User:Andrew Linden/Office Hours/2008 08 19

Transcript of Andrew Linden's office hours:

[11:03]	JayR Cela:	hi there Andrew & Dakota :_)
[11:03]	Dakota Schwade:	Hi. I'm here to listen in :-)
[11:03]	Andrew Linden:	Hello
[11:03]	Ilsa Munro:	Hi guys
[11:04]	JayR Cela:	hi there iLsa & Gaius :_)
[11:05]	Cummere Mayo:	back
[11:05]	Andrew Linden:	I have the transcripts from last Thursday's hour (which I didn not attend) but haven't yet formatted them and put them up on the wiki.
[11:05]	Andrew Linden:	Lessee... announcements announcements...
[11:05]	Andrew Linden:	I think the MONO deploy is supposed to start today, or has started.
[11:05]	Kitto Flora:	On MG?
[11:06]	Cummere Mayo:	i thought that got postponed?
[11:06]	Andrew Linden:	I haven't been paying attention to our #deply channel. Yeah on the main grid.
[11:06]	Andrew Linden:	er... #deploy
[11:06]	Kitto Flora:	This AM there's no mono running on Aditi
[11:06]	Cummere Mayo:	cg said soemthing about they were looking for a showstopper on 1.24
[11:06]	Andrew Linden:	Ah could be.
[11:07]	Andrew Linden:	Hang on... lemme quickly scan my email for an update..
[11:07]	Opensource Obscure:	CG also told: today services, wed some few regions , than thursd and frid the whole grid
[11:07]	Kitto Flora:	Also nothing on the Blog
[11:07]	Dakota Schwade:	http://forums.secondlife.com/showthread.php?p=2114092#post2114092
[11:07]	Kitto Flora:	I went to Aditi to see if I could test some more
[11:07]	Opensource Obscure:	but yes then he mentioned about Cummere said
[11:07]	Dakota Schwade:	There's a mention there of rolling restarts.
[11:08]	Andrew Linden:	Ah ha! It just arrived
[11:08]	Kitto Flora:	Hot Pizza?
[11:08]	Opensource Obscure:	could you check if Vulcano sim is still in the 'test' regions that get updated in the 1st bunch?
[11:08]	Rex Cronon:	hello everybody
[11:08]	Andrew Linden:	Looks like the MONO deploy has been delayed by one day
[11:09]	Andrew Linden:	There appears to be a simstate save/load bug they expect to fix quickly.
[11:10]	Andrew Linden:	Lessee, what else...
[11:10]	Andrew Linden:	Simon, do you have any announcements?
[11:10]	Simon Linden:	No, not muc
[11:10]	Andrew Linden:	Alright, the I guess that is all, unless I remember something.
[11:10]	Simon Linden:	.not much.
[11:11]	Cummere Mayo:	any more updates abotu 2708?
[11:11]	Cummere Mayo:	or 2807 or whatever that one is?
[11:12]	Andrew Linden:	http://jira.secondlife.com/browse/SVC-2708
[11:12]	Andrew Linden:	that is the one about simulator performance problems
[11:12]	Andrew Linden:	Any updates on that Simon?
[11:13]	Andrew Linden:	Last I heard (more than a week ago now) Don and Simon were thinking about trying to deploy some experimental code for getting better info about that problem.
[11:13]	Georgette Whitfield:	Hi all sorry I'm late
[11:13]	Andrew Linden:	I don't know if anything was actually deployed. If so, it would have been done to just a few regions (not the entire grid) temporarily
[11:13]	Georgette Whitfield:	Hey Dakota
[11:13]	Cummere Mayo:	ive been tracking soem times to pull from mainland sims too see if theres anythign we can figure out before that code is ready for deploy
[11:14]	Simon Linden:	No updates on SVC-2708 ... I think there were some back-end issues resolved last week that might have contributed to occasiaional asset rezzing delays
[11:14]	Cummere Mayo:	/ao off
[11:14]	Andrew Linden:	As far as I know none of their work is in the 1.24 upate, so this is a custom experiment, and the code would have to be merged with 1.24 once that deploy goes out.
[11:15]	Cummere Mayo:	whats the soonest you could deply it ?
[11:15]	Cummere Mayo:	*deploy
[11:15]	Andrew Linden:	That is, it would have to be merged, rebuilt and redeployed again after 1.24.
[11:16]	Simon Linden:	We have the development build currently on one of our internal grids, and hope to get it onto a few main grid sims later this week or next
[11:17]	Simon Linden:	There are a few minor performance tweaks in the new code, but the main benefit is some metrics gathering that can help us compare what's going on across all the regions on a system
[11:17]	Andrew Linden:	Ideally it would be done after the region ran on 1.24 for at least a few hours, to disambiguate the cause of any measurable performance changes
[11:17]	Cummere Mayo:	simon, would you, and andrew be open for a suggestion that might intitally amke your lives mroe difficult, but i think might pay off with much mroe isntant bug catches and overall faster sim deploys?
[11:18]	Andrew Linden:	Always open to listening to suggestions.
[11:18]	Simon Linden:	Our code is based on 1.23 right now, we have two branches (another based on Mono and 1.24) so we can test both
[11:18]	Cummere Mayo:	put together a program like BSI but for servers
[11:18]	Simon Linden:	The mono release is complicating it, but that said, that's the version we are going to need to work on
[11:18]	Andrew Linden:	BSI? Please expand that acronym.
[11:19]	Opensource Obscure:	bug-catchers
[11:19]	Cummere Mayo:	Batery streed ireegulars. we test alpha versions of the viewer
[11:19]	Cummere Mayo:	*irregulars
[11:19]	Simon Linden:	Cummere - we've talked about that internally and would like to do something along those lines
[11:19]	Andrew Linden:	Ah yes. We've had some discusions about that internally already.
[11:19]	Andrew Linden:	Aric Linden wants to do something like that
[11:20]	Andrew Linden:	although, his plans is to have some Linden owned regions running some bleeding edge versions
[11:20]	Andrew Linden:	and the regions would intially be open to BSI or a similar group
[11:20]	Cummere Mayo:	*nods*
[11:20]	Andrew Linden:	I asked him why the regions wouldn't be just made open to the public
[11:21]	Cummere Mayo:	i dunno it would be a good idea to have bsi do it. if were running both alpha viewers and alpha regions might be harder to determine what is causing what
[11:21]	Andrew Linden:	He's not against it, but sounds like they would initially be of limited access to start
[11:21]	Andrew Linden:	Right, not necessarily BSI but others who might be interested in experimenting with bleeding edge simulators
[11:22]	Cummere Mayo:	you could probably get a few sim owners to donote at least one sim to the cause
[11:22]	Cummere Mayo:	im sure karoline or suz would
[11:22]	Andrew Linden:	All that said, I don't know what kind of timeline would happen for that project.
[11:22]	Simon Linden:	One of the complications there is region stabilility and preventing any content loss ... it's easy to switch to another viewer, less so for simulator versions
[11:23]	Andrew Linden:	It is on Aric's list of things he'd like to do, and Don and Simon's early experimental deploy's of their code could help iron out some of the experimental deploy issues and help speed up the project.
[11:23]	Simon Linden:	In any case, the performance work is a small step in that direction. We're going to run some of the regions that have had problems on our test code, and hopefully that will give us more information and we can move towards better fixes
[11:23]	Cummere Mayo:	true but then again if you were to say run a clone of lsot angeles so that one of those is on the alpha server one is on the standared migth show great comparisons
[11:25]	Andrew Linden:	Right, sometimes it is difficult to produce the load required to test performance, so for that we'll have to try to find willing estate owner participants, or just deploy slower as we wait for updates to go through the pipe;ine
[11:25]	Andrew Linden:	pipeline
[11:25]	Andrew Linden:	however, Aric's project is not limited to performance testing
[11:26]	Andrew Linden:	it is more of a way to get the development edge of the simulator out for more exposure sooner
[11:26]	Andrew Linden:	In any case, I'm in favor of the idea and am going to try to help Aric when he starts
[11:26]	Andrew Linden:	Anybody have any other issues to bring up?
[11:27]	Andrew Linden:	I've got one subject to talk about, but I'll wait until later if we have other things on the agenda.
[11:27]	JayR Cela:	I'm relatively happy with overall performance the past 2 weeks
[11:27]	Ilsa Munro:	Just one of my dinky dumb vehicle script questions so I can wait till last
[11:28]	Georgette Whitfield:	I think SL is getting much more stable
[11:28]	Georgette Whitfield:	Thanks guys!
[11:28]	Georgette Whitfield touches head
[11:28]	Cummere Mayo:	um.... oen thing though. the tp logout bug is back for those of us on nigthly. is that a server side thing or viewer or both?
[11:29]	Andrew Linden:	TP logout bug? I'm unfamiliar with it, I think. Do you have a jira I can read?
[11:29]	Kitto Flora:	Spurious TPs seem to be more frequent again
[11:29]	Andrew Linden:	"For those of us on nightly"... what does that mean exactly?
[11:30]	Cummere Mayo:	the nightly viewer.
[11:30]	Kitto Flora:	You try TP some place, or cross a border - and the TP just hangs, then you get forced to log out
[11:30]	Cummere Mayo:	currantly 1.21.0(94327)
[11:30]	Andrew Linden:	Ah, it must be a viewer bug then. So a bug appears to have resurfaced in the viewer code
[11:30]	Cummere Mayo:	or the system logs you out
[11:31]	Cummere Mayo:	yep
[11:31]	Cummere Mayo:	it has
[11:31]	Kitto Flora:	"You have been disconnected from the region...." type message
[11:31]	Kitto Flora:	One of the several persistent view bugs
[11:31]	Andrew Linden:	Well, I don't know anything about it, but I'll make a note and try to find the right dev to mention it to.
[11:32]	Andrew Linden:	Cummere, do you know the day that the bug appeared?
[11:32]	Andrew Linden:	That would help us narrow it down
[11:32]	Kitto Flora:	Also the 'PC locks up - only recourse is to power cycle/reset' - and theres no debug data sent to LL
[11:32]	Cummere Mayo:	the day they released the second nightly in this series
[11:32]	Andrew Linden:	Mac or Windows?
[11:33]	Andrew Linden:	Kitto, are you talking about Mac or Windows?
[11:34]	Kitto Flora:	Windows for me
[11:34]	Cummere Mayo:	widnows. ill be brignign all this up at triage tomarrow
[11:34]	Andrew Linden:	Ok thanks. There was a bug on Mac that was causing the crash-reporter to fail, so I was curious.
[11:35]	Kitto Flora:	The PC-locks-up bug has been there for months - its now one of the few thats left
[11:35]	Dakota Schwade:	I'm on a Mac. I occasionally get a crash-on-quit, with no opportunity to report it. Can't narrow down when it began, however.
[11:35]	Andrew Linden:	Ok, so what I was going to talk about was the idea of "collision hints"... different collision shapes/approximations for complex shapes. This is something we've talked about in these office hours before
[11:35]	Andrew Linden:	but I think Simon was doing some legwork on outlining the project
[11:36]	Andrew Linden:	and I thought it would be fun to bring the discussion out into the open
[11:36]	Kitto Flora:	The disconnected-on-TP , and the spurious TPs - thats turned up in the latest viewer, when they made the dazzle/original option.
[11:36]	Andrew Linden:	For those who weren't in on the earlier discussions... the basic ideas are: wouldn't it be cool to be able to specify a different (simpler) shape for objects that just need to look complicated but don't need to collide as a complicated object
[11:37]	Simon Linden:	The idea is to be able to pick a simple shape (box or sphere) to represent an object in the physics simulation rather than the higher detailed mesh we use now
[11:37]	Simon Linden:	Andrew and I are having typing races
[11:37]	Cummere Mayo:	what about those that DO need to ?
[11:37]	Andrew Linden:	there are a few distinct ways to do it:
[11:37]	Simon Linden:	It would default to the existing behavior
[11:37]	Andrew Linden:	(1) replace individual prims with simpler shapes
[11:38]	Andrew Linden:	(2) replace whole linked objects with simpler collections
[11:38]	Andrew Linden:	(3) don't include certain prims in the collision shape (basically per-prim phantom)
[11:39]	Simon Linden:	3a - per-prim phantom, but no collision events
[11:39]	Andrew Linden:	lessee... any other distinct modes? I can't remember them all
[11:39]	Cummere Mayo:	sound sgood and btw when do we get a dodecahedron as a simple one prim build sahpe?
[11:39]	Rex Cronon:	regarding #1, aren't prims simple shapes?
[11:39]	Andrew Linden:	Uh... new shapes would be a different discussion I think
[11:39]	Simon Linden:	They are simple until you twist, cut or hollow them
[11:40]	Gaius Goodliffe:	Also once you link a bunch together in a linkset, it's not so simple anymore.
[11:40]	Andrew Linden:	well, by "simple" I mean simple for the physics engine. Concave shapes in general are complex for computing colision information
[11:40]	Kitto Flora:	Whats the impact on collision_start() events ?? Particularly in a linked object where only one prim has collision_start().?
[11:40]	Andrew Linden:	so a box is simple, as is a true sphere
[11:40]	Simon Linden:	If you keep a box looking like a box, or a sphere round, we can use the simpler shapes in the physics engine. Otherwise (most of the time) it becomes some sort of mesh
[11:40]	Andrew Linden:	convex shapes are simpler than concave
[11:41]	Andrew Linden:	Kitto, could you elaborate on your question? What is WHAT's impact on collision_start()?
[11:42]	Simon Linden:	Kitto - the events shouldn't really change, except for exactly when they happen due to the different shapes
[11:42]	Simon Linden:	If a builder finds they don't want the simpler shapes, they could pick to use the existing shapes we supply to the physics engine
[11:42]	Andrew Linden:	We've already talked about the various possible ways to do it... I was hoping to move the discussion into new territory... which should be done first, when do we think we could get it done?
[11:43]	Simon Linden:	So it's totally an opt-in choice for builders ... at least that's the design we're considering now
[11:43]	Andrew Linden:	How can be get them done (new feature) in an age of fixing stability (bug fixes).
[11:43]	Kitto Flora:	If one has 10 prims linked to makea complex object, and If the collision shape is adjusted by enclosing the whole linked set in a single box for the physics engine collisions.... then if ONE of the prims had a collision_start() event in it, will that one prim still collide correctly?
[11:44]	Andrew Linden:	Ah yes, actualy collide_start() may be an incompatible feature with these proposed ideas.
[11:44]	Kitto Flora:	Pity
[11:44]	Cummere Mayo:	that might be bad then ><
[11:44]	Kitto Flora:	For my products that makes the idea moot.
[11:44]	Andrew Linden:	That is, if you really want collide_start() on individual sub-prims, you might not be able to take advantage of a simpler collision shape.
[11:44]	Gaius Goodliffe:	per-prim phantom would seem to me to be the easiest way for a builder to specify a collision shape as precisely as they need. My only concern is the possible need for additional prims. If we went with per-prim phantom, could the limits be adjusted so that the limit of prims on a physical vehicle only counts physical prims?
[11:45]	Andrew Linden:	The content owner would have to decide, and we'd have to provide some sort of UI feedback when someone tried to use both features together.
[11:46]	Andrew Linden:	Actually, the right behavior would probably be to automatically disable the simpler collision shape if per-prim collision events are used.
[11:46]	Kitto Flora:	That makes it unusable for vehicles with bumpers
[11:47]	Andrew Linden:	As to the question about per-prim phantom being used to get around the prim limit for dynamic vehicles... I think that ideally that would happen...
[11:48]	Cummere Mayo:	ok youre gonan ahte me for this one... but would it also be possibel to implimant the oposite? simple prims with complex collision scripts to give a crumple effect?
[11:48]	Cummere Mayo:	*hate
[11:48]	Andrew Linden:	however one of the problems with high-prim vehicles is that region-crossings suffer... although the complex attachment workaround also exacerbates that
[11:49]	Gaius Goodliffe:	Possibly two limits -- 32 physical prims and no more than 32 non-physical?
[11:49]	Andrew Linden:	(I had always hoped to be able to work on region crossings before lifting the prim-limit for dynamic objects)
[11:49]	Kitto Flora:	Hehe - forget all the rest, work on the region crossing
[11:49]	Gaius Goodliffe:	Actually 32 physical is probably way more than needed.
[11:50]	Andrew Linden:	Yeah Cummere... I think I'll punt that "simple objects with complex collision physics" into the future
[11:50]	Cummere Mayo:	andrew you need more help then jsut one person. yous houdl go raiding univeristies for retiring computer scientists that would like a part time job lol
[11:50]	Simon Linden:	Cummere - I think you could get that effect with per-prim phantomness. Have a transparent object with the high-detailed shape detect collisions
[11:50]	Kitto Flora:	region crossing lag is order of magnitude more a problem that vehicle prim counts, or collisions.
[11:50]	Rex Cronon:	yes, if u have the rest of the prims phantom:)
[11:50]	Andrew Linden:	Can't work on that one right now, but yes, more interesting physics would be cool.
[11:50]	Andrew Linden:	Kitto is right
[11:51]	Andrew Linden:	Which is why I was wondering "How could we get this feature done during an era of bug fixes?"
[11:52]	Kitto Flora:	Which feature?
[11:52]	Cummere Mayo grins evilly
[11:52]	Ilsa Munro grins
[11:52]	Cummere Mayo:	tahts simple andrew.... make it part of a bug fix
[11:52]	Gaius Goodliffe:	:D
[11:52]	Andrew Linden:	While region crossings are a more serious issue... that is a bigger project. Some of these feature ideas aren't much work. Most of the work is implementing a clear User Interface.
[11:52]	Ilsa Munro:	Some creative JIRA'ing perhaps?
[11:52]	Gaius Goodliffe considers how to file a bug that would require per-prim phantom to fix.
[11:52]	Simon Linden:	I'm going to try and sell this (simpler physics shapes) as a performance issue, which we know needs work
[11:52]	Cummere Mayo:	ya know that would probably work
[11:53]	Cummere Mayo:	and you could get allot of user support since it WOULD help performance
[11:53]	Gaius Goodliffe:	Oh yes. I have some airships that are just killers on performance simple because extra doodads hanging off.
[11:53]	Kitto Flora:	Does simple collision shapes or per-prim phantom help with the 'bumps at the corners' problem?
[11:53]	Gaius Goodliffe:	*simply because
[11:54]	Andrew Linden:	Well, if it were automatic, and under the hood, then I could see how it could be considered a performance fix. But if it requires content creators to provide the hints, and is incompatible with collision_start() events then it sounds like a feature.
[11:54]	Andrew Linden:	What if we could do some geometry analysis of the linked prims in a collection and figured out which ones we could get away with NOT adding to the physics engine?
[11:55]	Cummere Mayo:	as long as the system is GOOD ild go for that
[11:55]	Ilsa Munro:	Such as small decoratives and the like? Low mass = low importance?
[11:55]	Gaius Goodliffe:	I don't see how it could be done automatically. Heck, I'm not sure if it could be done by a person without asking the creator...
[11:55]	Andrew Linden:	That is, what if we could analyze the script, verify that no collision events are needed for the complicated dash of a biplane, and just not include all of those little knobs and dials in the plane's shape?
[11:55]	Rex Cronon:	r u sure that making an AI that can do that is going to be easy:)
[11:55]	Andrew Linden:	THAT would be an automatic optimization that we could make under the hood, and most residents would never notice.
[11:56]	Gaius Goodliffe:	Looking at the "whiskers" on the airship behind you -- are they ignorable or should be collide?
[11:56]	Simon Linden:	Another idea to add to the list is mesh simplification - apply a bit of triangle reduction, but that might also require a UI
[11:56]	Andrew Linden:	No Rex, dunno how easy that sort of intelligence would be... sounds hard.
[11:56]	Gaius Goodliffe:	(I'd be thrilled if they could just be ignored -- but someone else might use the same arrangement for landing gear.)
[11:56]	Andrew Linden:	well, some shapes are entirely embedded in other larger shapes. Those could be detected and discarded
[11:57]	Andrew Linden:	especially for the knobs in the dash of a airplane
[11:57]	Simon Linden:	That's a good example - the feature working correctly really depends on it being set up properly for the specific content
[11:57]	Andrew Linden:	er... such as
[11:57]	Kitto Flora:	If no daughter prim collision<x>() events, then just used a bounding-box of the linked object for physics collision?
[11:58]	Rex Cronon:	so, if an ave sits inside a sphere, the avatar would have no collisions?
[11:58]	Cummere Mayo:	if soemthign is entirely burried within a physical prim chances are it doesnt need to collide. even jsut turning those off, i bet you could gain a couple percentage points in sim perfromance
[11:58]	Andrew Linden:	naturally, the autoatic algorithm couldn't detect ALL of the optimizations that a content creator could do, but applied across the board it might produce enough performance boosts to be worth it
[11:58]	Rex Cronon:	actually it might make more lag
[11:59]	Andrew Linden:	Sometimes people double-up, or triple-up one shape in the exact same spot. Such duplications could be detected and optimized.
[11:59]	Cummere Mayo:	yeah i did that for a staff i made
[11:59]	Andrew Linden:	yes Rex, a sitting avatar embedded in a larger shape could be optimized out.
[12:00]	Rex Cronon:	if the object creators would have a way to turn that off, maybe
[12:00]	Andrew Linden:	Well, such an automatic system would take longer than a simple per-prim-phantom hint provided by content creators
[12:00]	Rex Cronon:	why?
[12:01]	Andrew Linden:	Also, the shape analysis could take a while... some objects might not be able to be analyzed real-time... could take a significant fracction of a second... depending on how smart and fast the algorithm was
[12:02]	Kitto Flora imagines some baking process ...
[12:02]	Andrew Linden:	so in order to avoid blocking the simualtor for fractions of a second we'd have to offload it to another process, or crunch the numbers piecemeal over several frames
[12:02]	Andrew Linden:	Finally, such an algorithm would probably make a good paper for Siggraph
[12:03]	Andrew Linden:	(perhaps such a paper has already been written)
[12:03]	Georgette Whitfield:	I need to head y'all
[12:03]	Georgette Whitfield:	Thanks and bye
[12:03]	Rex Cronon:	bye
[12:03]	Andrew Linden:	Yeah, I've got to go too. Hour's up.
[12:03]	JayR Cela:	yep me too
[12:03]	Ilsa Munro:	D'oh
[12:04]	JayR Cela:	byeee everyone :_)
[12:04]	Ilsa Munro:	Quick question?
[12:04]	Rex Cronon:	bye everybody
[12:04]	Cummere Mayo:	wella ndrew you sold me on your ideas. if you want to try and sell your bsoses on it inwworld feel free to bug me to come ass my support
[12:04]	Rex Cronon:	have fun
[12:04]	Andrew Linden:	Simon, lets talk more about how to go forward on this later today, offline.
[12:04]	Georgette Whitfield:	Thanks so much for this great world dev guys
[12:04]	Gaius Goodliffe:	Yes, thank you gentlemen.
[12:04]	Andrew Linden:	ok cummere, I'll mention your name as a possible volunteer to the right people
[12:04]	Cummere Mayo:	*Ask
[12:05]	Kitto Flora:	Lunchtime!
[12:05]	Kitto Flora:	Byebye all
[12:05]	Gaius Goodliffe:	yay
[12:05]	Ilsa Munro:	Okay, I've got a single prim with a simple vehicle script. I want to have it lift straight up so I'm using llSetVehicleVectorParam(VEHICLE_LINEAR_MOTOR_DIRECTION, <0,0,1>);
[12:05]	Simon Linden:	sure, let's talk Andrew. Thanks everyone for coming
[12:05]	Ilsa Munro:	Going from a standing start in my head it should lift more or less perpindicular to the ground, but I seem to pick up forward motion as well
[12:05]	Ilsa Munro:	And it's not the easiest thing to search up in the forums I'm finding
[12:05]	Ilsa Munro:	Any hints?
[12:05]	Andrew Linden:	Ilsa, I think perhaps you have some other parameters set that are interfering with the motion you want...
[12:06]	Andrew Linden:	lemme see if I can remember them
[12:06]	Andrew Linden:	the deflection parameters in particular
[12:06]	Andrew Linden:	and also the friction
[12:06]	Andrew Linden:	VEHICLE_LINEAR_DEFLECTION_TIMESCALE
[12:06]	Aybabtu Aabye:	o damn, off by an hour
[12:06]	Andrew Linden:	and also VEHICLE_FRICTION_TIMESCALE
[12:06]	Ilsa Munro:	3
[12:07]	Ilsa Munro:	Which should mean it takes 3 seconds to start moving on any linear axis if I'm understanding
[12:07]	Aybabtu Aabye:	hehe any chance this is a two hour session?
[12:07]	Ilsa Munro:	Def Timescale
[12:07]	Ilsa Munro grins
[12:07]	Ilsa Munro:	You can listen to me whine beg and plead Aybabtu but it's not likely to be too enlightening
[12:07]	Andrew Linden:	Ilsa, that deflection parameter can be thought of as...
[12:08]	Andrew Linden:	hrm... I'm trying to remember if the deflection timescale is a vector or a scalar
[12:08]	Ilsa Munro:	Scalar
[12:08]	Aybabtu Aabye:	scalar
[12:09]	Aybabtu Aabye:	a question on friction vs. motors and timescales
[12:09]	Andrew Linden:	alright, so it only deflects into the vehicle's forward (backward) axis
[12:09]	Aybabtu Aabye:	is it possible to have the timescale be a vector on the motor
[12:09]	Ilsa Munro:	Ahhh, the linear settings only apply to the fb axis and not all non rotational axes?
[12:09]	Aybabtu Aabye:	one thing i come up against is the motor acts friction unintentionally
[12:09]	Andrew Linden:	so if your vehicle is moving UP... then it will redirect its velocity into its forward direction... with halflife approximately equal to the timescale you specify
[12:10]	Andrew Linden:	right, it is a linear deflection
[12:10]	Simon Linden:	I have to run ... bye everyone
[12:10]	Andrew Linden:	the ANGULAR_DEFLECTION
[12:10]	Ilsa Munro:	Thanks Simon
[12:10]	Aybabtu Aabye:	if i set the motor to <40,0,0> and the timescale to say 4
[12:10]	Aybabtu Aabye:	bye simon
[12:10]	Andrew Linden:	that one is different... it reorients the vehicle to point along its velocity
[12:10]	Ilsa Munro nods
[12:10]	Andrew Linden:	think of the angular deflection as the fins on the end of a dart
[12:11]	Andrew Linden:	if you tumble the dart forward, it will orient point first
[12:11]	Andrew Linden:	that is what the angular deflection timescale does
[12:11]	Aybabtu Aabye:	then it motor will take 4 seconds to get to 40 in the x direction but also damps y and z to 0 in 4 seconds
[12:11]	Aybabtu Aabye:	acting like friction when i dont want that
[12:11]	Andrew Linden:	to turn off the timescales set them very large... 1000 seconds, so that it takes forever for the effect to happen
[12:12]	Ilsa Munro:	Ahh, I think I had it nearly 100% backwards then
[12:12]	Andrew Linden:	The friction coefficient is a vector, so that you can apply different friction along the various axes
[12:12]	Ilsa Munro nods
[12:12]	Andrew Linden:	so you can specify low friction in the forward direction (X) and high friction in the sideways (Y)
[12:12]	Ilsa Munro:	ground vehicles have low friction on FB but high on LR
[12:13]	Andrew Linden:	right
[12:13]	Andrew Linden:	ok so sounds like that will help?
[12:13]	Ilsa Munro:	Got me pointed in the right direction I hope
[12:13]	Ilsa Munro:	Thanks a Ton Andrew
[12:13]	Andrew Linden:	cool. Good luck.
[12:13]	Ilsa Munro waves
[12:14]	Aybabtu Aabye:	andrew got time for one more after hours question?
[12:14]	Andrew Linden:	Alas no Aybabtu
[12:14]	Andrew Linden has much work to do.
[12:14]	Aybabtu Aabye:	hehe ok, ill try to catch up with you thursday, on time i hope