User:Andrew Linden/Office Hours/2008 02 05

From Second Life Wiki
Jump to: navigation, search

Transcript of Andrew Linden's office hours:

[11:02] Arawn Spitteler: I've an idea for a circular bench, that grows with occupation, but have to figure how to turn people about.
[11:02] Teravus Ousley: Hello everybody :D I'm actually in two meetings unfortunately ;P
[11:02] Arawn Spitteler: I think I just solved it, though.
[11:02] Kitto Flora: Hello Andrew
[11:02] Andrew Linden: Hi everybody.
[11:02] Teravus Ousley: Hello :D
[11:03] Anders Falworth: Hi, Andrew
[11:03] Ruud Lathrop: Hi
[11:03] Rex Cronon: hi
[11:03] Arawn Spitteler grovels politely in the discretely general direction of our new arrival
[11:03] Andrew Linden: I'm in a bad mood today, because I've been trying to fix @#$%^& avatar motion bugs
[11:03] Andrew Linden: that code is going to be the death of me
[11:04] Arawn Spitteler hopes he isn't serious, about catching his death of Code: Live by hte sword, and all
[11:04] Rex Cronon: u should buy a punch bag. helps u releave stress:)
[11:04] Kitto Flora: Whats buggy about avatar motion?
[11:04] Ruud Lathrop: Or grap a gun and go shoot somebody in SL :)
[11:05] Teravus Ousley: yes, or a pillow .. or at least a fax machine that you can bash in the parking lot
[11:05] Andrew Linden: we have a new update, held up by some last minute QA stuff going on right now
[11:05] Andrew Linden: hopefully today
[11:05] Rex Cronon: yes, go to a fighthing sim and create havok:)
[11:05] Ruud Lathrop: new update of server?
[11:05] Andrew Linden: yeah, new Havok4 preview update
[11:06] Andrew Linden: There will also be a viewer update sometime after that, in a few days
[11:06] Andrew Linden: but it is pending some MONO compatibility UI stuff
[11:06] Kitto Flora: Hi Simon
[11:07] Simon Linden: Hi folks
[11:07] Ruud Lathrop: Hi
[11:07] Rex Cronon: hi simon
[11:07] Anders Falworth: Hi, Simon.
[11:07] Rex Cronon: interesting, we have a mole here:)
[11:07] Teravus Ousley: Hello
[11:07] Andrew Linden: the stuff in the next Havok4 update was all stuff that was done last week
[11:08] Andrew Linden: so nothing new to report that will be in the update
[11:08] Arawn Spitteler: STATUS_ROTATE?
[11:08] Andrew Linden: what is NOT in the update are things such as the STATUS_ROTATE fix
[11:09] Andrew Linden: I fixed one of Cubey Terra's planes (the ultralight). NOT in the update
[11:09] Andrew Linden: And while I was at it I convexified the collision proxies for sculpy prims
[11:10] Andrew Linden: they were being proxied by tori
[11:10] Rex Cronon: what does that mean? i am talking about sculpties.
[11:10] Andrew Linden: but will eventully be convex hull's of tori
[11:10] Andrew Linden: Well, at the moment sculpties don't collide anything like their apparent shape
[11:10] Rex Cronon: yes, i know is torus
[11:10] Andrew Linden: the collide and roll around like a regular default torus, with scale
[11:11] Andrew Linden: but the torus uses the Havok "mesh shape" which is kinda CPU expensive
[11:11] Andrew Linden: and sculpties are usually used for their collision properties
[11:12] Andrew Linden: so instead of using the mesh shape I use the "convex vertices shape" (basically a convex hull)
[11:12] Andrew Linden: which is cheaper
[11:12] Andrew Linden: and still kinda rolls around like the torus (but with the hole filled in)
[11:12] Arawn Spitteler: So, Stairs won't need their invisible ramps?
[11:12] Rex Cronon: so now basically sculpties are ovals?
[11:13] Andrew Linden: no, stairs may still need their invisible ramps
[11:13] Andrew Linden: sculpties still won't collide like their apparent shapes
[11:13] Andrew Linden: but the shape the DO use will cause less lag
[11:13] Nitram Wei: Script run-time error
[11:13] Nitram Wei: Too Many Listens
[11:13] Teravus Ousley: always a positive :D
[11:13] Andrew Linden: yeah, they are kinda egg shaped... but symmetrical, no fat end
[11:14] Rex Cronon: elyptical?
[11:14] Arawn Spitteler thinks the lack of a fat end is why we can't open them
[11:14] Rex Cronon: hi nitram
[11:14] Andrew Linden: I had some subject that I was going to raise for discussion... but I've forgotten what it is.
[11:14] Nitram Wei: hi
[11:14] Andrew Linden: Hello Nitram, welcome.
[11:14] Nitram Wei: thnx
[11:15] Andrew Linden: oh yeah, I remember one thing...
[11:15] Teravus Ousley: Kind of like that.. ---> Points to the prim on the table
[11:16] Andrew Linden: A Resident named Areth Gall contacted me because he had been measuring the moments of inertia of hollow boxes
[11:16] Andrew Linden: and was coming up with numbers that did not match the theoretical values
[11:16] Andrew Linden: so when I looked into it... sure enough there was a bug
[11:16] Andrew Linden: so I fixed it (NOT in the update, but few would notice anyway)
[11:17] Andrew Linden: and was pleased that I've got such intrepid Residents out there testing stuff.
[11:17] Rex Cronon: they had too much mass?
[11:17] Andrew Linden: well... too much angular mass
[11:18] Andrew Linden: all of the diagonal elements of the inertia tensor were too large by 0.01
[11:18] Arawn Spitteler: That sounds rather precise
[11:18] Andrew Linden: really not very noticable... maybe 16% too big for a default cube
[11:19] Andrew Linden: yes, it was very precise. Turns out the default inertia was set to the identity before adding up the parts
[11:19] Andrew Linden: and there is a mysterious factor of 1/100 related to the discrepancies between the new "real" mass and the "legacy" Havok1 mass that you get when you call llGetMass()
[11:20] Arawn Spitteler: How do you measure "real" mass?
[11:20] Andrew Linden: well, you can multiply the legacy mass by 100
[11:21] Andrew Linden: but it is only used internally (in the physics engine) at the moment
[11:21] Andrew Linden: I'm not interested in adding a llGetRealMass()
[11:21] Arawn Spitteler wonders how the measurement was done.
[11:21] Andrew Linden: but maybe someday when we have LSL3 or whatever we can move toward more realistic masses
[11:22] Arawn Spitteler: So GetRealMass()= llGetMass()*100;
[11:22] Andrew Linden: How would you measure the moments of inertia of an object in SL? Anyone know?
[11:22] Rex Cronon: i thought that lsl will be replaced by mono
[11:22] Andrew Linden: MONO will initially have the same LSL interface... you'll still write in LSL
[11:22] Andrew Linden: but the execution under the hood is different
[11:23] Arawn Spitteler: I know that Rotational Inertia is the sum of mass* radius squared
[11:23] Andrew Linden: but what if someone were to write LLPython that uses different syntax
[11:23] Rex Cronon: but eventually u should be able to use c#, c++, java, or other languages
[11:23] Andrew Linden: and maybe different callback structure
[11:23] Andrew Linden: Yes Arawn, it is the integration of mass * radius^2, over the object's volume
[11:24] Domchi Underwood: Hi everyobdy...
[11:24] Ruud Lathrop: Hi
[11:24] Rex Cronon: hi domchi
[11:24] Andrew Linden: so for complicated objects the integral is messy
[11:24] Andrew Linden: it is often easier to just measure it
[11:24] Rex Cronon: .
[11:25] Arawn Spitteler: I think there's something about adding the Moments of Inertia, of each prim, as factored somehow by the square of that prim's radius. Generally, you'd llGetTargetOmega before and affter llApplyRotationalImpulse
[11:25] Andrew Linden: that is right Arawn
[11:26] Andrew Linden: In SL you can give the object hover so that it doesn't hit anything
[11:26] Andrew Linden: then apply angular impulses and measure the angular acceleration
[11:26] Andrew Linden: F = ma is the linear equation
[11:27] Andrew Linden: torque = inertia * angular_velocity is the angular equivalent
[11:27] Andrew Linden: however, the inertia is actually a tensor, but if you're spinning about one cardinal axis then it collapses into a scalar equation
[11:27] Arawn Spitteler: Angular Acceleration could get complicated
[11:27] Andrew Linden: and you can just measure each axis by itself
[11:28] Andrew Linden: The legendary Xylor Bayslef was the first person I knew of to measure the inertia tensors in that way.
[11:29] Ruud Lathrop: What are the major bug fixes in this release?
[11:29] Andrew Linden: lemme look them up...
[11:29] Arawn Spitteler: So, if this table were set Physical Hover, it would gain speed, each time one of us dismounted?
[11:30] Kitto Flora: It has no rotational speed
[11:30] Kitto Flora: cchange the sizze - still no r-speed
[11:31] Andrew Linden: first of all, the next update will bring Havok4 up to par with the recently released 1.19.0 server
[11:31] Simon Linden: The hover calculations use the current mass and are applied in each timestep, so changing the mass won't make it rise or fall
[11:31] Andrew Linden: There were a few small performance optimizations... I expect them to be measureable in a few content scenarios, but not very big.
[11:32] Andrew Linden: There was a duplication of selection not lining up bug that as fixed
[11:33] Andrew Linden: I tried to fix SVC-1089 and failed
[11:34] Andrew Linden: SVC-1239 is fixed (llSetLinkPrimitiveParams() on avatar) was breaking some boats and other things
[11:34] Andrew Linden: I'm hoping SVC-1272 (boats too low in water) is actually fixed
[11:34] Domchi Underwood: Andrew, I read your plan on megaprims... but can you comment on microprims? Will some action be taken regarding them?
[11:34] Andrew Linden: oops... I take that back
[11:35] Andrew Linden: I wasn't watching the revision numbers carefully and looked over where we stopped for this update
[11:36] Andrew Linden: sigh... which means everything since SVC-1089 is NOT fixed in this update
[11:36] Gaius Goodliffe: Hmm... Andrew, is SVC-1089 possibly related to other "dropping like a rock" issues, like, for example, how when flying at 640m, if you press forward, you drop like a rock?
[11:36] Andrew Linden: the dropping like a rock issue is fixed
[11:36] Andrew Linden: and is not related
[11:36] Gaius Goodliffe: oh cool.
[11:37] XerxesTiny.LITTLE BEE.Head: Touched.
[11:37] Ruud Lathrop: If a bug is 'fix pending' will that mean they are being released in next server/client update?
[11:38] Andrew Linden: sigh... I'm just not in the mood to read the logs and figure out where one update started and another ended
[11:38] Andrew Linden: do you remember anything Simon?
[11:38] Simon Linden: Ruud - it may not be the immediate next update. It depends on when it was fixed vs. when we built and start testing the update
[11:38] Andrew Linden: most of the stuff we announced as in the next update on Thursday is in this one
[11:39] Simon Linden: I'm not sure of the cutoff either ... I think most of what's about to go out was discussed Thursday
[11:39] Ruud Lathrop: Ok
[11:39] Sascha Vandyke: is an update version scheduled already ?
[11:40] Arawn Spitteler: So, how soon is a certain Cabriole ready for a test drive?
[11:40] Simon Linden: I do know I'm currently working on a vehicle bug, where it's going wildly out of control. I'm not sure how many vechicles it affects
[11:40] Andrew Linden: the update should happen today if all goes well
[11:40] Sascha Vandyke: ahh ok
[11:41] Sascha Vandyke: is there any blog to find the time or announcement
[11:41] Andrew Linden: no the blog announcement would go out after the update has happened and the beta is open
[11:41] Sascha Vandyke: ahh ok thx
[11:41] Andrew Linden: so you can tell when it is happening when the beta grid goes down
[11:42] Andrew Linden: but this update will also go out to any regions on the main grid that are using Havok4
[11:42] Andrew Linden: hrm... I wonder if they'll do that tomorrow or today
[11:42] Andrew Linden: Sidewinder knows.
[11:42] Kitto Flora: This is
[11:42] Sascha Vandyke: thx
[11:43] Kitto Flora: so when he 78103 is a larger number, then you are on a new h4 server version
[11:44] Domchi Underwood: Andrew, I have a question re your megaprim genocide plan. ;)
[11:44] Andrew Linden: ok ask
[11:45] Domchi Underwood: What about microprims... any plans about them?
[11:45] Andrew Linden: No microprims for a while.
[11:45] Andrew Linden: Er... no plans.
[11:45] Andrew Linden: The physics engine has a "collision tolerance" of 0.1 meter
[11:45] Sascha Vandyke: will also the 256x256 be gone? lol that would interest me personally
[11:45] Brookston Holiday: Megaprim genocide plan?
[11:45] Domchi Underwood: :) But eventually, will they have to go as well?
[11:45] Domchi Underwood: The plan to get rid of megaprims. :)
[11:46] Andrew Linden: objects smaller than that,or on that scale, will not collide very accurately
[11:46] Andrew Linden: There is no Megaprim genocide plan
[11:46] Andrew Linden: however... that reminds me...
[11:46] Andrew Linden: I think Kelly fixed a megaprim link bug with a hack to the linkability rules
[11:46] Gaius Goodliffe: w00t!
[11:46] Andrew Linden: so someone here had a blimp that did not link
[11:46] Andrew Linden: I believe that has been fixed
[11:47] Simon Linden: Yes, that was fixed ... I confirmed it
[11:47] Gaius Goodliffe: That was me. :D
[11:47] Andrew Linden: Here is the plan for megaprims...
[11:47] Rex Cronon: so any prim that is within 10m can be linked?
[11:47] Andrew Linden: (1) make it so parcel owners can return (or maybe just move?) objects that are overlapping their parcels
[11:48] Andrew Linden: (2) provide some UI feedback to builders when their objects extend outside of their allowed building boundaries
[11:48] Gaius Goodliffe: (I don't know if "just move" would work -- what if the megaprim is bigger than the owner's parcel? (thinking of First Land plots (and dating myself, I guess lol)))
[11:48] Andrew Linden: (3) give parcel owners some more options for automatic enforcement of their boundaries (off by default)
[11:49] Ruud Lathrop: That can also be applied to normal prims?
[11:49] Andrew Linden: (4) allow builders to build any size (and sometimes only as large as can fit on their parcel) AKA "megaprim liberation"
[11:49] Gaius Goodliffe: That would be very nice. :D
[11:50] Andrew Linden: yes, steps (1) throuth (3) are just about normal prims
[11:50] Andrew Linden: after that, Megaprims can be considered normal prims
[11:50] Andrew Linden: there are a few details in there...
[11:50] Andrew Linden: sculpties are difficult to query for parcel overlap
[11:50] Andrew Linden: so that is it own kinda big project
[11:50] Sascha Vandyke: ok what about a private estate with some parcels and i want a megaprim going over them
[11:51] Andrew Linden: btw... that is NOT the official LL plan on megaprims
[11:51] Andrew Linden: that is MY plan
[11:51] Sascha Vandyke: ahh ok lol
[11:51] Andrew Linden: I'm the only one with a plan on megaprims
[11:51] Andrew Linden: unfortunatly, I'm trying to finish Havok4 first
[11:51] Rex Cronon: don't sculpties have a bounding box?
[11:52] Andrew Linden: yes sculpties do have a bounding box, but it is not accurate
[11:52] Teravus Ousley: up to twice the radius that the sim currently knows about
[11:52] Teravus Ousley: and that's radius in terms of an ellipsis
[11:52] Andrew Linden: so as soon as you start trying to enforce prim encroachment then sculpties will be one of those corners of the possibility space that do not follow the expected rules
[11:52] Teravus Ousley: :D
[11:53] Andrew Linden: also... there is the problem of prims overlapping into neighboring regions
[11:53] Andrew Linden: that is harder to solve for the encroachment problem
[11:53] Andrew Linden: so that is its own kinda big project
[11:54] Andrew Linden: I describe my megaprim plan every other week in office hours nowadays
[11:54] Ruud Lathrop: :)
[11:54] LuckySweet Kondor: o.o walked in on somthing i see
[11:54] Domchi Underwood: And you also did on forums and blog... :)
[11:55] Arawn Spitteler: Mega-Sculpties will be nice, but their appearence shouldn't be allowed to damage their repute
[11:55] Gaius Goodliffe: prims of ill-repute? o.O
[11:55] Andrew Linden: I was thinking I'd like to work on making better collision approximations for sculpties
[11:56] Domchi Underwood: Ah, forums are working again... here's the link to original post Andrew wrote:
[11:56] Domchi Underwood:
[11:56] Andrew Linden: however recently I ventured into the simulator side code for sculpties (for the convexification project I was mentioning earlier today)
[11:56] Andrew Linden: and now I don't want to touch the sculptie code with a 10-foot pole
[11:57] Teravus Ousley: heh
[11:57] Arawn Spitteler: A ten foot pole wouldn't be Sculpty or Maga
[11:57] Andrew Linden: I might even rather fix vehicle and avatar motion bugs than try to work on sculpties
[11:57] Teravus Ousley: hehe
[11:58] Gaius Goodliffe: lol
[11:58] Sascha Vandyke: is better yes, i have found a lot of av bugs lately lol
[11:58] Teravus Ousley: the texture would be a property of the sculptie that would have to be 'stored with' the sculptie instead of requested on demand.. and you'd have to decode the image and generate a convex mesh from it.
[11:58] Rex Cronon: what did u find there that scared u so much:)
[11:59] Teravus Ousley: a scary proposition.. with lots of math
[11:59] Andrew Linden: it turns out the simulator doesn't know much about sculpties
[11:59] Harleen Gretzky: Then there is also animated sculpties to worry about
[12:00] Andrew Linden: but prims have a general purpose container for holding some "extra data"
[12:00] Teravus Ousley: lol, the idea of a body having to be rebuilt with an animation is @.@
[12:00] Andrew Linden: it is currently used for flexi, sculptie, and something else that I can't remember
[12:00] Gaius Goodliffe: I've always had the impression that sculpties were pretty much spheres with some interesting client-side drawing behavior...
[12:00] Andrew Linden: the interface to that extra data container is terrible (a mix of bad C and bad C++)
[12:00] Teravus Ousley: knowing what a change does to a physical body that interpenetrates something.. essentially 'explode kaboom'
[12:01] Teravus Ousley: like what happens when you unlink physical prim
[12:01] Teravus Ousley: :D
[12:01] Andrew Linden: right Gaius, so atm the simulator doesn't know much about the sculptie status
[12:01] Andrew Linden: it doesn't yet download the textures itself
[12:01] Gaius Goodliffe nods.
[12:01] Andrew Linden: but it would have to in order to generate the collision mesh
[12:02] Andrew Linden: or else we'd have to provide the collision mesh in some other way
[12:02] Andrew Linden: in fact, there is very little sculptie specific code in the simulator
[12:02] Andrew Linden: since it goes through a general purpose container
[12:03] Andrew Linden: but that general purpose container is just poorly implemented and I phear it
[12:03] Roxanne Hynes: Are there plans on completely removing some of the Megaprims, for example prims that are equal or go above 256x256? Or will nothing change for now with that?
[12:03] Gaius Goodliffe: hehe
[12:03] Andrew Linden: sorry, yes Roxanne. megaprims with sides > 256 will have their sides clamped to 256
[12:03] Gaius Goodliffe: Awww
[12:03] Sascha Vandyke: ahh then is fine
[12:03] Roxanne Hynes: ok thanks
[12:04] Andrew Linden: not exactly megaprim genocide, but definitely targeted murder
[12:04] Teravus Ousley: yeah, at this point, there's not a lot of good uses for megaprim beyond that size
[12:04] Sascha Vandyke: is there anything you noticed from havok 4 playing together with the new server code?
[12:04] Arawn Spitteler: I think it depends on whether you've more than one sim
[12:04] Simon Linden: Sascha - you mean moving from one sim to another?
[12:04] Gaius Goodliffe: Extropia Core has a good use for them -- check out their "low earth orbit" stuff if you're ever there. I think they use a 1024x1024x100 one to make it happen, though.
[12:05] Sascha Vandyke: yes, and some more lag now, think it has become worse after introducing 1.19 to the main
[12:05] Sascha Vandyke: i have avs coming to my sim and they just don't load sometimes
[12:06] Simon Linden: I haven't heard of any specific problems, but the next version we push out will be based on 1.19
[12:06] Sascha Vandyke: ahh ok
[12:06] Simon Linden: ... meaning up-to-date; we try to pick up any bug fixes that are pushed with the normal server so we don't fall behind
[12:07] Kitto Flora gusses: If you have an island sim and you put a 1024x1024 megaprim on it, and plaint it, you can make ground or sky scenery that extends further than the edge of the sim... effectivly to the horizon.
[12:07] Sascha Vandyke: ok thx. I also have the problem to report some bugs, jira is not so nice to find bugs for havok4, i already reported twice a double one
[12:08] Andrew Linden: Yeah, the bugs are piling up there.
[12:08] Gaius Goodliffe: Kitto: right. In Extropia they use one up high so you see a curved earth beneath you while you're in orbit.
[12:08] Andrew Linden: I'm going to try to make a pass and close some of the Havok4 bugs in the public jira
[12:08] Harleen Gretzky: How do you gus? as in [12:07] Kitto Flora gusses: If...
[12:08] Sascha Vandyke: would be nice
[12:09] Kitto Flora: SVC-1093 was fixed a while back, still shows open
[12:09] Kitto Flora does /me guesses
[12:09] Gaius Goodliffe guesses you should use /me. :)
[12:09] Harleen Gretzky: ahh
[12:09] Andrew Linden: I've got to go now, so we'll have to wrap up.
[12:10] Sascha Vandyke: have a nice day
[12:10] Ruud Lathrop: Ok
[12:10] Teravus Ousley: Good luck with the avatar movement, sir
[12:10] Arawn Spitteler has heard some debate, on whether Megaprims lag the sims
[12:10] Ruud Lathrop: All have a nice night :)
[12:10] Teravus Ousley: :D
[12:10] Kitto Flora: We await the update with baited bugtraps.
[12:10] Rex Cronon: bye everybody, i have to go to
[12:11] Simon Linden: Thanks everyone
[12:11] Teravus Ousley: take care :D