User:Andrew Linden/Office Hours/2010 10 01

Transcript

[16:00]	Moon Metty:	hey Ardy :)
[16:00]	Ardy Lay:	Hi
[16:01]	Kaluura Boa:	Shhhhh! I'm sleeping
[16:01]	Moon Metty:	/me squeeks
[16:02]	Moon Metty:	oh wait
[16:02]	Moon Metty:	there's Andrew!
[16:02]	Andrew Linden:	Hello.
[16:02]	Moon Metty:	so you could make it after all
[16:02]	Andrew Linden:	Simon won't be attending today. He's on vacation.
[16:02]	Andrew Linden:	Yeah, I was thinking about leaving early, but the occasion I was going to run off to has been delayed.
[16:03]	Moon Metty:	ok
[16:03]	Moon Metty:	hi Rex :)
[16:03]	Rex Cronon:	hi moon:)
[16:03]	Rex Cronon:	hi everybody
[16:03]	lonetorus Habilis:	hey andrew
[16:03]	lonetorus Habilis:	and the rest of ya too of course
[16:04]	Rex Cronon:	hi
[16:04]	Andrew Linden:	I finally got my maint-server work up on aditi today.
[16:04]	Andrew Linden:	It is on the following regions should anyone want to play with the SVC-93 (SVC-129) fix:
[16:04]
[16:04]
[16:04]	Moon Metty:	ah, cool
[16:04]	Andrew Linden:	Oatmeal 9, Oatmeal 13, Oatmeal 14, and Sandbox Cordova
[16:04]	Andrew Linden:	So, all of those regions on the beta grid.
[16:05]	reddot99 Republic:	see if we can break it or something?
[16:05]	Moon Metty:	ok
[16:05]	Moon Metty:	what's the syntax?
[16:05]	Andrew Linden:	I think this work will be available as a "Release Candidate" on agni in... another 1.5 weeks
[16:05]	Andrew Linden:	The param name is PRIM_ROT_LOCAL
[16:05]	Rex Cronon:	r those oatmeals specials, adrew:)
[16:06]	Moon Metty:	excellent :)
[16:06]	Andrew Linden:	I posted my internal testplan in the comments of SVC-93, and there is an example script there.
[16:06]	Rex Cronon:	andrew*
[16:06]	Andrew Linden:	Special in what way Rex?
[16:06]	Andrew Linden:	If they are special I don't know in what way.
[16:06]	reddot99 Republic:	the oatmeal regions often are where new functions get deployed for testing
[16:07]	Andrew Linden:	I replaced the old "Havok 7 Server" channel on aditi (which had a partially working llCastRay())
[16:07]	Rex Cronon:	i don't know. that is why i am asking. like ability to se who runs what scripts, what meme each scriptuses. stuff like that
[16:07]	reddot99 Republic:	wait, raycast is gone again?
[16:07]	Andrew Linden:	The Oatmeal regions are just test regions as far at I know. Owned by Oskar Linden.
[16:07]	Andrew Linden:	Yeah, llRayCast() is not available on aditi for the moment.
[16:08]	Jonathan Yap:	RayCast will be out with mesh beta Oskar said yesterday
[16:08]	Andrew Linden:	I'll see about getting it back up next week if anyone realy wants to play with it.
[16:08]	reddot99 Republic:	oh, thats good enough for me then
[16:08]	Andrew Linden:	Howerver, I know that Falcon has done some new work on llCastRay() and that version has not been available yet.
[16:08]	Moon Metty:	isn't raycast part of the mesh servers?
[16:09]	reddot99 Republic:	though, i still think it needs to be renamed, because its more like "cast line segment"
[16:09]	Techwolf Lupindo:	/me returns
[16:09]	Andrew Linden:	He's been trying to implement throttles for llCastRay() calls to address worries about new sim-lag modes
[16:09]	Andrew Linden:	yeah, I think Falcon is proceeding with llCastRay() in the mesh project.
[16:09]	Andrew Linden:	He might show up for this OH today. If so we can ask him.
[16:09]	reddot99 Republic:	hey andrew, there was something that i wondered if havok was capable of.
[16:09]	Moon Metty:	yes
[16:09]	reddot99 Republic:	called a sweep test
[16:09]	Andrew Linden:	/me looks around to see if Falcon is already here.
[16:10]	Techwolf Lupindo:	Any word on the huge number of reports of failed bakes and breakage of products due to breakage of the offline/online code?
[16:10]	reddot99 Republic:	not here yet,
[16:10]	Andrew Linden:	yes reddot99, Havok can do sweep tests, however they are not used or enabled in our physics engien ATM, I think.
[16:10]	Andrew Linden:	The sweep tests come at a cost... they can be laggy.
[16:11]	reddot99 Republic:	how intensive would they be if enabled for only single temp rez prims moving above a certain velocity?
[16:11]	Andrew Linden:	Also, one feature that is possible in my andrew-maint-server channel on aditi is the "Tri-Sampled Heightfield"
[16:11]	Andrew Linden:	which provides more correct collisions at a slight increase in computation costs
[16:11]	Moon Metty:	tri-sampled ..
[16:12]	Andrew Linden:	yeah, basically it actually produces triangles for terrain collisions
[16:12]	reddot99 Republic:	triple checks its work?
[16:12]	reddot99 Republic:	ah, now if only we could get caverns out of linden land
[16:12]	Andrew Linden:	the heightfield implementation we're currently using is called the "sampled heightfield" and just does a single height for a 1m^2 location.
[16:13]	reddot99 Republic:	that why terraforming gets so messy?
[16:13]	Andrew Linden:	No, these are heightfields which cannot do hollow (multiple heights per (x,y) point)
[16:13]	Andrew Linden:	however, mesh terrain might eventually be possible once we get mesh objects working
[16:13]	reddot99 Republic:	andrew, the caverns was wishful thinking :p
[16:13]	Pauline Darkfury:	With something like that, how feasible would it be to have that auto-disable and fallback to the less computationally intensive method when a sim gets loaded beyond a certain point?
[16:14]	Andrew Linden:	no reddot99, the messy terraforming has to do with our legacy "lossy compression streaming format" of the terrain data
[16:14]	Andrew Linden:	Hrm... that is a good question for Falcon Pauline.
[16:15]	Andrew Linden:	I'm not sure what the cost would be for swapping between a tri-sampled heightfield and a sampled heightfield.
[16:15]	Pauline Darkfury:	Either that, or make it an EM tunable, so we can sacrifice accuracy for performance at our own choice.
[16:15]	reddot99 Republic:	i think the fact that its essentially a 256x256 hieghtmap for an entire region comes into play a bit too
[16:15]	Andrew Linden:	that is, swapping real-time might not be profitable
[16:15]	Andrew Linden:	anyway, the tri-sampled heightfield won't be available for experimentation until next week
[16:15]	Pauline Darkfury:	Well, if both fields were computed and stored, hopefully not too expensive, as long as it is damped to avoid hysteresis
[16:16]	Andrew Linden:	I think it can only be enabled by an estate owner with the right viewer
[16:16]	Andrew Linden:	so if anyone is interested in messing around with it they should contact me or Falcon next week sometime
[16:16]	Moon Metty:	ok
[16:17]	Moon Metty:	could that also affect the visibiltiy of small objects close to the ground?
[16:17]	Andrew Linden:	yeah Pauline, however to swap between the two "shapes" there might be some overhead in the physics engine "world management"
[16:17]	Andrew Linden:	Falcon would know.
[16:17]	reddot99 Republic:	i dont think it'll affect the terrain rendering
[16:17]	Andrew Linden:	We can do real-time switching between the two, but atm we only do it upon a special command from the viewer UI
[16:18]	Andrew Linden:	Correct reddot99, it has nothing to do with terrain rendering, only physics collisions.
[16:18]	reddot99 Republic:	theres falcon, 130 meters away
[16:18]	Andrew Linden:	Let's see... other stuff that might be interesting in andrew-maint-server on aditi...
[16:18]	Andrew Linden:	the terrain collisions should be a bit faster. I did some optimizations there.
[16:19]	Andrew Linden:	A griefer rez-flood crash mode is fixed.
[16:19]	reddot99 Republic:	hey falcon
[16:19]	Falcon Linden:	hey folks
[16:19]	Pauline Darkfury:	hello :)
[16:19]	Moon Metty:	hey Falcon :)
[16:19]	Rex Cronon:	hi falcon
[16:19]	reddot99 Republic:	that the rez queue overflow? or something else?
[16:19]	Andrew Linden:	That's about it that is interesting to the public, except for the simconsole support (which is used to try the tri-sampled heightfield)
[16:20]	Andrew Linden:	reddot99, there is a rez mode where if you can request lots of stuff in a quick burst you can crash before the grey goo fence gets much chance to block it.
[16:20]	reddot99 Republic:	hey falcon, question about havok, how effeciently can it do sweept testing?
[16:20]	Jonathan Yap:	Will people be able to test the sim console on aditi even if they do not own a region there?
[16:20]	Falcon Linden:	Sweep testing?
[16:20]	reddot99 Republic:	yeah
[16:21]	Falcon Linden:	Meaning...?
[16:21]	Andrew Linden:	swept object collision queries, Falcon
[16:21]	Falcon Linden:	you mean continuous collision detection?
[16:21]	Andrew Linden:	yes
[16:21]	reddot99 Republic:	like for fast moving objects that currently jump over walls
[16:21]	Rex Cronon:	how does this sweep thing work?
[16:21]	Falcon Linden:	IT can do it very efficiently relative to other physics enignes.
[16:21]	Falcon Linden:	engines*
[16:21]	Falcon Linden:	But it's not the kind of thing you use for bullets :)
[16:22]	reddot99 Republic:	explosives like rockets you would want it to have though
[16:22]	Andrew Linden:	As I understand it, continuous collision queries near a tri-sampled heightfield could be a recipe for lag disaster.
[16:22]	Falcon Linden:	for the rocket itself, yeah, not for debris though
[16:23]	Falcon Linden:	andrew: oh yeah. heh.
[16:23]	reddot99 Republic:	i would be happy if it had a toggle that required temp rez and one prim
[16:23]	Falcon Linden:	No.
[16:23]	Falcon Linden:	:)
[16:23]	Falcon Linden:	There will be no continuous collision detection until every other lag point and feature in Second Life has been addressed. At least, that's my position. :)
[16:23]	Andrew Linden:	Another question that Pauline had for you Falcon: how hard would it be to switch real-time between the tri-sampled and sampled heightfields, in response to too much physics lag for example?
[16:23]	Jadee Saxena:	HIIIII
[16:24]	Techwolf Lupindo:	reddot gave me an idea. Will the sim console be aviable to non-owners/EMs to allow tech support folks to look at the stats to troubleshoot problems in the sim?
[16:24]	Falcon Linden:	Well, you can switch in real-time with the console. You odn't have to restart the sim anymore.
[16:24]	Falcon Linden:	I don't think.
[16:24]	Rex Cronon:	hi
[16:24]	Jadee Saxena:	hi rex
[16:24]	Jadee Saxena:	where am i??
[16:24]	reddot99 Republic:	that or actually just use a raycast to jump to the first collision point, that'd be good enough for me,
[16:24]	Andrew Linden:	Yes, but could you hook up the switch to the RCCS (runtime collision control system) for emergencies?
[16:24]	Rex Cronon:	welcome to andrew's office hour
[16:25]	Falcon Linden:	Techwolf: there are no official policies. It'll be up to the individual who implements a given feature to decide who gets access to it.
[16:25]	Falcon Linden:	andrew: yeah, that could certainly be done.
[16:25]	Falcon Linden:	reddot99: /that's/ the way you should be building things
[16:25]	Jadee Saxena:	what this???
[16:25]	Falcon Linden:	use a raycast to figure out where it will end up and send it there.
[16:25]	Rex Cronon:	is about techinical details of how everything around here works:)
[16:25]	Jadee Saxena:	wow
[16:25]	reddot99 Republic:	yeah, but that doesnt work when you cant get from one place to a nother in 1 hop, instantly
[16:25]	Andrew Linden:	Jadee, this is a discussion session about Second Life bugs and features.
[16:26]	Pauline Darkfury:	basically, from my point of view, improved terrain collisions sound great, but not if that slows already heavily loaded sims, so dynamically switching to the lower cost method under heavier load conditions would be ideal (with damped return to the tri-sampling to avoid hysteresis)
[16:26]	reddot99 Republic:	still stuck with the warppos method, or brute forcing a big list for a path
[16:26]	reddot99 Republic:	though the primitive params
[16:27]	Falcon Linden:	well, that's a separate issue we should resolve
[16:27]	Falcon Linden:	not create more hacks to handle that hack
[16:27]	Andrew Linden:	yes Pauline but a few people might want to trade some computational efficiency for more correct collisions, race-em regions perhaps.
[16:27]	Andrew Linden:	We're not sure, but we're making it available and we'll experiment.
[16:27]	Falcon Linden:	to be honest, the tri-sampled heightfield feature was never really intended to be a "feature"
[16:27]	Falcon Linden:	it was just a quick thing we threw in as a test case for the console
[16:28]	Andrew Linden:	We havent' even really benchmarked it much to see how much lag the tri-sampled heightfield introduces.
[16:28]	Falcon Linden:	because we were already doing work on the heightfield generally
[16:28]	Andrew Linden:	At least I haven't benchmarked it. Falcon may have done a little bit.
[16:28]	Chloee Brandi:	hi
[16:28]	Chloee Brandi:	urm
[16:28]	Chloee Brandi:	sorry
[16:28]	Chloee Brandi:	ive just been urm
[16:28]	Chloee Brandi:	in the bath
[16:28]	Chloee Brandi:	ill just change
[16:29]	Pauline Darkfury:	I guess the keep everyone happy method would be to allow tri-sampling to be on/off/auto
[16:29]	Andrew Linden:	Other news... I got the new linkability code written today -- not deployed yet.
[16:29]	Andrew Linden:	the goal with the linkability changes was to make the linkability checks more efficient
[16:29]	Rex Cronon:	is the max distance 64?
[16:29]	Andrew Linden:	I haven't benchmarked that yet either. I need to make some llSetLinkedPrimitiveParameterFast() test for it.
[16:30]	reddot99 Republic:	would that also include being more logical?
[16:30]	Falcon Linden:	I haven't benchmarked it either, Ijust know it's slow.
[16:30]	Jadee Saxena:	jadee saxena
[16:30]	Andrew Linden:	No Rex, I wanted to make it as small as possible without breaking legacy linked objects, so I made the bounding sphere 54 meters (for the prim centers)
[16:30]	Falcon Linden:	because it creates triangles on the fly. The real problem isn't that the implementation is slow, it's that our terrain in SL is far too highly tesselated
[16:30]	Andrew Linden:	so the effective outer boundary will be ~64m
[16:30]	Falcon Linden:	Honestly, I think mesh will revolutionize terrain anyway
[16:31]	reddot99 Republic:	wish you had made it larger, rather than trying for miniumal size
[16:31]	Andrew Linden:	It can always be expanded, but it can never be reduced.
[16:31]	Rex Cronon:	i guess we can finally make a ground that looks like dali drawing:)
[16:31]	Rex Cronon:	like in a*
[16:32]	reddot99 Republic:	theres a test tool for ya, andrew
[16:32]	Jadee Saxena:	Plank??
[16:32]	lonetorus Habilis:	nice plank
[16:32]	reddot99 Republic:	telescoping bridge
[16:32]	Jadee Saxena:	ohhh
[16:32]	lonetorus Habilis:	telescoping plank
[16:33]	Jadee Saxena:	please
[16:33]	reddot99 Republic:	would be simple enough to make it keep tring to expand and show how where it gets stuck,
[16:33]	Andrew Linden:	reddot99, how does that bridge benchmark it?
[16:33]	Andrew Linden:	Does it actually measure the time it takes to complete a shape change?
[16:33]	reddot99 Republic:	if it gets expanded too far, then reset, then it should be fine, and i can do that quick
[16:34]	Andrew Linden:	I didn't have time to find the jira issue about the slowness of llSLPPFast()
[16:34]	Andrew Linden:	but I think there is one, Kelly showed it to me a month ago or so.
[16:34]	Moon Metty:	yes, there is one
[16:34]	lonetorus Habilis:	in what cases?
[16:34]	Andrew Linden:	Well, maybe there is an example script on it. I'll look for it next week.
[16:35]	reddot99 Republic:	but look, theres times
[16:35]	Moon Metty:	the lsl function itself is fast, i think
[16:35]	Andrew Linden:	Yes, but the LSL function is limited by the cost of the code trying to figure out if the result is linkable or not.
[16:35]	Moon Metty:	SVC-5687
[16:35]
[16:36]	Andrew Linden:	The best benchmark object would probably have the maximum number of prims (256)
[16:36]	reddot99 Republic:	just add more prims to this and it works
[16:36]	Andrew Linden:	Thanks for that link Moon. I think that was the one I was looking for.
[16:37]	Moon Metty:	:)
[16:37]	lonetorus Habilis:	reddot can i have a copy to play with ?
[16:37]	Moon Metty:	i used the search words "delay visually" because i remembered the title
[16:37]	lonetorus Habilis:	ty :)
[16:38]	lonetorus Habilis:	/me replaces jira search function with moon
[16:38]	Moon Metty:	hehehe
[16:38]	Rex Cronon:	u r better than a database, moon:)
[16:38]	Moon Metty:	thank you Rex
[16:38]	reddot99 Republic:	but andrew, would the new linkablity rules allow prims in any allowed portion of the area to move to any other allowed position?
[16:39]	Andrew Linden:	Hrm... good question reddot99
[16:39]	reddot99 Republic:	thats what this would be useful to test
[16:39]	Andrew Linden:	I don't think I'd want to change that at the same time.
[16:40]	reddot99 Republic:	i'd rather it stop leaving things hanging in middle of nowhere on builds like this
[16:40]	Andrew Linden:	This change is primarily motivated by improving performance
[16:40]	Andrew Linden:	I really want to reduce the risk of changing LSL behavior, if possible.
[16:40]	Andrew Linden:	I'm sure I'll break something, or the change could get caught up in committee.
[16:40]	Moon Metty:	yes
[16:41]	reddot99 Republic:	andrew. currently to undo the effects of that, it takes a good size loop to correct for
[16:41]	Andrew Linden:	reddot99, you mean to work around the 10m position change limit?
[16:42]	reddot99 Republic:	that isnt the issue, within a object, thats seemingly nonexsistant
[16:42]	Andrew Linden:	I'm not sure I know of the exact LSL problem you're talking about then.
[16:42]	reddot99 Republic:	its where one prim your trying to move fails to move, because a nother prim farther down your loopp needs it to stay still to folllow linkablity rules
[16:43]	Andrew Linden:	Oh I see. No, the linkability rules would be much simpler under the new scenario.
[16:43]	Rex Cronon:	depending how good your lsl scripts r, u might be able to overcome those limitations:)
[16:44]	reddot99 Republic:	would it prevent that from happening though?
[16:44]	Andrew Linden:	Basically, the current plan is this: if the prim centers fit inside a 54m diameter sphere then the link would work -- irregardless of prim scales.
[16:44]	Andrew Linden:	That's the current plan. We'll see if it has to be modified.
[16:44]	reddot99 Republic:	rex, a simple script is preferable to a complex one if both are designed to achieve the same objective
[16:45]	Rex Cronon:	it doesn't have to be complex:)
[16:45]	reddot99 Republic:	every additioonal line of code addds to complexity
[16:45]	Andrew Linden:	Falcon, does llCastRay() currently work in the mesh regions on aditi?
[16:46]	Falcon Linden:	yeah, but I'm completely overhauling the API
[16:46]	Falcon Linden:	err
[16:46]	Falcon Linden:	actually
[16:46]	Falcon Linden:	not sure
[16:46]	Falcon Linden:	It depends on when the last deploy was done
[16:46]	Rex Cronon:	until we have llScale(uuid key, float percent), we will have complexity:)
[16:46]	Falcon Linden:	it will work in some way or another soon
[16:46]	Andrew Linden:	I see, but future deploys to mesh should eventually pick up llCastRay() again.
[16:46]	Falcon Linden:	definitely
[16:46]	Jadee Saxena:	please
[16:47]	reddot99 Republic:	speaking of mesh, how're their collisons calculated compared to normal objects?
[16:47]	Falcon Linden:	although the API is going to be mostly different
[16:47]	Andrew Linden:	Ok good. I replaced the old "Havok 7 Server" channel that had the llCastRay() prototype working.
[16:47]	Falcon Linden:	reddot99: that is a complicated topic
[16:47]	Falcon Linden:	You'll have a lot of options when you import meshes
[16:47]	reddot99 Republic:	as effective or significantly less than a cube
[16:47]	Falcon Linden:	in terms of controlling the physics representation
[16:47]	reddot99 Republic:	i mean, just the triangles versus a prim
[16:47]	Falcon Linden:	haha, actually, right now, when the server glitches when trying to grab an asset, you get a cube :)
[16:48]
[16:48]	Falcon Linden:	"triangles vs prim"? you mean, how efficient is a mesh collision compared toa prim collision?
[16:48]	reddot99 Republic:	yeah
[16:48]	Falcon Linden:	impossible to generalize.
[16:48]	Falcon Linden:	That's why we're introducing a new system for prim costs that will account for object-specific physics cost
[16:49]	reddot99 Republic:	so prim cost wont be based on appearance but on physics impact?
[16:49]	Falcon Linden:	A big mesh with just a few fat triangles? Dirt cheap. As cheap as a box or maybe cheaper.
[16:49]	Falcon Linden:	reddot99: it'll be based on both.
[16:49]	Falcon Linden:	A small mesh with lots of triangles? Very expensive.
[16:50]	Falcon Linden:	But you'll have a very important alternative. You won't need to use the mesh for the physics
[16:50]	Pauline Darkfury:	how does that new prim cost system impact on a sim with 15,000 prims of various configurations and shapes? Will all existing sims still be under the total limit? Or is it just for mesh stuff?
[16:50]	reddot99 Republic:	so your allowing us to use a seprate mesh for collision, or at least, possibly a much simplified version, and i'm thinking buildings,
[16:50]	Falcon Linden:	we'll be providing a utility to auto-generate a simplified physics representation, and you'll be able to turn off physics on a mesh completely and hand-build a proxy out of invisible prims
[16:50]	Falcon Linden:	reddot99: that's correct, yes.
[16:51]	reddot99 Republic:	ok, so same as people do with sculpts effectively?
[16:51]	Falcon Linden:	Pauline: Legacy prims will not be affected. For legacy prims, prim count = new prim resource cost. Total sim prim resources = 15,000. So while new objects could take more (or less) resources than their actual prim count, old objects will continue to be 1:1
[16:51]	Falcon Linden:	reddot99: except better. But yes.
[16:52]	Falcon Linden:	and there will be tools for generating physics proxies
[16:52]	reddot99 Republic:	will meshs be solid, or hollow for physics approximations?
[16:53]	reddot99 Republic:	like, the area between parallel faces on a object
[16:53]	Pauline Darkfury:	If it's all about physics cost under the new system, how about allowing legacy phantom prims to be free or say half cost?
[16:53]	Falcon Linden:	reddot99: hollow.
[16:54]	Falcon Linden:	well
[16:54]	Falcon Linden:	hmm
[16:54]	Falcon Linden:	they'll be hollow if you use a mesh. Solid if you use our simplification tools.
[16:54]	Falcon Linden:	Pauline: Because we aren't willing to chang ethe cost on existing items and because phantom is not the same as no shape.
[16:54]	Falcon Linden:	phantom prims still exist in the physics engine. Prims with the new physics shape type "none" do not.
[16:55]	Andrew Linden:	"phantom" actually means "collides with terrain only"
[16:55]	Pauline Darkfury:	ahh, ok, thanks for the explanation, makes sense
[16:55]	reddot99 Republic:	what about the volume detect falling through land bug,
[16:55]	Falcon Linden:	That's not a bug, it's a feature!
[16:55]	Falcon Linden:	...
[16:55]	Liisa Runo:	yea
[16:55]	Moon Metty:	hehe
[16:55]	Falcon Linden:	I hate volume detect.
[16:55]	Pauline Darkfury:	I guess we'll have an incentive to convert existing phantom stuff to no-shape mesh then ;)
[16:55]	reddot99 Republic:	been used like one, yes, but its not intended
[16:55]	Falcon Linden:	Pauline: yep :)
[16:56]	Falcon Linden:	no-shape won't just be for meshes, though
[16:56]	Falcon Linden:	it can be used on prims
[16:56]	Falcon Linden:	(just not root prims)
[16:56]	reddot99 Republic:	ok. that was my next question. lol
[16:56]	Pauline Darkfury:	Oh, cool, so we can potentially just set a new flag on prims, akin to phantom, which will both reduce sim load and cost?
[16:56]	Liisa Runo:	phantom and volumedetect both are very needed but we could use 3rd that is as phantom as volumedetect but without collision calculations
[16:57]	reddot99 Republic:	and falcon, i actually made a decent toy that uses both volumedetect and standard physics to activate two different modes of action,
[16:57]	Falcon Linden:	Liisa: yes, that's "no physics"
[16:57]	Falcon Linden:	err, "none"
[16:58]	Falcon Linden:	Pauline: Reduce sim load, yes. Reduce cost? Well, if the rest of your linkset is efficiently built, absolutely. But if it has a bunch of torii or twisted, cut prims that /aren't/ set to "none", the price will likely go way up :-P
[16:58]	reddot99 Republic:	if solid rootprim is hit, it goes off imediately, otherwise it waits till its passed to the far side of a wall,
[16:58]	Falcon Linden:	That's because we're using the new shape type flags as a way for resis to opt into the new accounting system
[16:59]	Falcon Linden:	red: cool
[16:59]	Pauline Darkfury:	right, and we'll be able to opt-out per object if we find an object has a horrific cost under the new system?
[16:59]	reddot99 Republic:	i wanted to make a rocket that exploded after going through a window or simialar gap
[16:59]	Falcon Linden:	pauline: per linkset, yes.
[16:59]	Pauline Darkfury:	sounds great :)
[17:00]	Falcon Linden:	I'd suggest in the cases of high costs in the new system that you improve it so it has a lower cost :)
[17:00]	Pauline Darkfury:	yeah, but being able to flip back and forth lets us quickly optimise until we have time to rebuild
[17:00]	Falcon Linden:	In the new system, a linkset can cost as little as 1/2 the number of prims it contains. (All numbers subject to change without notice...blah blah blah :P)
[17:00]	Falcon Linden:	Pauline: yep
[17:01]	reddot99 Republic:	also, falcon, thanks for putting up with all the questions about this, never heard about it before now,
[17:01]	Falcon Linden:	Also, I'm working right now on making it so setting a linkset physical, if it's opted into the new accounting system, will be determined only by its physics cost, not by number of prims.
[17:01]	Falcon Linden:	so you could have arbitrarily detailed vehicles as long as it had a very simple physics shape
[17:02]	reddot99 Republic:	so no more needing to make half attachment aircraft?
[17:02]	Andrew Linden:	I thought the render cost would also kick in if it were too high, Falcon.
[17:02]	Andrew Linden:	"arbitrarily detailed"... that doesn't sound like the Falcon I know.
[17:03]	Jadee Saxena:	jadee saxena
[17:03]	reddot99 Republic:	you can achieve as much detail with textures often as you can with sculpts
[17:03]	Falcon Linden:	andrew: Oh, don't get me wrong, it might cost you 15,000 prims
[17:03]	Falcon Linden:	it might be the only thing you can on your sim if it's detailed enough
[17:03]	Falcon Linden:	but if the physics is just a box, why not make it dynamic? :P
[17:03]	Andrew Linden:	Ah ok I see.
[17:03]	reddot99 Republic:	trying to watch the viewer drown in update messages?
[17:04]	Falcon Linden:	it's just a proper fix for the "Wear most of your ship" exploit
[17:04]	Falcon Linden:	the viewer should only get an update for the linkset as a whole.
[17:04]	reddot99 Republic:	though, that doesnt make it so that in mouselook you can see out, which is a half issue
[17:04]	Falcon Linden:	whatever.
[17:04]	Falcon Linden:	:P
[17:04]	Falcon Linden:	always complaining
[17:04]	reddot99 Republic:	lol,
[17:04]	Falcon Linden:	I give you skyships and you whine about mouselook :)
[17:05]	Moon Metty:	hehe
[17:05]	Pauline Darkfury:	lol
[17:05]	reddot99 Republic:	i'm thinking small planes, not giant ones
[17:05]	Andrew Linden:	Nah, they can just attach the skyship -- mouselook works again.
[17:05]	Falcon Linden:	oh, in that case, let me change the mesh schedule to figure out a way to integrate mouse look...for small planes ;)
[17:05]	Moon Metty:	well, it doesn't rain in SL, so we have to complain about other things
[17:05]	reddot99 Republic:	exactly what moon said
[17:05]	Falcon Linden:	hehe
[17:06]	Andrew Linden:	Thanks for showing up everyone. I'm going to depart.
[17:06]	Ardy Lay:	Bye Andrew
[17:06]	Psi Merlin:	Thanks
[17:06]	Moon Metty:	great meeting, Andrew, Falcon
[17:06]	Pauline Darkfury:	here's a fun video of someone turning around 8000-9000 prims physical simultaneously which you might enjoy: http://www.youtube.com/watch?v=Gdtkl732bik
[17:06]	Moon Metty:	happy weekend :)
[17:06]	reddot99 Republic:	also, the mouselook thing could be probably fixed by a viewer mod, i bet,
[17:06]	Falcon Linden:	cheers folks
[17:06]	Rex Cronon:	tc andrew
[17:06]	Pauline Darkfury:	thanks Andrew, Falcon :)
[17:06]	Rex Cronon:	tc falcon
[17:06]	Pauline Darkfury:	have a good weekend :)
[17:07]	Patnad Babii:	goodbye everyone have a nice weekend
[17:07]	Rex Cronon:	tc everybody
[17:07]	Rex Cronon:	and have fun:)

Generated with SLog Wikifier