User:Andrew Linden/Office Hours/2008 05 15

Transcript of Andrew Linden's office hours:

[17:03]	Arawn Spitteler:	Hi, Andrew
[17:03]	Sidewinder Linden:	wow... small party tonight :)
[17:03]	Andrew Linden:	hello
[17:03]	Kooky Jetaime:	Hey Andrew
[17:03]	Kooky Jetaime:	And Sidewinder
[17:03]	Arawn Spitteler:	Hi, Sidewinder.
[17:03]	Sidewinder Linden:	hi guys
[17:03]	Kooky Jetaime:	I found a place that is a good repro for the ground slide
[17:03]	Kooky Jetaime:	thats why I came
[17:04]	Andrew Linden:	We had a relatively large crowd on Tuesday.
[17:04]	Andrew Linden:	but it is early yet
[17:04]	Arawn Spitteler:	Was tuesday when you announced the wind-down?
[17:05]	Kooky Jetaime:	wind-down?
[17:05]	Arawn Spitteler:	H-Team is breaking up, in the next weeks.
[17:05]	Les White:	(hiyas)
[17:05]	Andrew Linden:	Yes, I mentioned that the Havok4 team would eventually split up and work on other stuff, over the next few weeks.
[17:05]	Sidewinder Linden:	hi les
[17:05]	Kooky Jetaime:	Solved the sinkin problem?
[17:06]	Andrew Linden:	That is, the othere developers would work on other stuff, and I would continue to fix Havok4 bugs for a while.
[17:06]	Zwagoth Klaar:	Should we hire grief counselors for the break up?
[17:06]	Sidewinder Linden:	although one thing to know is that that does not mean we won't continue working on physics and performance... mostly an internal organizational shift, and somewhat recognition that we're not "in release" and not working on the project to "get to release"
[17:06]	Sidewinder Linden:	simon has been tackling some thorny performance issues, and he's going to keep on that... we've been trying to localize the sources of "cyclic time dilation" and now have a few in hand
[17:07]	Andrew Linden:	Les, I've been meaning to talk to you about the #12 Stock Brick racecar...
[17:07]	Arawn Spitteler is still having flying train issues, frequently: I think the important part of the sliding stance issue, is that it might offer hints to the sinking.
[17:07]	Les White:	hmm, not sure that is one of mine, andrew. what's the issue?
[17:07]	Kooky Jetaime:	to much downforce on avis?
[17:07]	Andrew Linden:	I thought I got a copy from you... can't remember who made it.
[17:08]	Andrew Linden:	Perhaps it is in my inventory...
[17:08]	Andrew Linden:	No, I have it in the Beta database.
[17:08]	Andrew Linden:	Anyway... one of the reasons it is such a squirly car is that it has a rather small moment of inertia
[17:09]	Andrew Linden:	because it is really just a hollow shell of a vehicle
[17:09]	Arawn Spitteler:	Hollow would increase moment of inertia, relative to Mass.
[17:09]	Kooky Jetaime:	Andrew, does something force-neutralized sink? heh, did someone set gravity to 10.0 instead of 9.8?
[17:09]	Andrew Linden:	I modified it a bit as an experiment... added some mass and inertia tensor by adding some solid blocks to it and found it handled a bit better.
[17:10]	Andrew Linden:	No Kooky, gravity has not changed that I know of.
[17:10]	Les White:	it's not mine for sure. in the series i run the cars are equalized to 100 units of mass
[17:10]	Rex Cronon:	hello everybody
[17:10]	Les White:	hiya
[17:10]	Arawn Spitteler:	Hi, Rex.
[17:10]	Gravity check:	Hello, Avatar!
[17:10]	Areth Gall:	Hm. I may have a problem with my wireless card cutting out
[17:10]	Les White:	i have noticed with the cars in my series, all equalized scripts, wheel prims, size and mass...that some cars use 1.5 ms phys time and some as much as 8
[17:11]	Les White:	the ones with twisted torii and such really seem to lag the place up
[17:11]	Rex Cronon:	i should weight a lot heavier with all the metal on me:)
[17:11]	Andrew Linden:	yeah, concave vehicles add lag
[17:11]	Arawn Spitteler:	That depends, Rex, on whether the Metal is a massless attachment, or painted on to Shape.
[17:11]	Andrew Linden:	I'd really like to work on that.
[17:11]	Les White:	i'd love to see a llSetHull(vector bounding_box) kinda thing and make the vehicle prims non relavent
[17:12]	Andrew Linden:	Unfortunately we don't modify the mass of the avatar based on mass of attachments.
[17:12]	Andrew Linden:	I had hoped to eventually add that feature... but now it would just break stuff I'm sure.
[17:13]	Les White:	ok. some people are making a base car with a couple big prims and wearing the the rest as an attachment to keep things low lag. it works. though kinda messy for users
[17:13]	Gaius Goodliffe nods.
[17:13]	Gaius Goodliffe:	Basically, it's a way to hack per-prim phantom. You have non-phantom big prims, and then the attachment (which is phantom).
[17:13]	Les White:	over all i dont really have issues with h4 and vehicles...just the optimization of shapes. some making crazy phys time
[17:14]	Les White:	right Gaius
[17:14]	Andrew Linden:	Well, I have an idea for that... what I'd like to do is replace all concave objects with a list of smaller convex pieces that achieve the same collision shape.
[17:14]	Gravity check:	Touched.
[17:15]	Andrew Linden:	We've already tested it in Havok2 and found that the physics engine ran faster for such objects.
[17:15]	Les White:	for sure. that would clean up some things if it came into play
[17:15]	Arawn Spitteler:	Wouldn't effect the Prim Count?
[17:15]	Gravity check:	Touched.
[17:15]	Andrew Linden:	No this would be done "under the hood"
[17:15]	Les White:	it's simply not possible to allow some prims phatom in an otherwise solid link set?
[17:16]	Andrew Linden:	however... some features in SL make that project more complicated than it sounds...
[17:16]	Andrew Linden:	such as the per-prim collision callbacks in LSL
[17:16]	Kooky Jetaime:	Hmmm Mass * 9.8, applied as a force should be weightless right?
[17:16]	Andrew Linden:	but it is still doable.
[17:16]	Areth Gall:	Hm. So how does the physics engine for objects work? If you're saying that objects run faster or slower.. I thought all objects would be updated at the same time.
[17:16]	Andrew Linden:	sorry, I mean the CPU cycles per object are higher for concave shapes.
[17:17]	Les White:	right
[17:17]	Areth Gall:	ah. Sorry. My fault
[17:17]	Kooky Jetaime:	If thats right, then why is my box floating up at a near constant rate from a stop....
[17:17]	Andrew Linden:	so for a pile of convex objects... the real-time it takes for the physics engine simulation to run each step would be smaller
[17:18]	Creem Pye:	sorry :O
[17:18]	Les White:	creem attacked sofia!
[17:18]	Sofia Westwick:	lol
[17:18]	Andrew Linden reads the script....
[17:18]	Kooky Jetaime:	Ok
[17:19]	Andrew Linden:	I'm not sure why it creeps up... there is a little bit of noise in the physics simulation it seems.
[17:19]	Gaius Goodliffe:	I think that's a less than ideal solution, though. There are plenty of vehicle shapes (large V-tails or even wings, for example) where colliding on the convex shape might be... weird.
[17:19]	Sidewinder Linden:	sofia files a/r on creem :)
[17:19]	Kooky Jetaime:	Could that same noise account for the sink?
[17:19]	Sofia Westwick:	lol Sidewinder
[17:19]	Creem Pye blames the mask for making him do it
[17:19]	Kooky Jetaime:	granted its rising faster than the sink effect, I think, but.. Hmm
[17:19]	Andrew Linden:	Yes Gaius, but there may be ways to still simplify the object...
[17:19]	Gaius Goodliffe nods.
[17:20]	Andrew Linden:	If we had hierarchy then you could organize the vehicle parts into groups and convex hullify the various sub-parts
[17:20]	Gaius Goodliffe:	I think you need hints form the developer though. Let us specify which prims are important, for example.
[17:20]	Andrew Linden:	right, per-prim phantom would help here
[17:20]	Les White:	for sure
[17:20]	Creem Pye:	how about mixing physical and phantom prims inthe same linkset? (where phantom parts would have much lighter physics processing, I presume)
[17:21]	Creem Pye:	this is already done with flexi to a degree =)
[17:21]	Gaius Goodliffe:	I think that's the best solution, ultimately.
[17:21]	Andrew Linden:	right Creem, that is what I meant by "per-prim phantom"
[17:21]	Sofia Westwick:	I thought phantom prims could be physics
[17:21]	Sofia Westwick:	set
[17:21]	Zwagoth Klaar:	Dont you then have to decide if the phantom objects have weight though?
[17:21]	Creem Pye:	ah sorry, I need to catch up to the conversation =)
[17:21]	Andrew Linden:	no, phantom objects already have mass
[17:22]	Andrew Linden:	actually, right now "phantom" means "collides with terrain only"
[17:22]	Zwagoth Klaar:	okay, solves that question
[17:22]	Andrew Linden:	we set it that way to prevent phantom objects from falling through the heightfield and getting deleted
[17:22]	Arawn Spitteler:	Do Physical Phantoms still fall through the ground?
[17:22]	Andrew Linden:	No Arawn, they hit the ground.
[17:22]	Zwagoth Klaar:	only if somebody sits on them
[17:22]	Zwagoth Klaar:	then they fall
[17:22]	Rex Cronon:	how about volumedetect?
[17:22]	Andrew Linden:	volume detect is truely collisionless
[17:23]	Andrew Linden:	or... it should be
[17:23]	Andrew Linden:	turns out there is a bug there where some collections of VolumeDetect objects appear to collide with other objects.
[17:23]	Andrew Linden:	It is on my plate but I have yet to actually work on it.
[17:23]	Kooky Jetaime:	Areth, are you on my lap sweety?
[17:24]	Andrew Linden:	I was trying to work on some broken "swimmer" attachment today.
[17:24]	Areth Gall:	..Strange
[17:24]	Areth Gall:	Pardon me.
[17:24]	Kooky Jetaime:	thats ok
[17:24]	Les White:	that swimmer is in the news a lot. what part is actualy broken? can it be fixed with an update from the creator?
[17:24]	Kooky Jetaime:	your cute, so I'll let it slide :)
[17:24]	Rex Cronon:	that happened to me to. i usually end up on peoples heads:)
[17:24]	Andrew Linden:	It is not clear what is broken about the swimmer yet.
[17:24]	Sidewinder Linden:	from what i've heard, les, the creator is no longer around
[17:25]	Kooky Jetaime:	swimmer?
[17:25]	Andrew Linden:	I got sidetracked by some other stuff and a little crash-mode.
[17:25]	Les White:	siggy is gone? didnt know that
[17:25]	Sidewinder Linden:	well according to sean and another person, but it could just be that he's not answering them :)
[17:25]	Andrew Linden:	The swimmer is an attachment that makes you swim when you walk into the water.
[17:25]	Les White:	yeah.
[17:25]	Kooky Jetaime:	Hm
[17:26]	Andrew Linden:	Ok the one announcement I had was about megaprims...
[17:26]	Kooky Jetaime:	let me guess, sinking?
[17:26]	Andrew Linden:	I was talking about megaprims a bit on Tuesday, and there was an email discussion going on about them in the sldev mailing list
[17:26]	Andrew Linden:	so for those who have not been following the mailing list history...
[17:27]	Sidewinder Linden:	*braces for the discussion to veer hard to megaprims for the rest of the time* :)
[17:27]	Kooky Jetaime:	hehehe
[17:27]	Andrew Linden:	I just posted my revised outline of my "Megaprim Liberation" plan, which is not official LL policy, but merely the plan I would submit to others in LL should we ever have a meeting to talk about megaprims...
[17:28]	Kooky Jetaime:	I just know I liked being able to make a few..but they seemed to poof on me:/
[17:28]	Andrew Linden:	So the steps toward Megaprim Liberation are:
[17:28]	Andrew Linden:	(1) Add a feature that allows land owners to return any object that overlaps their land (aka Prim Encroachment)
[17:28]	Sofia Westwick:	thats a good one
[17:28]	Les White:	long over due
[17:28]	Andrew Linden:	(2) Make static objects collide across region boundaries
[17:29]	Sofia Westwick:	what does that mean Andrew?
[17:29]	Andrew Linden:	(3) Gradually (every few weeks or months) increase the default max prim size (and perhaps linkability)
[17:29]	Sofia Westwick:	(2)
[17:29]	Zwagoth Klaar:	Not sure 2 is such a great idea, it really increases the accidental damage rezzing one.
[17:29]	Andrew Linden:	so... to 16, then 32, then 64m etc... up to 256 max
[17:30]	Kooky Jetaime:	Why would cross region collisions be part of that ?
[17:30]	Andrew Linden:	do it incrementally so we can have time to fix any resulting problems (server or client-side) that crop up as a result
[17:30]	Rex Cronon:	why not go straight to 64
[17:30]	Kooky Jetaime:	Or simply to allow cross region parcel cross checking?
[17:30]	Kooky Jetaime:	At least go to 32 for the first step
[17:31]	Kooky Jetaime:	20x20 is very useful, a lot more useful than 16x16
[17:31]	Andrew Linden:	(4) Add a feature that allows Estate/Region owners to set void, distant horizon, and cloud layers/textures (to eliminate the need for >256m megaprim sky/horizon effects)
[17:31]	Les White:	hot
[17:31]	Andrew Linden:	(5) Crack down on megaprims with sides > 256m
[17:31]	Sofia Westwick:	I know mega prims as of now sometimes casue rendering issues, if you increase the default max prim size (and perhaps linkability) Andrew would that sold the rendering issues mega prims sometimes cause?
[17:31]	Andrew Linden:	fin
[17:31]	Gaius Goodliffe:	Andrew, given that encroachment and region boundry problems are no bigger for megaprims than they are for linksets, how about liberating meagprims up to the size of a linkset for now?
[17:31]	Kooky Jetaime:	Couldn't you crack down on them now?
[17:31]	Sofia Westwick:	solve*
[17:32]	Kooky Jetaime:	it should be a simple matter to scan the asset database for objects with a size in excess
[17:32]	Andrew Linden:	Kooky, the exact order of the steps is not clear. It might be possible to bump default limits up a little bit before (1) and (2) are completely finished...
[17:33]	Andrew Linden:	My opinions are not set in stone on this
[17:33]	Andrew Linden:	We'd have to come to some sort of consensus here at LL
[17:33]	Andrew Linden:	however, I don't expect much arguments against most of my proposed changes
[17:34]	Andrew Linden:	most of the steps in that list are things that everybody wants
[17:34]	Sofia Westwick:	I know mega prims as of now sometimes casue rendering issues, if you increase the default max prim size and linkability Andrew would that solve the rendering issues mega prims sometimes cause?
[17:34]	Kooky Jetaime:	Andrew, the final limit going to be 256x256x256?
[17:34]	Andrew Linden:	Right Soft, which is why I would like to gradually increment the sizes
[17:34]	Sidewinder Linden:	it seems that one of the internal discussions is to look at viewer performance impacts...
[17:34]	Andrew Linden:	so we can see what the rendering (and simulator) problems are, if any
[17:34]	Rex Cronon:	what happned to "sky is limit" for megaprim sizes?
[17:34]	Andrew Linden:	as larger prims/objects become very common
[17:35]	Sofia Westwick:	I love your mega prim leberation idea Andrew
[17:35]	Andrew Linden:	There are long-term problems with very large megaprims...
[17:35]	Sofia Westwick:	that would be great Sidewinder
[17:35]	Kooky Jetaime:	Honestly, I think a lot of builders would be happy for 20x20 right now.... for floor spaces that could result in a 75% reduction in prims
[17:35]	Arawn Spitteler:	Even at 10X10, I've seen blocks missing.
[17:35]	Andrew Linden:	for instance item (2) is difficult to solve well for megaprims that extend across 28 different regions
[17:36]	Gaius Goodliffe:	hehe
[17:36]	Kooky Jetaime:	the infinate checkerboard prim
[17:36]	Andrew Linden:	and for which a slight rotation might throw them across a different set of 33 regions
[17:36]	Rex Cronon:	i have no problem seeing megaprims, but 10x10 tend to not showup
[17:36]	Andrew Linden:	There is very little really cool content, in my opinion, that requires >256m prims...
[17:36]	Creem Pye:	the rendering issue is from a megaprim with a center just beyond the draw distance limit not showing up, right?
[17:37]	Andrew Linden:	assuming that we can do (4) which is the main way they are being used today
[17:37]	Les White:	yeah
[17:37]	Phantom Messenger:	volumedetect sim scanning
[17:37]	Sidewinder Linden:	there is the replacement sky and water to horizon issue, which seems to be the big reason for "huge mega's" in builds today
[17:37]	Sofia Westwick:	I totaly agree with what you ssaid Andrew
[17:37]	Andrew Linden:	The only other somewhat interesting content that has been suggested is a giant 4000m building or statue, a vertical pillar that was still in the region's physics engine
[17:38]	Andrew Linden:	however... I don't see why that object couldn't be made of a pile of 256m prims
[17:38]	Andrew Linden:	and it still produces a problem with (2)
[17:38]	Gaius Goodliffe:	If the prim is so large its center is outside your draw distance, you couldn't see the statue anyway
[17:38]	Andrew Linden:	I'd really like to simplify the problem with (2) with a simple 256 meter limit
[17:39]	Andrew Linden:	Right, renderwise it makes sense to break such an object up into pieces anyway
[17:39]	Andrew Linden:	pieces that are on par with the typical render distance
[17:39]	Gaius Goodliffe:	I think is sounds like sane policy. :)
[17:39]	Rex Cronon:	i guess things could be changed sometime in the distant future
[17:39]	Kooky Jetaime:	What is the "typical render distnace" ?
[17:40]	Andrew Linden:	So that is "Andrew's current Megaprim Liberation plan".
[17:40]	Andrew Linden:	If anyone else asks about megaprims I'll just point them at this office hour transcript.
[17:40]	Gaius Goodliffe:	Can I ask the dreaded "when" question?
[17:41]	Andrew Linden:	Besides (1) none of that is being worked on, or is in the schedule.
[17:41]	Andrew Linden:	I've been thinking about (2) and have some ideas on how to do it.
[17:41]	Andrew Linden:	And I would probably start (2) sometime shortly after I'm done fixing bugs.
[17:41]	Zwagoth Klaar:	I still see 2 as problematic with any size megaprims, and link sets in general, I always liked knowing I couldnt screw up another sim with content on the one I was on.
[17:41]	Arawn Spitteler:	For a 4km statue to be visible to a 64m viewer like mine, it only has to have its center within 64 meters, doesn't it?
[17:42]	Andrew Linden:	Yes Arawn, those are some of the rendering problems we'll have to deal with.
[17:42]	Andrew Linden:	That is still a problem with 256m prims + 64m draw distance
[17:42]	Arawn Spitteler:	A stack of 256 meters objects couldn't do that, but I don't know about a linked set.
[17:42]	Kooky Jetaime:	Andrew - perhaps something to ad
[17:42]	Kooky Jetaime:	er add to your list
[17:43]	Gaius Goodliffe:	At the rist of repeating myself, I'd just point out again that encroachment problems and sim border issues are no worse for smaller megaprims than for larger linksets. Expanding allowable prim size to the size of a linkset, then, could be done before (1) or (2) without real negative effect.
[17:43]	Andrew Linden:	What would that be Kooky?
[17:43]	Gaius Goodliffe:	*risk
[17:43]	Kooky Jetaime:	What about a graphical effect on the users screen if you have a physics collision with something not rendered?
[17:43]	Kooky Jetaime:	like hitting a force field or something
[17:43]	Kooky Jetaime:	so peopel don't get hit with apparent "phantom physics"
[17:43]	Andrew Linden:	well... perhaps that should be explicitly mentioned... but that is more of a bug with our interestlist code
[17:44]	Andrew Linden:	the presence of that object should be sent to the client so that it could render it
[17:44]	Andrew Linden:	so I view a problem like that as a bug that need to be fixed
[17:44]	Kooky Jetaime:	well, assuming the geometric center is not in range
[17:44]	Andrew Linden:	the "interestlist" code needs an overhaul anyway
[17:45]	Kooky Jetaime:	if I have a 10x10x256m box in front of me, while I'm standing at the bottom edge, it'll draw? even if my draw is set to say 64?
[17:45]	Kooky Jetaime:	Ok
[17:45]	Kooky Jetaime:	Teravus,... your uhm.. sitting on me
[17:45]	Andrew Linden:	that overhaul should probably be done in conjunction with any of the megaprim liberation steps in my list
[17:46]	Teravus Ousley:	sorry about that.
[17:46]	Arawn Spitteler:	Interestlist sounds like a similar algorhythm to Overlap
[17:46]	Kooky Jetaime:	its ok.. not the first time tonight
[17:46]	Andrew Linden:	Kook, it should work that way. Any surface near you should first be streamed from the simulator to the client
[17:46]	Andrew Linden:	and then rendered by the client
[17:46]	Arawn Spitteler:	Whether an object is in a place, rather than the location of its center.
[17:47]	Andrew Linden:	as long as a surface is within the draw distance it should show up -- otherwise I would consider it a bug
[17:47]	Teravus Ousley:	Hey Sidewinder, nice Bast :D
[17:47]	Sidewinder Linden:	heh thanks ter
[17:48]	Arawn Spitteler:	Is that Bast? I think my dragon looked more like me.
[17:48]	Teravus Ousley guesses from other conversations that the simulator uses frustum checks and distance
[17:48]	Kooky Jetaime:	I'm an Astrul :)
[17:48]	Kooky Jetaime:	er Astral
[17:49]	Andrew Linden:	A few new people showed up when I was in the middle of my outline of my megaprim plan. I'm assuming that they came to hear about megaprims.
[17:49]	Creem Pye:	would your step (2) for prim liberation be difficult if the object in the neighboring sim is physical and moving?
[17:49]	Andrew Linden:	The full transcripts of this meeting will be up on my wikipage after this meeting.
[17:49]	Mifune Thibaud:	actually i heard there was some discussion about child prims being assigned phantom properties
[17:49]	Andrew Linden:	And I sent an email this afternoon to the sldev mailing list with the same plan.
[17:49]	Mifune Thibaud:	but also interested in megaprim liberation
[17:50]	Kooky Jetaime:	I'd love having child prims with indipendiant physical status
[17:50]	Andrew Linden:	yeah, per-prim phantom is a feature that we talk about often.
[17:50]	Arawn Spitteler:	It would be independant Phantom Status.
[17:50]	Andrew Linden:	It would solve some problems with making vehicles with more prims
[17:50]	Andrew Linden:	and is one of the not-too-hard features I'd like to work on someday... after fixing more bugs
[17:51]	Gaius Goodliffe:	:D Not to hard. THat's hopefuly.
[17:51]	Mifune Thibaud:	hehe yes
[17:51]	Arawn Spitteler:	We could make solid block vehicles, with all the styling we wanted. Allowing Objects to have massless phantom attachments, would also be nice.
[17:51]	Les White:	yeah. it would be a real helpful thing
[17:51]	Mifune Thibaud:	it would also benefit the physics engine
[17:51]	Les White:	that's the help i mean :)
[17:51]	Teravus Ousley:	heh. It's an interesting concept.
[17:51]	Mifune Thibaud:	so simple non hollow prims can be used for the physics part
[17:52]	Andrew Linden:	Right, there are a few "physics hints" that content creators could provide that could help simplify the object's collision properties for the phyics engine.
[17:53]	Andrew Linden:	per-prim phantom is just one
[17:53]	Andrew Linden:	I've mentioned some of the others several times
[17:53]	Teravus Ousley:	presumably that would also be a client change to un-grey the physical checkbox on child prim.
[17:53]	Andrew Linden:	but at some point it feels like wishful thinking... all talk no substance.
[17:53]	Teravus Ousley:	.. err.. phantim checkbox
[17:54]	Andrew Linden needs to finish fixing bugs.
[17:54]	Teravus Ousley squishes the typo bug
[17:54]	Andrew Linden:	So we've got another few minutes to go
[17:55]	Andrew Linden:	and I don't have any other topics on my agenda
[17:55]	Areth Gall:	Indeed. Seems like it just flew by.
[17:55]	Kooky Jetaime:	Could the "noise" that caused my box to float u be responsible for the downward drag on everything else? avis, flying vehicles?
[17:55]	Creem Pye:	how difficult would it be to push back the horizontal boundaries of islands to +4096m instead of 256m?
[17:56]	Creem Pye:	(assuming that the extra space was just void water)
[17:56]	Areth Gall:	Any word on that bug I keep bringing up? Where for some reason, the rotational velocities of objects decrease without collision or deacceleration or another impulse applied?
[17:56]	Mifune Thibaud:	a megasim :)
[17:56]	Zwagoth Klaar:	dont think that will happen, ever, would require an entire rewrite of the map system and everything
[17:56]	Mifune Thibaud:	i think the current map format might be prohibitive
[17:56]	Mifune Thibaud:	although i think it'd be one of those good radical changes
[17:56]	Arawn Spitteler:	Then the Map could give ground.
[17:56]	Mifune Thibaud:	and make vehicles more functional
[17:56]	Gaius Goodliffe:	Actually, I know the map draws off-sim objects.
[17:57]	Andrew Linden:	Kooky, yes the creeping up may be related to slight damping of objects in general.
[17:57]	Arawn Spitteler:	Eventually, the Map will have to be re-coded, to accomodate the Open Grid.
[17:57]	Sofia Westwick:	you know it would be cool if we could have sims that were double or tripple the SQM size all as one sim not having to cross boarders
[17:57]	Andrew Linden:	I believe Areth has pointed out that hovering objects that rotate will slowly decelerate from some sort of internal friction.
[17:57]	Teravus Ousley:	Yes, unless it broke the massive region up into smaller regions as far as the map and other services go, it probably wouldn't be suggested.. there are also computational issues even with void areas
[17:57]	Creem Pye:	yep that's exactly what I'm getting at, Sofia =)
[17:58]	Les White:	i've been dying for a 512m sim for years.
[17:58]	Zwagoth Klaar:	didnt I give you something that showed that when it was still in beta?
[17:58]	Zwagoth Klaar:	the internal friction
[17:58]	Andrew Linden:	And the order of operations inside the physics engine's simulation step could cause slight disagreement between an anti-gravity force and gravity itself...
[17:58]	Andrew Linden:	depends on where the friction is applied
[17:58]	Sofia Westwick:	that would be leivly like have it the size of 4 sims but no boarders
[17:58]	Andrew Linden:	and the order of the "Action" apply and gravity integration step
[17:58]	Sofia Westwick:	like a square of 4 in size
[17:58]	Mifune Thibaud:	or even a special area on the grid specifically for those kinds of sims, which could sidestep the map issue (but with many moans and groans from land barons im sure)
[17:58]	Arawn Spitteler:	Would four void sims work togeather, if blocked?
[17:59]	Andrew Linden:	It would be Very Hard to expand the size of the region beyony 256.
[17:59]	Kooky Jetaime:	I think a lot of people would love to have a couple of sims that that are 1024x1024x1024 for sailing
[17:59]	Les White:	no, but the same slice of CPU should handle a true 512. in other words 1/4 of the c5 sims CPU
[17:59]	Kooky Jetaime:	or flying
[17:59]	Teravus Ousley nod nods. The client has things that require 256m also.. such as the heightfield rendering.
[18:00]	Andrew Linden:	it might be possible someday to expand the simulation to blocks of 256... but still Very Hard
[18:00]	Andrew Linden:	we would try to fix the bugs in the region crossing code first
[18:00]	Andrew Linden:	and solve collisions across region boundaries
[18:00]	Les White:	i'm sure you guys offered 512m buig sims for a very short time long ago (at 4 times the cost)
[18:00]	Teravus Ousley:	.. so you'd really have to treat each 256m area as a different sim as far as communcations go..
[18:00]	Les White:	was it all a dream?
[18:00]	Kooky Jetaime:	I think peopel would gladly give up verticle space in exchange for a lot bigger footprint to allow vehicle movement for cruising unhindered
[18:00]	Creem Pye:	I think those were really just 4 normal 256x256 sims sold as 1 unit
[18:01]	Andrew Linden:	No Les, I don't believe we ever had 512m regions.
[18:01]	Les White:	i see Creem
[18:01]	Sofia Westwick:	would even be nice if region crossing wasn't so rough.. often its easier to teleprot across then walk across.
[18:01]	Andrew Linden:	Oh... we do run 4 256^2 regions on one CPU in some cases.
[18:01]	Andrew Linden:	Er... four 256^2 regions
[18:01]	Andrew Linden:	those are often called "void regions"
[18:02]	Creem Pye:	asset pooling of sims running on the same server could make region crossings pretty smooth, I think (at least opensim has a feature like that)
[18:02]	Andrew Linden:	or "open space regions"
[18:02]	Andrew Linden:	they get a smaller CPU budget, and have a smaller prim budget
[18:02]	Arawn Spitteler:	BoundingBoxExit() would also be handy, so we can shut Physics down, until changed is called.
[18:02]	Andrew Linden:	I'm going to have to run soon.
[18:02]	Kooky Jetaime:	hmm
[18:02]	Teravus Ousley:	ok, :D take care then. Thanks. :D
[18:02]	Kooky Jetaime:	I don't seem to float down on this one
[18:02]	Les White:	thanks for your time Andrew
[18:03]	Les White:	thanks Sidewinder
[18:03]	Andrew Linden:	Thanks for comming everyone.