Chat log from the meeting on 2024-03-26

From OpenSimulator

Jump to: navigation, search
[11:09 AM PT]  Ubit Umarov: welcome :)
[11:09 AM PT]  Joe Magarac: Hello again.
[11:09 AM PT]  Joe Magarac: I've been testing region crossings in Sharpview.
[11:09 AM PT]  Ubit Umarov: and still alive.. how good ;)
[11:10 AM PT]  Joe Magarac: Yes.
[11:10 AM PT]  Joe Magarac: I've been driving the little yellow car at Ubittest2 back and forth with no serious problems.
[11:10 AM PT]  Lyr Lobo: hehe
[11:10 AM PT]  Ubit Umarov: yes that is a simple car script
[11:10 AM PT]  Joe Magarac: Sharpview's motoin extrapolation is off somehow. Works for SL, but sometimes way off for OS.
[11:10 AM PT]  Ubit Umarov: .. or has...
[11:11 AM PT]  Joe Magarac: Tried the corner at Sharpview/sharpview2 etc, where 3
[11:11 AM PT]  Joe Magarac: regions meet.
[11:11 AM PT]  Joe Magarac: Mostly works.
[11:11 AM PT]  Joe Magarac: Sometimes parts of the vehicle disappear until the next region crossing.
[11:11 AM PT]  Joe Magarac: This indicates that some of the object updates got lost.
[11:11 AM PT]  Vincent.Sylvester Given that going right over the edge is hard that it works this well is surprising itself
[11:12 AM PT]  Joe Magarac: Sharpview does not have the optimization Firestorm has to copy them over from the losing sim.
[11:12 AM PT]  Joe Magarac: Sharpview does a full rebuild whenever an object crosses regions.
[11:13 AM PT]  Joe Magarac: All this is in very lightly loaded regions. It may work worse in busy areas.
[11:13 AM PT]  Joe Magarac: This is preliminary. I'll have more data soon. With good data collection on both server and viewer side, we should be able to fix any problems in this area. I want to push through on that.
[11:14 AM PT]  Vincent.Sylvester Plus hitting a corner you mostly get two crossings unless you hit the exact corner, so that it doesn't blow up more is a good sign
[11:14 AM PT]  Kayaker Magic: Or when the two regions are not on the same server, more internet delays cause more race conditions.
[11:14 AM PT]  Joe Magarac: Right.
[11:15 AM PT]  Joe Magarac: Probably useful to bring up a region on Linux with the Linux feature turned on that introduces network delay.
[11:15 AM PT]  Vincent.Sylvester You could test over on ZetaWorlds, I got a few adjacent regions that have about 250ms of roundabout distance between them
[11:15 AM PT]  Vincent.Sylvester That definitely gives you a good bump
[11:16 AM PT]  Joe Magarac: Sounds good. I think I found out why I couldn't log in over there.
[11:16 AM PT]  Joe Magarac: I only allow 5 secs for the login handshake and some login servers get really slow.
[11:16 AM PT]  Vincent.Sylvester It should reply quite quickly unless you inventory is big, so that's odd
[11:17 AM PT]  Vincent.Sylvester Then again is across the pond from you
[11:19 AM PT]  Joe Magarac: I want to try only regions that have the latest version of Open Simulator. No use rediscovering bugs already fixed.
[11:19 AM PT]  Joe Magarac: I hope to have a new Sharpview out next week that allows sitting, so others can try this.
[11:19 AM PT]  Joe Magarac: That's my stuff for today.
[11:19 AM PT]  Joe Magarac: Thanks.
[11:20 AM PT]  Ubit Umarov: ty
[11:20 AM PT]  Ubit Umarov: well on code changes,  i accepted a simple ossl function
[11:20 AM PT]  Vincent.Sylvester To the eternal dismay of Jeff apparently hehe
[11:21 AM PT]  Vincent.Sylvester The naming conventions are all over the place anyways
[11:21 AM PT]  Ubit Umarov: osGetLinkInventoryItemKey that then i did rename as osGetLinkInventoryItemAssetKey
[11:21 AM PT]  Ubit Umarov: and now im abotu to rename again as osGetLinkInventoryKey
[11:21 AM PT]  Ubit Umarov: so it is coerent with llGetLinkInventoryKey
[11:22 AM PT]  Kayaker Magic: Why not just use llGetLinkInventoryKey?
[11:22 AM PT]  Ubit Umarov: i just said that:p
[11:22 AM PT]  Ubit Umarov: im abotu to comit that second rename
[11:23 AM PT]  Vincent.Sylvester It's turned into a big mess
[11:23 AM PT]  Ubit Umarov: bc for ll  the key does mean the asset
[11:23 AM PT]  Vincent.Sylvester Then there is the issue of it taking the inventory name as search parameter, but multiple items can have the same name, different type
[11:23 AM PT]  Cuga.Rajal the spec on that function is messier than most
[11:23 AM PT]  Ubit Umarov: there is no itemkey on objects inventory LSL
[11:24 AM PT]  Kayaker Magic: Ugh.
[11:24 AM PT]  Ubit Umarov: in fact im still not sure why we (me?) did accept that one that exposes the itemkey
[11:24 AM PT]  Vincent.Sylvester For the give function I think
[11:25 AM PT]  Vincent.Sylvester Dunno, but really the way both lsl and ossl have grown to have different naming schemes drives me nuts regularly
[11:25 AM PT]  Ubit Umarov: well ill commit it as GetLinkInventoryKey  soon
[11:25 AM PT]  Ubit Umarov: possible ll will had similar one day
[11:26 AM PT]  Vincent.Sylvester Never mind that we have to grow lsl spec as it changes even if it doesn't make sense the way it is implemented
[11:26 AM PT]  Vincent.Sylvester It makes me just want to add a language override that is concise in naming schema
[11:26 AM PT]  Vincent.Sylvester No more ll or os, just something else and then a logical function name for all of them
[11:27 AM PT]  Vincent.Sylvester But good luck getting that adopted everywhere or even getting everyone to agree on a schema
[11:27 AM PT]  Vincent.Sylvester One day...
[11:27 AM PT]  Kayaker Magic: Then having to support all the old names for all the old code anyway.
[11:28 AM PT]  Vincent.Sylvester You can leave those in place. Technically would only need to make a new stub and point to old stuff
[11:28 AM PT]  Vincent.Sylvester Well oversimplified, but yeah adding a "language" to it not exactly difficult, just time consuming
[11:29 AM PT]  Vincent.Sylvester Speaking of, I continue writing a new migration system this week and have made quite some progress
[11:29 AM PT]  Vincent.Sylvester I have now changed to just determining the table version based on versioned schema files and apply migrations accordingly. Think that is the most prone to actually work
[11:30 AM PT]  Vincent.Sylvester Some migrations we can ditch as they are impossible, so that actually makes it much less than the version numbers lead on
[11:30 AM PT]  Vincent.Sylvester Discovered the other day that sqlite is going to require its own logic as it doesn't support the information_schema
[11:30 AM PT]  Vincent.Sylvester Which then brings me to. We got just more issues with Postgres then. I can't get that to work locally at the moment
[11:31 AM PT]  Vincent.Sylvester There is no one to maintain it and the connector we use doesn't work with the latest version on linux either
[11:32 AM PT]  Vincent.Sylvester I don't know who uses Postgres still, but it would be good to know whether it is worth supporting it going forward or if the time isn't better spent improving mysql/mariadb
[11:33 AM PT]  Lyr Lobo: someone mentioned using it
[11:33 AM PT]  Cuga.Rajal Gavin used to use it, he migrated to mySQL
[11:33 AM PT]  Cuga.Rajal I think bc the latest version of Postgres had issues
[11:34 AM PT]  Ubit Umarov: ( did commit the rename of the function )
[11:34 AM PT]  Vincent.Sylvester On windows I cannot seem to switch the auth plugin over. On nix that works, but then you get issues with the connector being incompatible, which has a whole thread on mantis about what's going on with that
[11:35 AM PT]  Ubit Umarov: well mysql / maria are spliting
[11:35 AM PT]  Ubit Umarov: i may change the connector to the opensource one..
[11:36 AM PT]  Ubit Umarov: i mean not from oracle :)
[11:36 AM PT]  Vincent.Sylvester Yeah that's part of why I am wondering
[11:36 AM PT]  Vincent.Sylvester If we can take advantage of some improvements in either that be good too, but it might mean splitting that up too, which would add one db engine to maintain
[11:37 AM PT]  Vincent.Sylvester Only so much time in a day
[11:37 AM PT]  Ubit Umarov: forget the name, but it is been done with the notion that mariadb and mysql are now diferent
[11:38 AM PT]  Kayaker Magic: I keep using MySQL from Oracle instead of MariaDB because MySQL supports bit-logical operaions on blobs and Maria's documentation is ... vauge on the subject.
[11:39 AM PT]  Kayaker Magic: Any one know if Maria supports bit operaions on blobs?
[11:39 AM PT]  Vincent.Sylvester Like editing parcel data?
[11:40 AM PT]  Ubit Umarov: i was talking abotu the connector..
[11:40 AM PT]  Ubit Umarov: about even..
[11:40 AM PT]  Lyr Lobo: not sure, Kayaker, but Spinoza uses Maria
[11:42 AM PT]  Vincent.Sylvester What intrigued me with Maria was the uuid thing, but turns out that's not as impressive as it seemed
[11:43 AM PT]  Vincent.Sylvester For the most part they are still similar, with both adding their own syntax and functions though
[11:43 AM PT]  Vincent.Sylvester Hasn't blown up yet, given our queries tend to not be too esoteric as long as basic sql is adhered to it shouldn't
[11:44 AM PT]  Ubit Umarov: no idea what bit logic operation on blobs means but sounds very undefined :p
[11:45 AM PT]  Vincent.Sylvester We only have a few spots with blobs too, so I wonder where you use them
[11:45 AM PT]  Kayaker Magic: It allows you to exclusive-or blobs with each other, for example. Which previously only worked on integers and BIGINT was the biggest thing you could XOR.
[11:46 AM PT]  Ubit Umarov: not sure what will the the result of Cats XOR Dogs
[11:46 AM PT]  Ubit Umarov: ;)
[11:47 AM PT]  Vincent.Sylvester Doing that directly on database, eh might not be wise either, always validate cause OpenSim won't heh
[11:52 AM PT]  Ubit Umarov: another minor change was that git no longer includes a default System.Drawing.Common.dll set for windows
[11:52 AM PT]  Ubit Umarov: runprebuild.bat does copy it now, line the does for linux
[11:53 AM PT]  Lyr Lobo: Thank you for the meeting. Must dash to an appointment *waves*
[11:53 AM PT]  Lyr Lobo waves
[11:53 AM PT]  Ubit Umarov: (like -> like )
[11:53 AM PT]  Ubit Umarov: have fun Lyr :)
[11:53 AM PT]  Lyr Lobo: thanks
[11:54 AM PT]  Ubit Umarov: in last meeting there was some talk abotu ll changes on voice
[11:55 AM PT]  Ubit Umarov: they are replacing vivox by their own internal service,  built aroung webrtc opensource parts plus their own code
[11:56 AM PT]  Vincent.Sylvester Ah yes, that'll be fun
[11:56 AM PT]  Joe Magarac: Yes. Please do it the same way on OS, so I only have to implement this once. Thank you.
[11:56 AM PT]  Ubit Umarov: they made public all the opensource libs they are using..
[11:57 AM PT]  Ubit Umarov: ofc viewer code is gpl, so tpvs should get all of it
[11:58 AM PT]  Ubit Umarov: since they are using opensource code they no longer need do connect to a external slvoice local server, that was close source
[11:58 AM PT]  Web Rain: Hi there :)
[11:58 AM PT]  Ubit Umarov: they jsut direct link the lower level webrtc deps, under GPL
[11:58 AM PT]  Vincent.Sylvester In any case need to make a new module for it, for the proxy part
[11:59 AM PT]  Kayaker Magic: Welcome Web
[11:59 AM PT]  Vincent.Sylvester Backend could be difficult for the spatial audio part since that's not open at the moment
[11:59 AM PT]  Ubit Umarov: we will need to wait for opensim viewers with that, to see
[11:59 AM PT]  Joe Magarac: It's not open in WebRTC?
[12:00 PM PT]  Ubit Umarov: they made their own spacial codec for their server
[12:00 PM PT]  Joe Magarac: Who, Linden Lab?
[12:00 PM PT]  Ubit Umarov: yes
[12:00 PM PT]  Joe Magarac: is it closed source?
[12:00 PM PT]  Ubit Umarov: a addon module for janus..
[12:00 PM PT]  Ubit Umarov: yes
[12:00 PM PT]  Joe Magarac: WTF?
[12:01 PM PT]  Ubit Umarov: they are stil consider if they will release it as opensource or keep closed
[12:01 PM PT]  Ubit Umarov: but that is server side.. not viewer..  viewer is bound by GPL rules
[12:01 PM PT]  Ubit Umarov: reason why slvoice was that external service only connected via internal tcp
[12:01 PM PT]  Vincent.Sylvester There is also the question of whether to put webrtc server end into OpenSim or keep is as own service, there are arguments for both sides
[12:02 PM PT]  Joe Magarac: Indeed.
[12:02 PM PT]  Ubit Umarov: in fact they have 3 type os servers
[12:02 PM PT]  Ubit Umarov: spacial voices ones, per regions
[12:02 PM PT]  Ubit Umarov: "peer to peer" ones for chat..
[12:02 PM PT]  Web Rain: in both cases , webrtc without ssl will just add more vuln to the box :(
[12:03 PM PT]  Ubit Umarov: and conversation ones for group chat
[12:03 PM PT]  Vincent.Sylvester Web, which is why I am gravitating towards own service approach rather than being part of OpenSim itself, just proxying over to a service that is secured properly
[12:03 PM PT]  Ubit Umarov: peer to peer seems to be just STUN..  but im guessing
[12:04 PM PT]  Web Rain: is what i invested a bit of time to explore and test :)
[12:04 PM PT]  Ubit Umarov: they will not use real p2p to protect users IPs
[12:05 PM PT]  Web Rain: my last PR will help you to do ^^ self signed certs for testing and maybe keep it in the main release
[12:05 PM PT]  Ubit Umarov: for us, this does open more possibilities that the close source, dead end  vivox
[12:05 PM PT]  Vincent.Sylvester When vivox first closed the registration page it's what I started looking into, so I have some working prototypes, sort of. This means some ground work towards the server end. OpenSim module is a matter of digging through protocol. The big questionmark is the codec and getting that to work with the viewer once it actually connects
[12:05 PM PT]  Ubit Umarov: and also no need for viewer side code like mumble does require
[12:06 PM PT]  Ubit Umarov: ofc we need to see how standard webrtc their ( non spacial )  implementation is
[12:07 PM PT]  Ubit Umarov: and if we can just glue up some STUN server for p2p etc..
[12:08 PM PT]  Vincent.Sylvester Google has a whole set of things for that, minus the spatial thing
[12:08 PM PT]  Ubit Umarov: ofc the server side opensource solutions they are using have no windows version...  linux only
[12:09 PM PT]  Vincent.Sylvester All the more reason not to try and stuff it into OpenSim heh
[12:09 PM PT]  Web Rain: if opensim work like a proxy for the voice this will be more safe and may give a room to deep customizations... also the client will get only the opensim ip and not the others ips
[12:09 PM PT]  Vincent.Sylvester That's how it works at SL too, so we have to follow that anyways
[12:10 PM PT]  Vincent.Sylvester The biggest reason for having it proxy though is geolocation since you want to have something close to the participants ideally
[12:10 PM PT]  Ubit Umarov: Beq did share with us the information she got from ll ->
[12:10 PM PT]  Vincent.Sylvester Voice lag is the most annoying thing having run Teamspeak for years I know that painfully well
[12:11 PM PT]  Vincent.Sylvester As with anything though we should wait a bit for SL to fall on its face a couple times and work out the bugs first xD
[12:11 PM PT]  Ubit Umarov: the cost of good low latency servers will be a problem
[12:11 PM PT]  Ubit Umarov: not many are vivox or discord..
[12:12 PM PT]  Ubit Umarov: btw discord is also webrtc based..  but their own variants.. ofc
[12:13 PM PT]  Ubit Umarov: bet sl is adding a nice pool os aws servers for that...
[12:13 PM PT]  Vincent.Sylvester Burning money is fun, you should try it lol
[12:14 PM PT]  Ubit Umarov: i did that a lot.. i used to smoke :p
[12:14 PM PT]  Vincent.Sylvester Still a better use than AWS I'd say heh
[12:14 PM PT]  Web Rain: XD
[12:17 PM PT]  Ubit Umarov: well cloud servers is a use case for voice ones
[12:17 PM PT]  Ubit Umarov: since the load is unpredictable, and more servers may be needed just a rush hours
[12:19 PM PT]  Ubit Umarov: even if we do manage to make voice work, i do not expect the kind of performance we do get with vivox..
[12:19 PM PT]  Vincent.Sylvester Outside of AWS though that type of scaling isn't exactly rocket science anymore either
[12:19 PM PT]  Web Rain: just for info i tested a bit the SSL implementation and seems all is working fine on a standalone... unfortunately i am not equipped for HG and grid tests, it requires several machines or a powerful machine
[12:20 PM PT]  Cuga.Rajal Discord's group voice if way better than Vivox
[12:20 PM PT]  Cuga.Rajal is way better
[12:20 PM PT]  Ubit Umarov: oh i did not look into those PRs still
[12:20 PM PT]  Cuga.Rajal but not spacial
[12:20 PM PT]  Ubit Umarov: well but they pay for the servers.. that is my concern :)
[12:20 PM PT]  Web Rain: don't merge now i have some commitus for the grids support
[12:20 PM PT]  Cuga.Rajal yes that is a major issue...
[12:21 PM PT]  Ubit Umarov: you wil not go far with a home server
[12:21 PM PT]  Cuga.Rajal Not a real solution, but I know people that just use Discord for their voice while in OS
[12:22 PM PT]  Cuga.Rajal they dont care about spacial
[12:22 PM PT]  Ubit Umarov: possble not even with a lost cost server on a servers provider
[12:22 PM PT]  Ubit Umarov: lost?? i meant low lol
[12:22 PM PT]  Cuga.Rajal its bandwidth issue more than server speed
[12:22 PM PT]  Joe Magarac: The trouble with a low-cost server provider is random latency.
[12:22 PM PT]  Ubit Umarov: and latency yes
[12:22 PM PT]  Joe Magarac: If you're sharing a server, other stuff can delay you randomly.
[12:23 PM PT]  Joe Magarac: Which really matters for voice.
[12:23 PM PT]  Ubit Umarov: yeaps..
[12:23 PM PT]  Joe Magarac: Something the size of OSgrid probably needs a dedicated server.
[12:23 PM PT]  Ubit Umarov: ( and high bandwith != low latency at all )
[12:24 PM PT]  Ubit Umarov: fear that more than one..
[12:24 PM PT]  Web Rain: spacial is just a sound volume lvl ? no? if is the case i think is possible to do that on opensim side no? (i think)
[12:24 PM PT]  Ubit Umarov: but not sure if low cost OHV kinda servers will even do it...
[12:25 PM PT]  Ubit Umarov: spacial can be a very complex things..
[12:25 PM PT]  Ubit Umarov: no idea how they did it
[12:26 PM PT]  Joe Magarac: Spatial should include level and delay at least. You can go much further and do better.
[12:26 PM PT]  Web Rain: volume by distance
[12:26 PM PT]  Ubit Umarov: and phase
[12:26 PM PT]  Ubit Umarov: mono to stereo with phase into account, etc
[12:27 PM PT]  Vincent.Sylvester You got things like kubernetes and all manners of similar things that make setting up a distributed voice network not that hard. There are even books about this
[12:27 PM PT]  Ubit Umarov: no idea how.. really with variable latency etc etc :)
[12:27 PM PT]  Vincent.Sylvester The biggest issue is scale, because for that to make sense you need to actually have use for it. If you only have five people talking then you just find their location and get something close rather than a whole network
[12:27 PM PT]  Vincent.Sylvester Sysadmin stuff anyways, not a programmers worry hehe
[12:28 PM PT]  Joe Magarac: Right.
[12:28 PM PT]  Joe Magarac: Have to go, another meeting. Thanks.
[12:28 PM PT]  Ubit Umarov: a lot harder that on a single application like viewers that do have it ofc
[12:28 PM PT]  Cuga.Rajal tc Joe
[12:28 PM PT]  Ubit Umarov: ohh we are way past the hour :)
[12:29 PM PT]  Ubit Umarov: any other opensim issue?
[12:29 PM PT]  Kayaker Magic: I'm good
[12:29 PM PT]  Cuga.Rajal nothing new
[12:30 PM PT]  Web Rain: no for me :) if you have a bit of time take a look in my repo this week
[12:30 PM PT]  Ubit Umarov: ok so lets end this meeting..  hope to see you all next week..   remember the hour change on EU, etc..
[12:30 PM PT]  Web Rain: if there is some thing that break SSL let me know
Personal tools
About This Wiki