @hg.zetaworlds.com: straight off a cliff
[12:13 PM PT] Ubit Umarov forgot where the ecc pills are
[12:15 PM PT] Andrew Hellershanks: They aren't going to help if you can't remember where you put the.
[12:15 PM PT] Andrew Hellershanks: s/the/them/
[12:15 PM PT] Orbert.Tatham @hg.zetaworlds.com: Ok, folks - Ubit needs his meds, and I need to head out
[12:15 PM PT] Gavin.Hird @grid.xmir.org:8002: the location was purged from cache
[12:15 PM PT] Andrew Hellershanks: I also have something I need to get back to.
[12:15 PM PT] Kayaker Magic: I bought a book "how to improve your memory", haven't read it yet, can remember where I put it...
[12:15 PM PT] Andrew Hellershanks: On that note, thank you all for coming. See you again next week.
[12:15 PM PT] Ubit Umarov: yeha rl also calling... cya later
