Chat log from the meeting on 2019-04-09

From OpenSimulator

(Difference between revisions)
Jump to: navigation, search

Latest revision as of 12:50, 9 April 2019

[10:59] Gavin.Hird The perceived windows size is 1680 x 1050, but it renders at 3360 x 2100
[10:59] Bob.Wellman is online.
[10:59] Telehub: Welcome back to, Dev Outreach, Andrew Hellershanks.
[10:59] Selby.Evans is online.
[10:59] Selby.Evans is online.
[11:00] Andrew Hellershanks: Hello, everyone.
[11:00] Kayaker Magic: So the Mac filters down to the actual display resolution?
[11:00] Selby.Evans hi everyone
[11:00] Gavin.Hird so very crisp, but 4x as much work
[11:00] Gavin.Hird Hi Andrew and Selby
[11:01] Gavin.Hird actually the screen resolution is 5k
[11:01] Gavin.Hird 5120 x 2280
[11:02] Gavin.Hird 5120 x 2880
[11:02] Gavin.Hird Hi Bill
[11:02] Bill Blight: Howdy
[11:03] Gavin.Hird so usually when you render at say 1680 x 1050 the system will upsample it, but here I render at true screen resolution
[11:04] Kayaker Magic: Hmm, how about rendering it higher than the screen DPI and using the GPU to downsampe to get a higher quality image?
[11:05] Kayaker Magic: More CPU work, but better image without more bandwidth to the display
[11:05] Arielle Popstar: does it potentially increase perceived lag?
[11:05] Gavin.Hird you can do that too, but you get a very crips image when rendering at true screen rez on a 5k display
[11:05] Andrew Hellershanks: Hello, Arielle.
[11:05] Arielle Popstar: hey andrew
[11:06] Gavin.Hird FPS drops to about 1/3 of normal rendering
[11:06] Kayaker Magic: OUCH!
[11:06] Gavin.Hird so yes
[11:06] Kayaker Magic: But Moore's Law will save us....
[11:06] Gavin.Hird right now it does 26.7 FPS
[11:07] Arielle Popstar: i've noticed my FPS seems dependent on which monitor i have the viewer screen on
[11:07] Andrew Hellershanks: Hello, Sandy
[11:07] Sandy Beachcomber: hi
[11:07] Gavin.Hird that is probably true
[11:07] Gavin.Hird Hi Sandy
[11:08] Kayaker Magic: There should be a corollary to Moore's law: Every 18 months the speed of our processors will double, but every 18 months lazy programmers writing bloatware will cause aps to run 4 times slower.
[11:08] Andrew Hellershanks: Here comes Ubit.
[11:08] Arielle Popstar: ;
[11:08] Arielle Popstar: )
[11:08] Gavin.Hird Most Macs have pretty weak GPUs in the first place
[11:08] Sandy Beachcomber: hi Arielle
[11:08] Arielle Popstar: Hi Sandy
[11:08] Andrew Hellershanks grins at Kayaker.
[11:08] Sandy Beachcomber: hi Kayaker
[11:08] Gavin.Hird so I have turned Retina resolution rendering off as default
[11:09] Ubit Umarov: hi
[11:09] Gavin.Hird Hi Ubit
[11:09] Andrew Hellershanks: Hello, Ubit.
[11:09] Sandy Beachcomber: hi Gavin, Andrew
[11:09] Sandy Beachcomber: hi Ubit
[11:09] Gavin.Hird There is a version for Mac users to test at
[11:10] Kayaker Magic: My first question: How is YEngine doing? The documentation says "  ;; ONLY SUPORTS ONE REGION PER INSTANCE at this point". Has that been fixed?
[11:10] Gavin.Hird good question
[11:10] Andrew Hellershanks: I was just about to ask if you had a question, Kayaker. :)
[11:10] Kayaker Magic: Nope. More like 6 questions!
[11:11] Andrew Hellershanks: ok. We'll take them one at a time.
[11:11] Gavin.Hird you have 10 mins per question then
[11:11] Andrew Hellershanks: Unless some other people have some questions they want to tak.
[11:11] Andrew Hellershanks: s/tak/ask/
[11:11] Kayaker Magic: tick tick tick
[11:12] Sandy Beachcomber: i'm here to listen
[11:12] Ubit Umarov: well processors speed did stall
[11:12] Ubit Umarov: recently some intel push to 5ghz but... those numbers don't mean much
[11:13] Ubit Umarov: my 4-7ghz cpu is not that faster than my 3.7 one
[11:13] Kayaker Magic: Yeah, I was joking about being saved by Moore's!
[11:14] Andrew Hellershanks: Clock rate is not a good indication of which processor may be faster than another.
[11:14] Gavin.Hird it just waits faster
[11:14] Andrew Hellershanks: :)
[11:14] Kayaker Magic: lol
[11:14] Ubit Umarov: i still remember how a 50Mhz ibm kicked a 250 dec alpha in performance :)
[11:15] Ubit Umarov: well a few years ago.. coff cofff
[11:15] Andrew Hellershanks: Kayaker, People are usually advised to not run more than one region per instance. If they do so, the current limitation in YEngine is not an issue. I don't recall seeing any recent changes in code base that removes that limitation.
[11:16] Gavin.Hird I remember when our entire VAX cluster had 8 Mhz in total shared by about 700 concurrent users
[11:16] Ubit Umarov: and not much a limitation anymore
[11:16] Bill Blight: I'm not sure how anybody thinks running more than one region per instance was ever a good idea
[11:16] Ubit Umarov: just normal thing.. don't abuse on number of regions per instance
[11:17] Bill Blight: Like saying, how many regions can I crash in one shot
[11:17] Ubit Umarov: yeah not a good idea
[11:17] Kayaker Magic: OK, how about the YEngine documentation that says ";; DANGER, do not use with HG, don't leave regions running alone with it."? I have been using YEngine in an OSGrid region which is open to HG.
[11:17] Ubit Umarov: is there to scare you
[11:17] Ubit Umarov: :p
[11:18] Andrew Hellershanks: I've seen 9 regions being run in a single instance.
[11:18] Bill Blight: I've see 100 so there is that
[11:18] Ubit Umarov: tp in from x should now preserve the state
[11:18] Ubit Umarov: tp out to X will not
[11:18] Kayaker Magic: cool!
[11:18] Bill Blight: but still a bad idea from a stability standpoint,
[11:19] Andrew Hellershanks: Bill, 100 in one instance. I hope those were mostly water regions.
[11:19] Kayaker Magic: I wasn't even going to ask about state preservation X<->Y !
[11:19] Ubit Umarov: opsys does place limits per application etc etc etc
[11:19] Bill Blight: One bad thing on one region you crash all regions in that instance
[11:19] Gavin.Hird I have found 15 regions per instance works reasonably well and stable
[11:19] Ubit Umarov: and code is still not that good
[11:19] Bill Blight: You have not spent much time on the mantis have you Andrew
[11:20] Ubit Umarov: most devs only considered a single region case most time
[11:20] Andrew Hellershanks: Bill, I read the reports that hit my inbox.
[11:20] Sandy Beachcomber is scared
[11:20] Ubit Umarov: we are on Y here
[11:20] Kayaker Magic: Doesn't running several regions in one instance conserve server resources over running several instances?
[11:21] Ubit Umarov: think bill only uses Y now :)
[11:21] Bill Blight: Multiple regions per instance is just asking for trouble, and from a production standpoint is is like putting all your eggs in one basket, and hoping the bottom does not fall out ..
[11:21] Bill Blight: Only Y
[11:21] Ubit Umarov: may preserve some yes kayaker , but also mess it up also :p
[11:22] Bill Blight: and it some cases an instance with 10 regions will use more resources than 10 separate instances
[11:22] Andrew Hellershanks: Kayaker, It can save a bit of memory as you avoid having several copies of the OpenSim code loaded but that doesn't use a lot of memory.
[11:22] Ubit Umarov: well im running 2 regions per instance on my small test grid and seem ok
[11:22] Gavin.Hird overhad in running multiple instances of mono
[11:22] Bill Blight: which is not much
[11:23] Bill Blight: well worth the safety net of stability at not crashing a dozen regions when one has issues
[11:23] Andrew Hellershanks: Right. If something goes wrong in one region it can, and often will, affect all other regions running in the same instance.
[11:24] Gavin.Hird I find mono does a pretty good job utilizing the processor available on the system when you have multiple regions, but of course if the instance goes down...
[11:24] Ubit Umarov: also less load isolation
[11:24] Bill Blight: might as well say it WILL affect them, there is no might about it,
[11:24] Ubit Umarov: not that the OpSys does add more
[11:24] Andrew Hellershanks: It also means restarting a region means restarting all of the ones being run by the instance at the same time.
[11:24] Bill Blight: yep , no rolling restarts
[11:24] Andrew Hellershanks: That can be very disruptive in a grid.
[11:25] Ubit Umarov: the worse is our own code, as i said many devs only considered single region case
[11:25] Bill Blight: There is a lot of stuff in the code that does not like to be shared
[11:25] Bill Blight: amazed it works at all
[11:25] Gavin.Hird you can restart a single region in a multi region instance, but of course it will be on the same codebase, so no rolling upgrade
[11:25] Ubit Umarov: we had bad modules sharing, mixing thigns from dif regions etc
[11:25] Arielle Popstar: if they all belong to different people but not such a biggie if all for the same person
[11:25] Ubit Umarov: NO NO
[11:26] Ubit Umarov: restart one regions is a NO NO
[11:26] Bill Blight was waiting for Ubit to blow his stack on that one
[11:26] Arielle Popstar: ?
[11:26] Ubit Umarov: restart the entire instance or none
[11:26] Andrew Hellershanks: I'm not sure how well that inworld restart feature works.
[11:26] Ubit Umarov: andrew i already answered that :p
[11:27] Arielle Popstar: i'm referring to 10 regions in one instance vs 10 seperate instances
[11:27] Andrew Hellershanks: Ubit, you typed that response faster than I did mine. :)
[11:27] Gavin.Hird it works (most of the time) as a stop-gap till it is more convenient to restart the instance, but it might also crash the entire instance
[11:28] Ubit Umarov: it just fakes it works :p
[11:29] Sandy Beachcomber: how many instances can a robust support?
[11:29] Ubit Umarov: restart the entire instance
[11:29] Bill Blight: depends on your robust
[11:29] Bill Blight: How many does osGrid support
[11:29] Bill Blight: thousands of regions
[11:29] Andrew Hellershanks: IIRC, there is a menu entry in the viewer for requesting the restart of a region. Might be a good idea to remove that entry.
[11:30] Arielle Popstar: depends how many avi's on each
[11:30] Sandy Beachcomber: osgrid robust has distributed services?
[11:30] Gavin.Hird why, it also works fine on single region instances
[11:30] Kayaker Magic: I had several dual region instances for quick testing of border crossing. OK, I'll never do multiple regions/instance again!
[11:30] Bill Blight: actually avies on the region don't have that big of an affect on robust once they are logged in and have their inventory
[11:30] Kayaker Magic: What was the resolution for YEngine and HG together?
[11:31] Sandy Beachcomber: lol its a lot of code updates
[11:31] Bill Blight: Robust does the heavy lifting on login and inventory delivery, once that is done the regions handle it from there mostly
[11:31] Sandy Beachcomber: i suppose it can be automated
[11:31] Bill Blight: I run one script that updates all my instances
[11:32] Sandy Beachcomber is learning
[11:33] Andrew Hellershanks: Kayaker, scroll back a ways. :)
[11:33] Bill Blight: people HG to this meeting all the time and it runs Y
[11:33] Gavin.Hird but now that I can run a region on a Pi, the cost for HW per region is pretty slim
[11:33] Bill Blight: you can run 100 or so instances on a good modern server
[11:34] Arielle Popstar: can you run a 30 person event on it?
[11:34] Andrew Hellershanks: Gavin, nebadon has run OS on a Pi for some time.
[11:34] Bill Blight: (well on linux)
[11:34] Gavin.Hird no you can't
[11:34] Ubit Umarov: and all 100 empty :p
[11:34] Bill Blight: no Ubit
[11:34] Gavin.Hird but you can run a var with a few avatars and quite a bit of pirms
[11:34] Arielle Popstar: oh did he have wright plaza on a Pi?
[11:34] Gavin.Hird which is what most people do for opensim
[11:35] Andrew Hellershanks: I have run up to about 130 regions on a single server, each in their own instance..
[11:35] Ubit Umarov: empty of avatars i meant
[11:35] Bill Blight: but when I say modern server, I mean something with a LOT of ram and major CPU's
[11:35] Sandy Beachcomber: example Bill?
[11:35] Gavin.Hird it is up and running on for visitors
[11:36] Bill Blight: but 30 - 50 instances don't seem to make my 6 core opterons with 64 gig breathe heavy
[11:36] Arielle Popstar: i've run a 100 region instance on a dual core 2 GB machine before
[11:36] Andrew Hellershanks: You need lots of RAM, and decent CPU's. Depends on how you class "major" CPUs.
[11:37] Arielle Popstar: depends what you using it for
[11:38] Arielle Popstar: i just used it for car racing for 3-4 people
[11:38] Gavin.Hird or lots of ocean for sailing
[11:38] Arielle Popstar: that too
[11:38] Gavin.Hird filler sims
[11:38] Arielle Popstar: nod
[11:38] Andrew Hellershanks: 100 regions with only 2GB of RAM? Most of those regions must have been empty.
[11:39] Arielle Popstar: pretty much
[11:39] Sandy Beachcomber: big vars would be better
[11:39] Bill Blight: that math does not work out though, even running in one instance, that is only allowing 2meg of ram for each region, that is just silly
[11:39] Arielle Popstar: less to run into that way
[11:39] Sandy Beachcomber: fewer region crossings?
[11:39] Arielle Popstar: nope
[11:40] Andrew Hellershanks: hm... with var regions that would be a 2560x2560.
[11:40] Arielle Popstar: one instance on .8.2 required 100 MB with each additional region on the instance needing 10 MB
[11:40] Bill Blight: *20mb
[11:40] Arielle Popstar: so empty 100 region var or mega only needs about 1100mb
[11:40] Andrew Hellershanks: 100MB seems a bit low.
[11:40] Bill Blight: where did you ever get that info
[11:40] Arielle Popstar: from task manager
[11:41] Bill Blight: LMAO
[11:41] Ubit Umarov: lol a single region on 8.2 is 600mb :p
[11:41] Andrew Hellershanks: I was going to say you should allow at least 300 to 500 Meg per region, minimum.
[11:41] Arielle Popstar: maybe in your world :)
[11:41] Bill Blight: maybe in the real world
[11:41] Ubit Umarov: memory usage is not that simple to see lol
[11:41] Andrew Hellershanks: I have seldom seen regions using less than around 300MB.
[11:41] Bill Blight: and windows is horrible at reporting mem usage right
[11:42] Ubit Umarov: both are
[11:42] Bill Blight: true
[11:42] Bill Blight: but on linux you can dig for more info
[11:42] Andrew Hellershanks: Right
[11:42] Bill Blight: But you can't argue with the expert
[11:43] Ubit Umarov: this one thinks it is using 357MB
[11:43] Arielle Popstar: just started up an instance attached to osgrid and it is using 64 MB
[11:43] Bill Blight: surrrre it is
[11:44] Ubit Umarov: my test linux
[11:44] Ubit Umarov: MEMORY STATISTICS
Heap allocated to OpenSim  : 58 MB
Heap allocation rate (last/avg): 0/0.001MB/s
Process memory: Physical 195 MB Paged 0 MB Virtual 5453 MB
[11:44] Ubit Umarov: my test win
[11:44] Ubit Umarov: MEMORY STATISTICS
Heap allocated to OpenSim  : 58 MB
Heap allocation rate (last/avg): 0/0.001MB/s
Process memory: Physical 195 MB Paged 0 MB Virtual 5453 MB
[11:45] Kayaker Magic: I'm using up more than my alloted 10 mins per question... I'm going to squeeze in another one:
[11:45] Bill Blight: and you see this part 5453 MB
[11:45] Kayaker Magic: I probably should submit a mantis for this one: In the Statistics bar (ctrl-shift-1) the Script Events number is always 0 in YEngine.
[11:45] Ubit Umarov: identical instances, 2 regions on var 1km
[11:45] Bill Blight: that is what will be swapped in and out of actuall ram
[11:45] Ubit Umarov: oops
[11:45] Ubit Umarov: failed the win paste
[11:46] Ubit Umarov: MEMORY STATISTICS
Heap allocated to OpenSim  : 56 MB
Heap allocation rate (last/avg): 0.065/-0.013MB/s
Process memory: Physical 167 MB Paged 146 MB Virtual 4893 MB
[11:46] Bill Blight: paged and virtual are not nothing, they are what get swapped out to HD but will swap back to real ram on load, so you have to take that into memory usage calculations
[11:47] Ubit Umarov: win in 32bit
[11:47] Gavin.Hird the 2x2 var restarted on the PI (debian strech) reports
[11:47] Ubit Umarov: MEMORY STATISTICS
Heap allocated to OpenSim  : 41 MB
Heap allocation rate (last/avg): 0.087/0.089MB/s
Process memory: Physical 137 MB Paged 114 MB Virtual 465 MB
[11:47] Gavin.Hird MEMORY STATISTICS
Heap allocated to OpenSim  : 23 MB
Last heap allocation rate  : 0 MB/s
Average heap allocation rate: 0 MB/s
Process memory: Physical 95 MB Paged 0 MB Virtual 237 MB
Peak process memory: Physical 138 MB Paged 0 MB Virtual 250 MB
[11:47] Ubit Umarov: not using warp3d :p
[11:47] Bill Blight: hehe
[11:47] Gavin.Hird not using warp3
[11:47] Ubit Umarov: Peak process memory: Physical 486 MB Paged 0 MB Virtual 5517 MB~
[11:47] Ubit Umarov: mine with warp3d :)
[11:48] Gavin.Hird then it was about 100 MB more
[11:48] Andrew Hellershanks: Kayaker, Yes, please do file a mantis about that issue so it doesn't get lost.
[11:48] Andrew Hellershanks: If you are using warp3d for map tiles memory requirements go up.
[11:48] Bill Blight: This region ..
[11:48] Bill Blight: MEMORY STATISTICS
Heap allocated to OpenSim  : 114 MB
Heap allocation rate (last/avg): 0.008/0.012MB/s
Process memory: Physical 358 MB Paged 0 MB Virtual 4723 MB
Peak process memory: Physical 410 MB Paged 0 MB Virtual 4764 MB
[11:48] Ubit Umarov: in 32bit Peak process memory: Physical 348 MB Paged 328 MB Virtual 664 MB
[11:49] Ubit Umarov: ahh both in release mode
[11:49] Gavin.Hird so I start the instance to generate maps, then restart and the maps will not be generated on the second startup and you save about 700 MB for an instance with 15 regions
[11:49] Arielle Popstar: try windows
[11:49] Ubit Umarov: numbers in debug are dif
[11:49] Kayaker Magic: Should I worry about high Script Events numbers from my scripts? I have a big script that only runs on a 1 sec timer evet, but then it sends a bunch of llMessageLinked messages that wake up events in other scripts, most of which just check "is that for me?" then exit. Does several hundred Script Events a second put undue load on OpenSim?
[11:50] Ubit Umarov: of course :p
[11:50] Ubit Umarov: on x that fills up all x engine threads each with several jobs
[11:50] Andrew Hellershanks: Kayaker, I would worry more about what activity is triggered by the hundred events rather than the actual events.
[11:51] Ubit Umarov: amasing how ppl forget the cost of "trigger" lol
[11:51] Gavin.Hird interrupt handling and context switching is what kills performance in the end
[11:52] Kayaker Magic: Most of the events do little but check and exit.
[11:52] Ubit Umarov: then "trigger" costs a lot more than your code
[11:52] Andrew Hellershanks: Kayaker: Is that being done in one of your sailing objects?
[11:52] Kayaker Magic: Yup
[11:53] Andrew Hellershanks: I can see that being a big problem if you have several people operating sailboats in the same region at the same time.
[11:53] Bill Blight: yep
[11:53] Kayaker Magic: I chose to send a message to all children, so you could have any number of sails (for example) and they would all respond.
[11:55] Ubit Umarov: situation on Y is a bit dif.. there are only 2 exec threads per region ( currently )
[11:55] Andrew Hellershanks: Sending a messaging to all keeps the code simple. You may be better off doing a query on script reset to determine the number of sails and note the prim numbers. You can adjust the list as needed as someone raises/lowers a sail.
[11:55] Ubit Umarov: so all those events wait i queue for their timeslice
[11:56] Bill Blight: Back to Kayaker's previous question Royale's grid runs nothing but Y, and HG works fine
[11:56] Bill Blight: as well as mine
[11:56] Royale.Mobian Yeah Bill is right.
[11:57] Ubit Umarov: Y doing plain LSL seems ok
[11:57] Sandy Beachcomber: how do you run Y engine?
[11:57] Arielle Popstar: whats the advantage of Y?
[11:57] Royale.Mobian we stopped using XEngine when Y was still XMR
[11:57] Ubit Umarov: error reporting on compile can be cryptic :)
[11:57] Andrew Hellershanks: I was just about to ask if anyone else has question before we get to more of Kayaker's questions. I just noticed how we are at the top of the hour.
[11:58] Ubit Umarov: yes.. but Y is not xmr !!!
[11:58] Bill Blight: Not anymore
[11:58] Bill Blight: red headed step child
[11:58] Ubit Umarov: i did not just renamed it :p
[11:58] Kayaker Magic: YEngine fixes some important design flaws in X, for example llSleep in X locks up a thread and can bring a region to its knees.
[11:58] Ubit Umarov: yes that is a major dif
[11:59] Bill Blight: and just feels faster in script execution
[11:59] Arielle Popstar: tyhat couldnt be fixed in X?
[11:59] Kayaker Magic: no
[11:59] Kayaker Magic: design flaw from the bottom up
[11:59] Ubit Umarov: yes the fixe was named xmr :p
[11:59] Royale.Mobian ubit i was saying we used it before the change into y happened. wasn't saying they're the same.
[11:59] Bill Blight: Like fixing a steam engine, to be like a jet engine
[11:59] Sandy Beachcomber: Sandy Beachcomber: how do you run Y engine?
[11:59] Ubit Umarov: or aurora script engine or phalox
[12:00] Ubit Umarov: all fixes to x :p
[12:00] Ubit Umarov: well "fixes" :)
[12:00] Andrew Hellershanks: Sandy, it is a configuration option in the ini files.
[12:00] Kayaker Magic: YEngine is a new option in OpenSim.ini
[12:00] Sandy Beachcomber: ok ty
[12:00] Sandy Beachcomber: not spotted that
[12:00] Ubit Umarov: don't forget to set the default engine also
[12:00] Kayaker Magic: Only on recent versions of OpenSim
[12:00] Arielle Popstar: all scripts would have to be redone if one was to use it?
[12:01] Bill Blight: no, only scripts with a few incorrect syntax issues
[12:01] Kayaker Magic: All scripts SHOULD work, if not SUBMIT A MANTIS!
[12:01] Ubit Umarov: yes like always
[12:01] Royale.Mobian Bill is right
[12:01] Arielle Popstar: lol
[12:01] Bill Blight: there are some syntax differences
[12:01] Ubit Umarov: no kayaker
[12:01] Bill Blight: things that X allowed
[12:01] Bill Blight: that Y doesn't
[12:01] Ubit Umarov: some will only work o X ever
[12:02] Arielle Popstar: so it isnt backwards compatible
[12:02] Bill Blight: for instance list mylist = ""; was valid in X, but in Y it has to be the correct list mylist = [];
[12:02] Andrew Hellershanks: Y includes fixes for some parsing issues that affect X.
[12:02] Ubit Umarov: and like at sl, ppl just code testing this and that ending up with code very specific to the engine
[12:02] Bill Blight: just as an example
[12:02] Kayaker Magic: I can live with the syntax changes, often they are bugs that X didn't catch.
[12:03] Ubit Umarov: without options Y is more like SL sintaxe
[12:03] Sandy Beachcomber: Ari asked about advantages?
[12:03] Ubit Umarov: it runs SL scripts directly a lot better
[12:03] Bill Blight: or just lazy syntax
[12:03] Kayaker Magic: Or type conversions I would rather have the compiler force me to make explicit.
[12:04] Andrew Hellershanks thinks this should be extracted and turned in to a wiki page for :)
[12:04] Ubit Umarov: Y will need a full section
[12:04] Bill Blight: Y , more sl Like, faster, many caveats from X are corrected
[12:04] Gavin.Hird I would say so
[12:04] Ubit Umarov: specially when its extentions are defined
[12:04] Bill Blight: yeah those extensions, need documentation
[12:04] Arielle Popstar: the questions i asked were from people who wondered about the whole reasoning behind Y engine so hopefully they read the log and will see the ansers
[12:04] Ubit Umarov: a subset ov xmr ones
[12:04] Bill Blight: a LOT of it
[12:05] Bill Blight: the safe ones
[12:05] Bill Blight: LOL
[12:05] Gavin.Hird does it roun the security orb script that is one of the top downloads on outworldz?
[12:05] Gavin.Hird run*
[12:05] Ubit Umarov: like arrays, continue, break, etc
[12:05] Ubit Umarov: you tell us gavin :p
[12:06] Gavin.Hird it runs on X, but does not do anyting
[12:06] Andrew Hellershanks: Now and then I have missed not having switch or break when writing scripts.
[12:06] Gavin.Hird which is why I wounder why it is on the top download
[12:06] Arielle Popstar: so is development on Y finished?
[12:07] Ubit Umarov: no
[12:07] Kayaker Magic: I've never seen a script I could not get to work on YEngine with a few minor syntax changes. I have seen a lot of people who see "type conversion not specified" and then give up saying "This script does not work!"
[12:07] Ubit Umarov: needs better error reporting
[12:07] Andrew Hellershanks: Gavin: number of downloads doesn't equate to how much it is actually being used.
[12:07] Bill Blight: nothing .9x is finished
[12:07] Bill Blight: 8x was not finished until .9x was released
[12:07] Ubit Umarov: and the options need cleaning up.. better definition etc
[12:07] Gavin.Hird absolutely Andrew
[12:07] Ubit Umarov: ( to ignore for now.. use it with plain LSL )
[12:08] Arielle Popstar: is it far enough along that a release could be considewred?
[12:08] Bill Blight: 9x will not be finished until it is replaced with the next version
[12:08] Ubit Umarov: some of the xmr ones are there, but may change
[12:09] Andrew Hellershanks: Does anyone else, besides Kayaker, have any questions?
[12:09] Andrew Hellershanks: It won't be long now before people start having to leave.
[12:09] Kayaker Magic: LOL, I'll just do one more!
[12:09] Kayaker Magic: I'm working on a way to test the piss poor region startup time from Warp3D. Here is my testing methodology: I'm measuring startup time in an empty region, then rezzing 1000 items in random locations, measuring again, rezzing 1000 more and repeat. I'm hypothesizing the time will go up faster than linear with number of objects. Plot a graph. Does this sound like a good plan?
[12:10] Bill Blight: depends on the objects
[12:10] Ubit Umarov: sounds a waste of time :p
[12:10] Gavin.Hird hehe
[12:10] Kayaker Magic: would prims vs mesh be an issue?
[12:10] Bill Blight: Mesh will slow it down , piss poor mesh will slow it down exponentially
[12:10] Gavin.Hird mesh will slow it down in my experience
[12:10] Andrew Hellershanks: I've often thought it would be nice to have an external version of Warp3D to generate maptiles. I know that is not really going to happen.
[12:11] Ubit Umarov: it needs to convert all to warp3d
[12:11] Kayaker Magic: bad mesh is epidemic in the Metaverse.....
[12:11] Bill Blight: plague
[12:11] Ubit Umarov: but BIG time is when assetscache is empty
[12:11] Bill Blight: omg yes
[12:11] Andrew Hellershanks: What is needed is better reporting of bad mesh. Current error messages don't make it easy to identify offending objects.
[12:11] Bill Blight: what is needed is to not make bad mesh
[12:11] Ubit Umarov: it will read almost all meshes and textures
[12:11] Gavin.Hird Warp3D for Robust that could execute on itself and just generate map tiles
[12:12] Kayaker Magic: How about an option in Warp3D to just draw a single pixel for each mesh object and not try to display it?
[12:12] Ubit Umarov: and that can take ages :)
[12:12] Bill Blight: how about turn off render meshes
[12:12] Andrew Hellershanks: Bill, that would be nice.
[12:12] Ubit Umarov: ( well it is time is needed in any case, but without warp is as users need the assets
[12:12] Sandy Beachcomber: dare i ask about server side baking?
[12:12] Ubit Umarov: will wapr will ask for all at start
[12:13] Arielle Popstar: thought i read that LL fixed their bad meshes by using a script to clean them up and make it more efficient. That way they didnt haeve to break content
[12:13] Bill Blight: no
[12:13] Arielle Popstar: cant do the same here?
[12:13] Bill Blight: HAHAHA
[12:13] Ubit Umarov: no one can do that lol
[12:13] Bill Blight: their multi million dollar physics engine does that
[12:13] Ubit Umarov: just look to the mess viewers do on some uploads
[12:13] Gavin.Hird they control the db, in opensim meshes comes from all over
[12:14] Bill Blight: we don't have a multi-million dollar physics engine
[12:14] Arielle Popstar: ok whatever it was they did do that then
[12:14] Ubit Umarov: and THEY REFUSE a lot of uploads
[12:14] Bill Blight: yep
[12:14] Arielle Popstar: we have Ubit!
[12:14] Ubit Umarov: just try fs with the new checks on and see
[12:14] Bill Blight: which is more apparent in the new FS
[12:14] Sandy Beachcomber: maybe our code should?
[12:14] Ubit Umarov: most those new fs checks are like sl servers ones
[12:14] Bill Blight: FS by default is attempting to enforce the LL mesh standards these days
[12:14] Sandy Beachcomber: efuse bad mesh?
[12:15] Gavin.Hird latest Dayturn is stricter too
[12:15] Bill Blight: and of course Opensimulator is getting the blame
[12:15] Ubit Umarov: yeak and fs refuses to set the size of a prim more than 256m
[12:15] Ubit Umarov: that is irritating
[12:15] Arielle Popstar: people are already reverting to .8.2 instances to upload meshes that wont pass inspection\
[12:16] Bill Blight: and that is why bad mesh is a plague
[12:16] Sandy Beachcomber: i see
[12:16] Arielle Popstar: then oar'ing them over
[12:16] Ubit Umarov: yes some people is just stupid.. nothing to do on that
[12:16] Bill Blight: because instead of fixing your mesh, you go to a place you can upload crap
[12:16] Arielle Popstar: well not everyone is a blender expert
[12:16] Gavin.Hird copybotters are hardly able to fix their meshes
[12:16] Ubit Umarov: bad mesh means BAD
[12:16] Ubit Umarov: gezz
[12:17] Bill Blight: then maybe you should not be uploading mesh
[12:17] Ubit Umarov: how stupid some ppl can be
[12:17] Bill Blight: and blaming opensim because your mesh sucks
[12:17] Ubit Umarov: well sorry folks but some things are just.. well stupid :p
[12:17] Andrew Hellershanks: It isn't just a question of how you use Blender but also how the mesh objects were uploaded.
[12:17] Ubit Umarov: and 0.91 does almost ZERO tests
[12:17] Bill Blight: People used to think tobacco was good for them, they learned, better, same with bad mesh, learn or die
[12:18] Arielle Popstar: seems more an art then a skill considering the lack of documentation
[12:18] Arielle Popstar: or confusing documentation
[12:19] Arielle Popstar: anyway maybe could consider how to fix the mesh on upload
[12:19] Gavin.Hird top creators don't want to document it for competitive reasons
[12:19] Bill Blight: there are millions of pages of documentation on making good mesh for 3d games, and most of it , for low poly, can be applied to opensimulator, just a matter of the desire to learn
[12:19] Ubit Umarov:
[12:19] Arielle Popstar: who going to go through millions of pages
[12:19] Andrew Hellershanks: Avoid the confusion by reading the information on the site and ignore the many variations on upload procedure posted elsewhere.
[12:19] Ubit Umarov: fs dev talking about the new checks
[12:20] Ubit Umarov: ( we don't fill up those detailed costs )
[12:20] Bill Blight: people who want to learn to do it right will
[12:20] Bill Blight: others are just lazy and stupid
[12:20] Ubit Umarov: ( to do that would need to check meshes a bit more and REFUSE more :p )
[12:20] Arielle Popstar: hows the animesh code coming?
[12:20] Gavin.Hird agreed Bill, but it does not necessarily come asy and most people don't have the time to invest
[12:20] Bill Blight: then they need to stop
[12:21] Bill Blight: anything worth doing is worth doing right
[12:21] Gavin.Hird Ariell, there is animesh on the table
[12:21] Bill Blight: and outside
[12:21] Ubit Umarov: aniwhat seems done for now
[12:21] Kayaker Magic: Arielle: No more questions! Time is up! :)
[12:21] Andrew Hellershanks: They don't have the time to invest then everyone else suffers.
[12:21] Arielle Popstar: cool
[12:21] Gavin.Hird Dayturn has it now both for windows and mac version
[12:21] Bill Blight: yes andrew
[12:21] Bill Blight: but they don't care
[12:21] Bill Blight: ME ME ME ME
[12:22] Arielle Popstar: if they only running it on their grid then it really doesnt matter does it?
[12:22] Arielle Popstar: they will have to invest in the extra hardware
[12:22] Bill Blight: it matters when they hand the crap out to everyone and that crap ends up on other peoples grids
[12:22] Arielle Popstar: how do you tell a bad mesh if inworld?
[12:22] Bill Blight: extra hardware?
[12:23] Arielle Popstar: well assume thats what is going to be needed if a mesh is bad to render it
[12:23] Bill Blight: in world? Bad LOD's overly heavy on physics, huge display weight, improper physics models
[12:23] Arielle Popstar: more ram, more cpu
[12:23] Bill Blight: many a way to tell
[12:23] Ubit Umarov: beq also has something about the fs tools to check meshs
[12:23] Ubit Umarov: on that blog
[12:23] Bill Blight: yep
[12:24] Andrew Hellershanks: Arielle, it often shows up on the maptiles as a large U shaped object.
[12:24] Bill Blight: new fs has things to visualize some of it
[12:24] Ubit Umarov: how to check LODs and physics shape
[12:24] Gavin.Hird In Dayturn, on the Develop, Render Metadata menu there is a Triangle Count menu item that will give you one idea of the quality of the mesh
[12:24] Arielle Popstar: maybe once people learn how to recognize what is bad, they will be quicker to get rid of it
[12:24] Gavin.Hird FS should have the same
[12:24] Ubit Umarov: ( that physics is almost useless if using bullet.. it will not match in several cases )
[12:25] Sandy Beachcomber: i agree Ari
[12:25] Ubit Umarov: gavin consider adding those tools on edit
[12:25] Bill Blight: how selfish and self centered is it to say, "I don't want to learn how to make something right, and just shove whatever crap I can get in world there, and let others deal with it. Because I don't care if it causes them problems."
[12:25] Sandy Beachcomber: we need to be able to recognise it
[12:25] Ubit Umarov: they are damm nice
[12:25] Ubit Umarov: and the singu http usage also
[12:26] Arielle Popstar: how selfish and self centered to expect everyone else to be at your level of expertise or have the time to invest to be so
[12:26] Gavin.Hird the next windows version will have better cache stats too
[12:26] Andrew Hellershanks: Triangle count doesn't come in to play in the modelling program I use. I don't have any idea of triangles until I get it in to Blender. At that point I need to use Blender tools to reduce the count.
[12:26] Gavin.Hird the Mac version already has it
[12:26] Ubit Umarov: to be able to check LODs without having to fly around is cool
[12:26] Ubit Umarov: and see a simgle prim physics shape also
[12:27] Gavin.Hird yes, there is a lot of goodies there
[12:27] Bill Blight: not at all Miss "Always Right", expecting others to respect the experience of others and not force their crap on them , is a respect issue, not an education issue.
[12:27] Bill Blight: Have a little pride in your creations
[12:27] Arielle Popstar: maybe Bill, people just dont realize the effect it has on others
[12:27] Sandy Beachcomber: but we need to be able to find the bad mesh
[12:27] Bill Blight: bad mesh is easy to find, with viewer tools
[12:27] Sandy Beachcomber: example?
[12:27] Bill Blight: we just talked about that
[12:27] Bill Blight: read back
[12:28] Arielle Popstar: now that the viewers are getting the tools TO recognize it
[12:28] Bill Blight: you could find most of it before with the "More Info" tool
[12:28] Royale.Mobian I know plenty of people who make bad mesh and know its bad and don't care because it don't effect them
[12:28] Bill Blight: you select a mesh and it has a million physics and a 2 million display weight, that is kind of bad
[12:28] Andrew Hellershanks: Hm... one option for the viewer is to question the user about uploading a model if the triangle count of the physics model is the same as, or a significant percentage of, the triangle count of the model.
[12:29] Gavin.Hird or even higher :-)
[12:29] Royale.Mobian there may be people who don't know its bad but there is also those who know and don't care
[12:29] Arielle Popstar: Nods @ Royale
[12:29] Bill Blight: 90% of this bar is mesh, it has a physics of under 1000 and a display of 20k .. have seen a cube with 100x that
[12:30] Royale.Mobian same here Bill
[12:30] Bill Blight: have seen mesh trees with a million display weight for ONE Tree
[12:30] Arielle Popstar: well then maybe just need to get the word out to people what to look for
[12:30] Ubit Umarov: and the lods do nice trasitions also
[12:30] Bill Blight: the word has been out for years
[12:30] Sandy Beachcomber: examines the bar
[12:30] Bill Blight: just people never listened
[12:30] Royale.Mobian ^
[12:30] Ubit Umarov: unlike most on opensim
[12:31] Andrew Hellershanks: We are making up for last weeks short meeting. :)
[12:31] Bill Blight: people do things like, "I'm not going to learn to fix my mesh, I'm just going to use a version that lets me upload the crap"
[12:31] Arielle Popstar: told Ubit i was going to come because the meetings were getting too short :p
[12:31] Ubit Umarov: yes you did :)
[12:32] Gavin.Hird :-)
[12:32] Gavin.Hird :-)
[12:32] Selby.Evans Must go -- bye all.
[12:32] Arielle Popstar: tc Selby
[12:32] Ubit Umarov: well one was longer.. but log was censored :p
[12:32] Royale.Mobian I had a modeler who used to be on my grid that got mad when we moved to ubode because he used that physics bug in bullet
[12:32] Bill Blight: when people keep spewing and supporting nonsense like, "Just use an older version it will let you upload your crap" that does not help the case
[12:32] Andrew Hellershanks: I was wondering when we would start to lose people. Any last minute questions before we wrap up todays meeting?
[12:32] Sandy Beachcomber: if grid owners knew what to look for in bad mesh it would be in their interest to eliminate it
[12:32] Arielle Popstar: oh wonder what that was about
[12:33] Bill Blight: I do eliminate it
[12:33] Sandy Beachcomber: i would
[12:33] Arielle Popstar: figured there was censoring going on by the time stamps
[12:33] Sandy Beachcomber: but i dont make mesh
[12:33] Royale.Mobian grid owners do if they know about it but it shouldn't be there job to do so
[12:33] Sandy Beachcomber: yet
[12:33] Ubit Umarov: ( good censoring.. )
[12:33] Ubit Umarov: andrew take note.. fix timestamps next time :)
[12:33] Kayaker Magic: Kitely has some sort of bad mesh detector that disables and notifies users.
[12:34] Sandy Beachcomber: i think grid owners would be happy to eliminate it
[12:34] Andrew Hellershanks grins at Ubit.
[12:34] Bill Blight: When people spread nonsense on how to use Opensimulator, and they are loud enough people then do that, and blame Opensimulator for not working right, when they are listening to people who have no clue ..
[12:34] Ubit Umarov: well fs has it also now
[12:34] Ubit Umarov: petty it does make almost impossible to upload anything :p
[12:34] Gavin.Hird You probably can do it in Excel with fill down on the timestamp column
[12:35] Arielle Popstar: Bill talking about me
[12:35] Sandy Beachcomber: a simple tool would be good!
[12:35] Bill Blight: yep
[12:35] Arielle Popstar: ;)
[12:35] Bill Blight: and others
[12:35] Ubit Umarov: that white jeep is also a exemple of a not so good meshs
[12:35] Bill Blight: right
[12:36] Ubit Umarov: look to the top bar as disntace changes
[12:36] Bill Blight: some of the parts have horrible LODs
[12:36] Ubit Umarov: yeap
[12:36] NPC 2014 Jeep Wrangler - Driveable - Beach Version v6: :: is tuned for ubODE
[12:36] Ubit Umarov: all just use viewers first try
[12:36] Ubit Umarov: and that is not good most time
[12:36] Sandy Beachcomber: i would like to right click it and be able to see if it was good bad or ugly
[12:36] Kayaker Magic: WOW! Great meeting! I have to go now....
[12:36] Arielle Popstar: tc Kay
[12:37] Sandy Beachcomber: bye Kayaker
[12:37] Andrew Hellershanks: ok, Kayaker. We can get to your last two questions next week.
[12:37] Ubit Umarov: on this LOD just zoom in and out
[12:37] Ubit Umarov: with good LODs we should see no changes, or just minor
[12:38] Arielle Popstar: Bill, is that sim stat script available somewhere?
[12:38] Bill Blight: it is just a combination of a couple I found and modded
[12:38] Bill Blight: I'll make it copy/mod, gimme a second
[12:38] Arielle Popstar: kk ty
[12:39] Andrew Hellershanks: sim stat scripts are easily found
[12:39] Ubit Umarov:
[12:39] Arielle Popstar: oh you dont think he would be seeing it?
[12:39] Ubit Umarov: beq on LODs
[12:39] Bill Blight: ok should be copy now
[12:39] Ubit Umarov: ( beq was the dev that did most mesh changes on FS )
[12:39] Gavin.Hird I need to hop to take care of something RL
[12:40] Arielle Popstar: yes but that one is better then any others i have seen
[12:40] Gavin.Hird have a great week everyone
[12:40] Arielle Popstar: thank you Bill
[12:40] Sandy Beachcomber: ty Bill
[12:40] Bill Blight: np
[12:40] Bill Blight: see my colored clock on the bar center Ubit
[12:40] Bill Blight: LOL
[12:40] Arielle Popstar: going to put it on the 68 MB region and see if it agrees
[12:40] Bill Blight: told you I was going to make a clock out of the new function
[12:41] Arielle Popstar: waves and poofs
[12:41] Ubit Umarov: yeah i did add a ossl thing to give PST/PDT time
[12:41] Royale.Mobian later everyone.
[12:41] Ubit Umarov: since so many want it and not so good scripts doing it
[12:41] Andrew Hellershanks: Bye, Arielle, and Royale.
[12:42] Bill Blight: it is an easy thing to do in scripts but the new function is welcome
[12:42] Sandy Beachcomber: bye
[12:42] Ubit Umarov: or making or server believe it is in california :)
[12:42] Bill Blight: that too
[12:42] Sandy Beachcomber: thanks for the education everyone
[12:42] Andrew Hellershanks: Bye, Sandy.
[12:43] Andrew Hellershanks: Time to wrap up todays meeting. See you all again next week.

Personal tools
About This Wiki