User:Babbage Linden/Office Hours/2009 04 29
< User:Babbage Linden/Office Hours
Jump to navigation
Jump to search
Revision as of 04:15, 6 May 2009 by Nock Forager (talk | contribs) (New page: Transcript of Babbage Linden's office hours: '''Topics''' * http-texture increase texture download speed at 100% rate: LINK {| <span id="chat1"><...)
Transcript of Babbage Linden's office hours:
Topics * http-texture increase texture download speed at 100% rate: LINK
[3:22] | Babbage Linden: | hi everyone
|
[3:22] | Laurent Bechir: | greeting, Babbage
|
[3:22] | Babbage Linden: | sorry i'm late
|
[3:22] | Nock Forager: | Hi Babbage.
|
[3:22] | WolfPup Lowenhar: | mornin babage
|
[3:22] | Kephra Nurmi: | moin babbage
|
[3:22] | Robertus Bobak: | hello Babbage
|
[3:22] | Thoys Pan: | G'Day Babbage
|
[3:23] | Babbage Linden: | not a lot of script news from us this week
|
[3:23] | Babbage Linden: | we're currently working on http textures in brighton
|
[3:23] | Babbage Linden: | which is looking good
|
[3:23] | Babbage Linden: | should improve texture download by 100% on average
|
[3:23] | Imaze Rhiano: | yay
|
[3:23] | koen3 Bing: | nice
|
[3:24] | WolfPup Lowenhar: | that will mean that soon we will be able to show actual webpages in SL its sefl
|
[3:24] | Laurent Bechir: | nice , yes :)
|
[3:24] | Babbage Linden: | should halve the time for texture downloads on average
|
[3:24] | Babbage Linden: | and potentially open the door to hosting textures externally
|
[3:24] | Babbage Linden: | eventually
|
[3:25] | Kephra Nurmi: | this would require a new client for everyone?
|
[3:25] | Babbage Linden: | it also moves a lot of data off of our creaky udp message system
|
[3:25] | Thoys Pan: | thats good, will it still send out textures with discard level? , or must we wait till the complete texture is loaded before we see it?
|
[3:25] | Kephra Nurmi: | or would it also run with old clients?
|
[3:25] | Babbage Linden: | it will need a new client to take advantage of it
|
[3:25] | Babbage Linden: | so we'll need to support the old udp texture download for a long time
|
[3:26] | Cristopher Lefavre: | Babbage, will this imply that textures are loaded over TCP/80 instead of a high UDP port?
|
[3:26] | Babbage Linden: | until all supported viewers are downloading textures over http
|
[3:26] | Babbage Linden: | not sure about the ports cristopher
|
[3:26] | WolfPup Lowenhar: | your saying that we could eventualy have it so that we would be able to store texttures out of SL then simply have that texture available inworld ?
|
[3:26] | Babbage Linden: | currently we're using the caps system to send the textures in cleartext
|
[3:26] | Babbage Linden: | but that may change too
|
[3:27] | Babbage Linden: | as using caps means we have a url for every <viewer, texture> pair
|
[3:27] | Babbage Linden: | where ideally we only want 1 url per texture
|
[3:27] | Laurent Bechir: | sorry if I don't understand completly :) does it means that objects could load their textures from a personnal server ?
|
[3:27] | Babbage Linden: | actually, 1 url per <sim, texture> pair
|
[3:27] | Babbage Linden: | laurent, potentially, yes
|
[3:27] | WolfPup Lowenhar: | with textures comeing in in chat that might be whay some textures fial to load at all
|
[3:27] | Babbage Linden: | you would want to wait until all supported viewers to http texture download though
|
[3:28] | Babbage Linden: | otherwise people with older viewers wouldn't be able to see your texture
|
[3:28] | Qie Niangao: | does the http-texture data pass through SL's network (I hope) or directly from the http source to the client (which would mean lots of connections--and security concerns)?
|
[3:28] | Babbage Linden: | qie, the former
|
[3:28] | Qie Niangao: | thanks. :)
|
[3:28] | Babbage Linden: | laurent was asking about the latter
|
[3:28] | Kephra Nurmi: | so the current idea is: client request capabilities, and receives a caps url for the textures?
|
[3:28] | Babbage Linden: | (which will come later ;-)
|
[3:29] | Babbage Linden: | kephra, yes
|
[3:29] | Kephra Nurmi: | *ah* nice
|
[3:29] | Babbage Linden: | this improves the simulator performance too
|
[3:29] | Kephra Nurmi: | so non clients wont have the caps and can not just simply download the textur
|
[3:29] | Babbage Linden: | currently the sim has to read textures from disk
|
[3:29] | Babbage Linden: | then chop them up and send them via udp
|
[3:29] | Babbage Linden: | in future it will just give the viewer a cap
|
[3:30] | Babbage Linden: | and the texture transport will be off loaded to an apache proxy on the sim host
|
[3:30] | Babbage Linden: | so it takes load off the sim too
|
[3:30] | Kephra Nurmi just thinks about the textures from 3rd party servers
| |
[3:31] | Kephra Nurmi: | would it be possible to run a capability server on own system, and to use this for other things than textures?
|
[3:31] | Imaze Rhiano: | isn't there plan to change texture baking too?
|
[3:31] | Babbage Linden: | texture baking is where most of the compexity comes with this
|
[3:32] | Babbage Linden: | i'm currently talking to phoenix about whether we want to change baked textures
|
[3:32] | Babbage Linden: | or support the current system with the http transport
|
[3:32] | Thoys Pan: | do textures also have some kind of checksum?
|
[3:33] | Thoys Pan: | so you could see if you really downloaded it right?
|
[3:33] | Babbage Linden: | well, TCP and HTTP take care of that
|
[3:33] | WolfPup Lowenhar: | i would think they would have to for them to load right as it is now
|
[3:33] | Babbage Linden: | and the UDP transport takes care of it too
|
[3:34] | Babbage Linden: | so, that's what we're up to at the moment
|
[3:35] | Babbage Linden: | anything you'd like to talk about?
|
[3:35] | WolfPup Lowenhar: | i was wonderin how the scrip memory project is going
|
[3:35] | Imaze Rhiano: | any data about script memory limits yet?
|
[3:35] | Indeterminate Schism: | Any more thoughts on the proposed memory-management tools?
|
[3:35] | Babbage Linden: | still waiting for the numbers from xan
|
[3:35] | Babbage Linden: | should be here this week
|
[3:35] | Babbage Linden: | i reminded him last week for them
|
[3:36] | Babbage Linden: | the plan is still to look at the numbers, work out what we can support for script limits
|
[3:36] | Babbage Linden: | then publish those proposed numbers
|
[3:36] | Babbage Linden: | then add tools for discovering where memory is being used
|
[3:37] | Babbage Linden: | and finally add the enforcement code after people have had time to tidy up
|
[3:37] | Babbage Linden: | enforcement is unlikely to happen before the end of the year
|
[3:37] | WolfPup Lowenhar: | it would be good to be able to see how much memory a script actualy uses
|
[3:37] | Thoys Pan: | also , babbage, would the http way make it free to upload/set textures?
|
[3:37] | Qie Niangao: | yeah, we could use those tools even if we didn't know the target values.
|
[3:38] | Babbage Linden: | yes, there are still discussions to have around the actual memory accounting
|
[3:38] | Babbage Linden: | we don't want to support 2 VMs indefinitely
|
[3:38] | Babbage Linden: | so need to make sure we don't accidentally incentivise using LSO scripts
|
[3:39] | Babbage Linden: | which is tricky as in some cases, LSO scripts can be the more memory efficient option
|
[3:39] | Babbage Linden: | (although not often)
|
[3:39] | Babbage Linden: | the problem gets worse when we go to 64 bit, which makes mono scripts use more memory, but not LSO scripts
|
[3:40] | Babbage Linden: | hopefully we can set the memory limits at levels where this is not an issue for 99% of people though
|
[3:40] | Babbage Linden: | and only really stops really abusive misuse of scripts
|
[3:40] | Babbage Linden: | for griefing
|
[3:40] | Cristopher Lefavre: | Any news on the htp server implementation, Babbage?
|
[3:40] | Babbage Linden: | we'll see
|
[3:40] | Xaria Mistwallow: | I have a 64bit PC
|
[3:40] | WolfPup Lowenhar: | < is on 32 bit systems
|
[3:40] | Babbage Linden: | 64 bit PCs don't affect scripts xaria
|
[3:40] | Xaria Mistwallow: | i do see drastical decrease in mono scripts, not sure why it doesnt effect me then
|
[3:41] | Kephra Nurmi: | *hm* does the simulator run in a 32bit chroot environment?
|
[3:41] | Thoys Pan: | with 64 bit, would floats be more precise?
|
[3:41] | Babbage Linden: | memory use for scripts is only affected by the simulator binary
|
[3:41] | Babbage Linden: | at the moment simulators are 32 bit
|
[3:42] | Babbage Linden: | running on 64 bit debian machines via emulation
|
[3:42] | Babbage Linden: | we'd like to move them to 64 bit
|
[3:42] | Xaria Mistwallow: | im on vista unfortunatly
|
[3:42] | Babbage Linden: | but that affects Mono memory usage
|
[3:42] | WolfPup Lowenhar: | when the simulators goto 64 bit that should me the perform better
|
[3:42] | Imaze Rhiano: | how about instead of memory limit - you would calculate abstract "performance value" for script - that would combined from different parameters - like script's memory usage, script's VM, what functions it calls, etc...
|
[3:42] | WolfPup Lowenhar: | mean*
|
[3:42] | Babbage Linden: | imaze, yes we could go down that route
|
[3:42] | Babbage Linden worries that it would be more confusing though
| |
[3:43] | Babbage Linden thinks of the mystical dwell calculation and shudders
| |
[3:43] | Thoys Pan: | ah, so it has to do with the os, and what if you use debian 32 systems for now?
|
[3:43] | WolfPup Lowenhar: | vist already dose that in a scale of 0 to 10
|
[3:43] | WolfPup Lowenhar: | vista*
|
[3:43] | Babbage Linden: | we have been using 64 bit debian machines for a long time
|
[3:43] | Babbage Linden: | but running 32 bit simulators on them
|
[3:43] | Xaria Mistwallow: | i know, i got probably the worst combination there is
|
[3:44] | Babbage Linden: | we'd like to make everything 64 bit
|
[3:44] | Xaria Mistwallow: | unfortunatly i found out AFTER i got this =(
|
[3:44] | Thoys Pan: | or you cant get above 3036 mb ram memorry when you have 32 bit right?
|
[3:45] | WolfPup Lowenhar: | actualy max on a 32bit system is 4GB physical but the top GB is used by the system
|
[3:45] | Kephra Nurmi: | applications cant get more then 2gb, and server cant have more than 3gb without bad tricks, on a 32bit system
|
[3:45] | Xaria Mistwallow: | Babbage: does that mean 64bit will become more stable aswell? I crash all the time now without any warning or knowing why, in the end it comes down to 64bit with SL doesnt work that well together
|
[3:45] | Xaria Mistwallow: | I do have to say that alot of games have this problem unfortunatly
|
[3:45] | Qie Niangao: | (RL calls. Thanks Babbage. Have fun all.)
|
[3:45] | Kephra Nurmi: | xaria your problem is client side ... because of 32bit/64bit opengl emulation
|
[3:45] | Nock Forager: | (me using Win7 64bit and works fine. Rarely carashed, Xaria).
|
[3:46] | Babbage Linden: | xaria, the simulator move to 64 bit won't affect the client stability
|
[3:46] | Babbage Linden: | the good news is that both the viewer and simulator stability have improved markedly recently
|
[3:46] | Babbage Linden: | the simulator crash rate went down ~50% after the last release
|
[3:46] | Babbage Linden: | which was great news
|
[3:46] | Xaria Mistwallow: | too bad, I just hope this could be fixed somehow, it gets really annoying when you crash all the times just becouse of that issue
|
[3:46] | Thoys Pan: | yes
|
[3:46] | WolfPup Lowenhar: | that is good
|
[3:47] | Kephra Nurmi: | well ... simulator patches could be bad for clients ... e.g. the saturday change on search cause me to crash reliable when ever i clicked (search)
|
[3:47] | Thoys Pan: | or theres a bug in the crash reporter since than lol
|
[3:47] | WolfPup Lowenhar: | < is helping test the new viewer right now
|
[3:47] | Xaria Mistwallow: | its not the sims im worried about, client with 64bit is :p
|
[3:47] | Xaria Mistwallow: | I did not have any problems on the OnRez viewer tho
|
[3:47] | Babbage Linden: | xaria, yes i understand, unfortunately i'm not the linden to ask about the viewer
|
[3:48] | Babbage Linden: | i think q has office hours and is very involved with the viewer work
|
[3:48] | Xaria Mistwallow: | 8am for him I noticed, yeah. ill go ask him about this same thing
|
[3:48] | Babbage Linden: | there's a bear for those that wanted one
|
[3:49] | Babbage Linden: | for sale for 0L$
|
[3:49] | Xaria Mistwallow: | ♥ Thank Youuuuuuuuuu!! ♥
|
[3:49] | Xaria Mistwallow: | :)
|
[3:49] | Xaria Mistwallow: | or take copy :)
|
[3:49] | WolfPup Lowenhar: | ♥ Thank Youuuuuuuuuu!! ♥
|
[3:49] | Babbage Linden: | any more questions or topics?
|
[3:49] | Xaria Mistwallow: | not so far i know of :p
|
[3:49] | Kephra Nurmi has progress on his 'using mediawiki as a cloud database for secondlife' project ;)
| |
[3:49] | Cristopher Lefavre: | The http server project?
|
[3:50] | Xaria Mistwallow: | I do not know much about scripting etc. I do try to learn tho
|
[3:50] | Nock Forager: | Watched your video on lang.net. The nubmers are very interested.
|
[3:50] | Babbage Linden: | glad you checked it out nock
|
[3:51] | Nock Forager: | So many scripts running in the Grid!
|
[3:51] | Babbage Linden: | i wrote up the conference from an sl perspective here: http://jimpurbrick.com/2009/04/27/langnet-3-years/
|
[3:51] | Babbage Linden: | yes, it's amazing: 1000 scripts per logged in avatar
|
[3:51] | koen3 Bing: | woow
|
[3:52] | Imaze Rhiano: | many 255 prim hairs with scripts :P
|
[3:52] | WolfPup Lowenhar: | dang
|
[3:52] | Xaria Mistwallow: | lol, i hardly have any. atleast not that i can mod =)
|
[3:52] | Babbage Linden wonders how many will be cleaned up with script limits
| |
[3:52] | Nock Forager: | Resize script should be removed after size fixed. ;-)
|
[3:52] | Babbage Linden: | part of the reason for so many scripts is that there's no limit
|
[3:53] | Babbage Linden: | i imagine some percentage are not needed
|
[3:53] | Babbage Linden: | and just haven't been cleaned up
|
[3:53] | Babbage Linden imagines the big SL spring clean
| |
[3:53] | WolfPup Lowenhar: | i wonder if megas actualy have hidden scripts in them
|
[3:53] | Xaria Mistwallow: | <<<hihi>>>
|
[3:54] | Kephra Nurmi: | megas are nearly normal prims, from scripting point of view
|
[3:54] | Indeterminate Schism: | Having a llSetLinkPrimitiveParams() where we could target a range of prims would be good, so we didn't need scripts in all the child-prims when the changes need to be synchronised. I think you've discussed that though
|
[3:54] | Babbage Linden: | yes, it would often be handy for scripts to be able to operate on lots of prims
|
[3:55] | Babbage Linden: | to avoid lots of scripts everywhere
|
[3:55] | WolfPup Lowenhar: | true but then you could also end up with one prim in a linkset that use a massive amount of memory
|
[3:56] | Babbage Linden: | if scripts can request large amounts of memory that's not a problem wolfpup :-D
|
[3:56] | Imaze Rhiano: | Doesn't MONO have now CIL verifier? Wouldn't that make possible C#?
|
[3:57] | Babbage Linden: | it has a bytecode verifier
|
[3:57] | Babbage Linden: | i'm not sure their metadata verifier is complete
|
[3:57] | Kephra Nurmi: | *oups* is this server special ... just got an 'can not upload script' message while editing a script in inventory
|
[3:57] | WolfPup Lowenhar: | true but someone might find a way to use that for greifing as well by making a script that kept making linked prims and having all of them setting with divverent prams
|
[3:57] | Babbage Linden: | which also needs to be done before we can use it
|
[3:58] | Indeterminate Schism: | Well you *can* do that now Wolf, you just can't change 'only these' efficiently
|
[3:58] | Babbage Linden: | ok, i need to head off
|
[3:59] | Babbage Linden: | thanks for coming everyone
|
[3:59] | Indeterminate Schism: | Thank you Babbage
|
[3:59] | Laurent Bechir: | thank you Babbage and bye
|
[3:59] | Babbage Linden: | hopefully see you all next week
|
[3:59] | Imaze Rhiano: | thank you
|
[3:59] | WolfPup Lowenhar: | you tc babage and thanks for having these office hours
|
[3:59] | koen3 Bing: | :)
|
[3:59] | Cristopher Lefavre: | Thanks, bye! |