User:Zero Linden/Office Hours/2007 Feb 22

From Second Life Wiki
< User:Zero Linden/Office Hours
Revision as of 20:23, 28 March 2007 by Zero Linden (talk | contribs) (better formatting)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Transcript of Zero Linden's office hours:

[7:33] Zha Ewry: Morning, Zero!
[7:34] Zero Linden: sorry all
[7:34] Zero Linden: running a bit late in SF
[7:34] Vitis Obviate: np - Morining Zero
[7:34] Zero Linden: hmmmm.... silly login packets unseating me
[7:34] Zha Ewry: Heh, much earlier for you, on the left coast, glad your're making it here
[7:35] Coffee Mug whispers: Ahh! Fresh Hot Coffee
[7:35] Zero Linden: well, I'm normally in the SF office at this time on Thursdays
[7:35] Zero Linden: which is why I set this
[7:35] Zero Linden: Let see, anyone here from Eurpoe?
[7:35] Vitis Obviate: works well for me
[7:35] Zero Linden: *Europe
[7:36] Zha Ewry sighs. Morning people.
[7:36] Coffee Mug whispers: Ahh! Fresh Hot Coffee
[7:36] Vitis Obviate: ty for posting the transcripts
[7:36] Zero Linden: Welome
[7:36] Zha Ewry: Yes, very much indeed. Good reading there
[7:37] Zero Linden: So - I'll be posting this one as well
[7:37] Zero Linden: (obligatory warning!)
[7:38] Zero Linden: Okay then - let's see were all of you at Tuesday's hours?
[7:38] Tree Kyomoon: yes
[7:38] Zha Ewry: yes
[7:38] Vitis Obviate missed it...but is reading the transcript now
[7:38] Zero Linden: Ah - okay - just wondering if I should continue twice a week, or make it once a week, but for a longer time...
[7:38] Zha Ewry was busy perstering you about first lok, and AV
[7:39] Vitis Obviate nods in suport of Zah's concern about 2048 text limit on httpRequest
[7:39] Zha Ewry: I prefer more often, but I'm a glutton for punishment
[7:39] Zero Linden: So, let's take that up
[7:39] Tree Kyomoon: me too
[7:39] Zha Ewry: (I've added you to my organizer through YE07)
[7:40] Zero Linden: Yeah - The two of you are gunning for "Zero Groupie" status!
[7:40] Zero Linden: Okay, so HTTP request limit
[7:40] Zero Linden: now, how much bigger do you think you could handle?
[7:40] Zha Ewry: Hmm. YOu'd mentioned it ties to the LSL limits
[7:40] Zero Linden: Well, the whole script has only 16k
[7:40] Zha Ewry: and I was thiinking about that.
[7:40] Zero Linden: and since strings are immutable
[7:41] Zero Linden: doing any operation on a string sort of requires about 2x available
[7:41] Zero Linden: (not quite, but to be conservative)
[7:41] Zha Ewry: There are two issues, in here One is total data, and the other is the server side
[7:41] Zha Ewry: Sure, be conservative. Otherwise we crash scripts
[7:41] Zha Ewry: If we had clean chaining, and possibly non text mimetypes,
[7:41] Tree Kyomoon: would be nice to have a slower client peer to peer option that isnt limited
[7:42] Zha Ewry: the 2048 is much less of a problem
[7:42] Tree Kyomoon: so we could pass datdirectly to the client
[7:42] Zero Linden: That's why we put in the limit - I didn't want an external script gone awry (Apache being so fond of generating HTML error responses) to crash your inworld script
[7:42] Zha Ewry: I'm mostly trying to move larger chunks of data into the world, without lots and lots of odd hackery
[7:42] Zero Linden: Zha / Tree - where is the destination of that data?
[7:42] Tree Kyomoon: in world
[7:43] Vitis Obviate: offWorld
[7:43] Zero Linden: Seems for in-world, most other communications channels are limited to 256 characters anyway
[7:43] Tree Kyomoon: I need to get 32-64 sometimes 100k XML fiiles into the world
[7:43] Zha Ewry: Depends. We've serveral cases. We're building molecular models with 5000+ prims
[7:43] Zero Linden: Tree - my gosh - what do you with a 100k XML file in LSL?
[7:43] Tree Kyomoon: build learning modules
[7:43] Zha Ewry: and right now, we end up using cvs files in notecards
[7:43] Tree Kyomoon: dynamically generated simulations
[7:43] Zha Ewry: csv, boring comma seperated things
[7:44] Zero Linden: just wondering how one parses XML reasonably in LSL
[7:44] Vitis Obviate: tedious hackary
[7:44] Zha Ewry: Which is painful, and ends up hand crafted
[7:44] Tree Kyomoon: its not so bad
[7:44] Tree Kyomoon: i can parse as I go
[7:44] Zero Linden: Our expectation was that people would put the heavy, XML parsing on the server, and only deliver very customized, preformatted data to scripts
[7:44] Zha Ewry: Yes, me too. And I don't need messy XML, mostly.
[7:45] Zha Ewry: Yes, its how we do it. Far more that we have large volume than complex format.
[7:45] Zero Linden: Since, to do XML, (whcih I'm a big fan of), requires that you really "do it right" to get all the benefit
[7:45] Vitis Obviate: We are running xslt on the server to convert our xml into whatever we need here
[7:45] Zha Ewry does not want to do schema validation in LSL, this week
[7:45] Zero Linden: Now, when passing data onto other prims, again, you can't push it even in 2k chunks.... so it seemed a reasonable match
[7:45] Zero Linden: Welcome Gus -
[7:45] Gus Plisskin: thx Zero
[7:46] Zha Ewry: Hmm. How much better does all that get with Mono?
[7:46] Zero Linden: The memory constraints won't get better at all
[7:46] Zha Ewry: I know some of it can't go away, or we'll turn all the sims into compute farms
[7:46] Tree Kyomoon: Im just trying to not have to rebuild our existing API
[7:46] Zero Linden: Since our model is still that there can be 1000s of script running in a sim at a time, we still have to limit them in memory usage
[7:46] Zha Ewry: Hmm.
[7:46] Zha Ewry: As an estate owner, I'd be happy to have a knob that let me trade that off
[7:47] Tree Kyomoon: me too
[7:47] Zha Ewry: But, in general, sure
[7:47] Tree Kyomoon: welcome gus!
[7:47] Zero Linden: Tree - I think, even with mono, we aren't expecting there to be that much compute power in world - it is far better to do the bulk of the application logic on the server and use SL as the display
[7:47] Zha Ewry: (with 100 avatars, and 5 scripted things a bit.. yes, scripts at up)
[7:47] Zha Ewry: add
[7:47] Gus Plisskin: hi Tree
[7:47] Tree Kyomoon: I think having a more robust LSL will help make that possible
[7:47] Zero Linden: LIke this region has 977 scripts -
[7:48] Zero Linden: and that is low
[7:48] Tree Kyomoon: righ now its very clumsy and laborious, needs arrays and proper objects
[7:48] Zha Ewry: The other sorts of things which sneak in, are textures and sonds, and such. Which are a real hassle to get into the world progamtically
[7:48] Zero Linden: Since each region has 512M of RAM to work with, we need to limit usage of scripts
[7:48] Zha Ewry: Woah. That low?
[7:48] Tree Kyomoon: 512???
[7:48] Zero Linden: Ah, well, those are another issue altogether
[7:49] Tree Kyomoon: Can I pay extra for more RAM?
[7:49] Vitis Obviate sips his coffee and contemplates that
[7:49] Zero Linden: Yes - that is what we've allocated per region: 512M and one CPU core
[7:49] Zha Ewry: (Which will sneak into the use cases you asked for last week)
[7:49] Zero Linden: no, you can't because we need the HW to be fairly generic
[7:49] Zha Ewry: Hmm. so about 2-4GB for each server you use
[7:49] Zero Linden: See, your region doesn't run on a specific computer - just a computer with a certain set of computation resources
[7:49] Zha Ewry: I can see that, actualyl
[7:50] Tree Kyomoon: Im amazed they work so well on 512, well done folks!
[7:50] Zero Linden: So we need to be able to move regions to other computers
[7:50] Zha Ewry: Kernel, core stuff, then 1/2Gb for the region. Snug fit, yes.
[7:50] Zero Linden: You see, with 2k physical machines - HW failure is a regular occurance
[7:50] Zha Ewry: Yes, statistical even
[7:51] Zero Linden: Zha - exactly - and until recently, that memory limit was kind of an upper bound on stock hardware
[7:51] Zero Linden: Actually, I understand Google has just released an interesting paper on disk failures.... meaning to read it
[7:52] Zha Ewry: One thing, to mplify Tree's point, is that if you had class X servers, with more memory per sim, or fewer sim per server, some of us would pony up for them
[7:52] Zha Ewry: Although, I get that from a system amangement point of view, every asymetry you add hurts
[7:52] Tree Kyomoon: I had planned on spending a lot more for a 3d simulation tool from Sony PS or MS
[7:53] Zero Linden: I'm sure! But, on the other hand, every additional class of server means less liquidity in our server pool, and more "different configuations" to manage
[7:53] Zha Ewry: Yes, my comment, exacly. Each new class, is system management pain
[7:53] Zero Linden: So, at present we take a "one size fits all" approach - though we keep upping the size
[7:53] Tree Kyomoon: but some extra cash to hire more help :)
[7:54] Zero Linden: Well- you'd be surprized at how hard it is to hire good sys. admin. help!
[7:54] Zha Ewry chuckles. Sure, tree, but it's expensive help. More testing, more fuss when loading racks, etc
[7:54] Gus Plisskin: stardardization maybe more important than a little more cash
[7:54] Tree Kyomoon: Oh, I know...Im hiring mode myself, very challenging
[7:54] Gus Plisskin: *standardization
[7:54] Zero Linden: On the other hand, architectural changes in the future should allow this kind of thing with greater ease
[7:54] Tree Kyomoon: yes, I agree Gus, point taken
[7:55] Zero Linden: We are developing management software for this sort of thing now because we
[7:55] Zero Linden: are going to want to be able run regions on different versions of the simulator
[7:55] Zero Linden: since the simulator has a host of support processes (dataservers, apache config, backbone, squid config....)
[7:56] Zha Ewry: ok, so, assuming we're stuck on memotry, the thing that strikes me, is if we could get good (standard driven) chaining, we could probably
[7:56] Zero Linden: a whole machine has to be at a given release level at a time
[7:56] Zero Linden: so we need to manage pools of machines at different release levels
[7:56] Zha Ewry: get a much cleaner serve/LSL side coupling
[7:56] Zero Linden: Tell me more - I'm not sure what you mean by chaining? keep-alive support?
[7:57] Zero Linden: Or actually doing things like Jabber does and
[7:57] Zero Linden: syncing the request/response streams
[7:57] Zha Ewry: There's so stuff in the html spec space 9(I'll drop you the links, later today, on a notecard) which lets you manage a series of get requests, in a chani
[7:58] Zero Linden: You can e-mail me directly if you want:
[7:58] Zha Ewry: None of that works on the LL servers, at the moment. Each request is a seperate get, with no chaining from the serverrs side, so we have to do chaning.
[7:58] Zha Ewry: Heh.
[7:58] Zha Ewry: Sure
[7:58] Zha Ewry: You afre in danger of getting use cases with pretty pictures, one day
[7:58] Zero Linden: Zha - send me the links - This isn't in HTTP/1.1 yet, right?
[7:58] Vitis Obviate smiles
[7:58] Zha Ewry: I think some of it is
[7:58] Zero Linden claps
[7:58] Zha Ewry: I'll check
[7:59] Zero Linden: I'd *love* use cases with pretty pictures
[7:59] Tree Kyomoon: question about chatting in open air..are we on a specific channel that is region tied, and can this chat be made accessible via external chat clients ?
[7:59] Zero Linden is an XP / Agile Methods kinda guy
[8:00] Zero Linden: Chat doesn't go through our IM system at all - so it isn't like IM systems
[8:00] Zero Linden: In this case, it is region tied - but also spacially located - and the sims cooperate at borders so it feels seemless
[8:00] Zha Ewry: Hmm. I've build a simple hack that pumps anything an object heres onto our internal IM client So not that hard
[8:00] Zero Linden: In other words, you hear stuff over the sim borders
[8:01] Zero Linden: We would be unlikely to ever tie chat into an externally accessible chat systm
[8:01] Zero Linden: IM, on the other hand,
[8:01] Zero Linden: is ripe for external connection
[8:01] Zero Linden: Zha yes, you could -
[8:01] Zha Ewry: Sure, the conttact on chat, is implictly, only peple I can see here me
[8:01] Gus Plisskin: FlipperPA did an IRC to SL IM system
[8:01] Zero Linden: but I think you having a repeater object is sort of like an explicit dump -
[8:01] Tree Kyomoon: seems like all youd need is a bunch of litle listneing / broadcasting objects
[8:02] Zero Linden: whereas us making all sim chat externally monitorable sort of changes the dynamic
[8:02] Zero Linden: Such objects exist -
[8:02] Tree Kyomoon: Im only thinking of attending meetings like this via blackberry/ or other devices if possible
[8:02] Zero Linden: often for group events that span sims
[8:02] Tree Kyomoon: that dont run the full SL client
[8:02] Zero Linden: That is a curious issue
[8:02] Zero Linden: Clearly thinner clients would be nice
[8:02] Gus Plisskin: Some groups get everyone's permission, then broadcast chat
[8:03] Zero Linden: but there is a trade off - part of the value of SL is your presence in the world
[8:03] Zero Linden: perhaps you could attend via blackberry, and you AV would appear here, in sort of "capsule" form or some such
[8:03] Zha Ewry: html chaning use case, in the pixels. (Sifu's been doing html chaining and maknig much mutterin)
[8:03] Tree Kyomoon: true, but you might still be able to be "present" to others in the fat client
[8:04] Zero Linden: But of course, you don't need all of SL just to have a group chat....
[8:04] Sifu Moraga: Hi all
[8:04] Tree Kyomoon: like a hologram
[8:04] Tree Kyomoon: yes, it does seem like an awfully massive process to text chat
[8:04] Sifu Moraga: Yes we've been struggling with the limits of llHTTPRequest
[8:04] Zha Ewry: We've been discussing, on and off, how to getbigger data chunks in world
[8:04] Zero Linden: Ha ha - like Princess Leia in the first Star Wars when broadcast by R2D2
[8:04] Tree Kyomoon: sorry for the tangent there
[8:04] Zero Linden: you could appear 1/2 size and translucent
[8:05] Tree Kyomoon: lol
[8:05] Vitis Obviate: could we perhaps add xml- RPC to this discussion?
[8:05] Zero Linden: Yes -lets
[8:05] Zha Ewry: Indeed, the other half of the problem :-)
[8:05] Vitis Obviate: :)
[8:05] Zero Linden: So, incoming XML-RPC is very difficult for us to maintain
[8:05] Zero Linden: What would need to be able to geet rid of it?
[8:06] Vitis Obviate: get rid of it? was hoping to make it better :)
[8:06] Zero Linden: Well - here's the thing- as current framed
[8:07] Vitis Obviate: You meant get rid of perpectually opened listener?
[8:07] Zero Linden: it creates a non-distributable load on the system - an XML-RPC channel is essentially a global identifier for an objec t that can appear anywhere in world
[8:07] Zero Linden: This is difficult to maintain
[8:07] Zero Linden: So, what if, for example
[8:07] Sifu Moraga: afk
[8:08] Zero Linden: Instead of XML-RPC, you had instead the ability for a script to be an HTTP listener
[8:08] Zero Linden: you'd get the path and the body of a request and could reply to it
[8:08] Zero Linden: but here'd be the catch:
[8:08] Zero Linden: the URL would be tied to the region
[8:08] Zero Linden: if you're object moved about - it'd end up with a new URL
[8:08] Zha Ewry: Hmm. The cost, being that you have to keep track of the object wherever it ends up in world, as the sisms and avatrs move?
[8:08] Zero Linden: Zha - exactly
[8:08] Zha Ewry: (with the current scheme)
[8:08] lyndell Aleixandre: Yay!
[8:08] lyndell Aleixandre: webhosting within SL lol...
[8:08] Vitis Obviate: well...if we could initiate an xml-rpc request to a trusted IP outworld...and got a response back in finite time..that would limit the amount of time such an object would need to be manageed
[8:09] Tree Kyomoon: so youd have to own the region to use it
[8:09] Zero Linden: Zha - and we don't really know when an object is "done" with it's channel
[8:09] Zha Ewry: SO, now you do it, and if we move to an hhtp Listenr, we do it.
[8:09] Zero Linden: Tree- no, not at all
[8:09] Zha Ewry: right. No way to garbagae collect
[8:09] Zero Linden: but the URL might be like:
[8:10] Zero Linden: where your script would get "/foo/bar", and the body
[8:10] Zha Ewry: So, you shed the tracking and load balance trickery , and we have to re-validate our inbound connections more often
[8:10] Zero Linden: exactly
[8:10] Zero Linden: but it seems with outbound HTTP, this should be easy to re-register your objects
[8:10] Tree Kyomoon: thats fine with me, but how is it not limited to entire region owners?
[8:10] Zero Linden: it basically shifts the burden of tracking from us globally to each of you, per application
[8:11] Zha Ewry: If, especially, if, it was neatly tied together with some upgrades to the http stuff, I can see it as a net win.
[8:11] Zero Linden: Because that URL could address any object in the region - the object needn't be owne by the region owner - or even the parcel owner
[8:11] lyndell Aleixandre: that sounds good
[8:11] Zero Linden: we let object land permissions take care of that
[8:12] Zha Ewry: One could imagine actually keeping the basic two way pipe as close to normal web usage as possible, and less assymettic
[8:12] Tree Kyomoon: but who determines the URL that is listened to in the first place?
[8:12] Zero Linden: I was thinking of a call, essentially, llGetMyURL()
[8:12] Zha Ewry: Well, the listening object presumbably gets it's inbound URL as part of the create process
[8:12] Zha Ewry has to stop chanelling Zero.
[8:12] Zero Linden: no no - keep at it!
[8:13] Zha Ewry singgers
[8:13] Zero Linden: This way, it could possibly work for attachements
[8:13] Tree Kyomoon: so the URL isnt region determined, its object determined
[8:13] Zero Linden: whose URL would need to follow the agent, not the region
[8:13] Zha Ewry: Hmm. Yes.
[8:13] Tree Kyomoon: ok sorry I missed that somehow
[8:14] Zero Linden: (see, I've given this a little thought...!)
[8:14] Zha Ewry: Being able to not have to re-register on every sim boundary change would be nice.
[8:14] Vitis Obviate: yes
[8:14] Zero Linden: Well, Tree, if you have an attachement that wants to be a server too - that URL can't be tied to the region or it would change every time you region cross
[8:14] Zha Ewry: 90% of my use cases, are for static objects, which sit in places
[8:14] Tree Kyomoon: it would but it could be reregistered silently without the owners involvement right?
[8:14] Sifu Moraga: (back again)
[8:14] Zero Linden: Exactly - so that is why I want to remove the very expensive overhead of XML-RPC's global location service
[8:14] Zha Ewry: but, about 10% want to be on huds, or object worn by avatars.
[8:15] Vitis Obviate: I have an important project that using higly mobile hud-based stuff
[8:15] Zero Linden: Right - and 0.01% are objects that are autonomously mobile
[8:15] Tree Kyomoon: this would be awesome Zero
[8:15] Zero Linden: Those would have to re-register
[8:15] Zero Linden: Now -
[8:15] Tree Kyomoon: a definite step closer to the browser on a prim
[8:15] Zero Linden: On the other hand, Donovan Linden
[8:15] Zero Linden: has demonstrated the "long poll" technique in world
[8:15] Vitis Obviate: the web services infrastructure person?
[8:15] Zero Linden: This is where, if you want a script to be able to receive requests
[8:16] Zero Linden: you make instead, make a llHTTPRequest out to a server... which just "hangs", until it has somethign to send you
[8:16] Zero Linden: When it does, it finally responds, and your script gets the response
[8:16] Zha Ewry: Hmm.
[8:16] Zero Linden: If the whole thing times out - you just make the request again
[8:16] Sifu Moraga: A blocking request
[8:16] Zha Ewry: The REST commuinityu is screaming, already
[8:17] Vitis Obviate: like TEST?
[8:17] Vitis Obviate: REST
[8:17] Zero Linden: Well - remember that the script doesn't block...
[8:17] Zero Linden: Yes, it isn't very REST
[8:17] Zha Ewry: no, but the long poll is very non idempotent
[8:17] Zero Linden: But, it is the basis of many an AJAX system
[8:17] Vitis Obviate thinks about AJAX
[8:17] Vitis Obviate: yikes
[8:17] Zero Linden: And Donovan has shown this to work in world - he's got a web page you go to and type in
[8:17] Zha Ewry has been on the phone wwith web 2.0 and REST people at the same tmie. Not pretty
[8:18] Zero Linden: and an object in world speaks what you've said
[8:18] Zero Linden: what's nice is this is fairly low drain on the system
[8:18] Zha Ewry: But yes, it should work just fine
[8:18] Zero Linden: Your viewer is in fact doing this technique right now!
[8:18] Zero Linden: "See, you're soaking in it."
[8:19] Vitis Obviate: can't see the water for the fish we are Zero
[8:19] Zha Ewry: Heh. And that's also why the client ghosts from time to time, :-)
[8:19] Zero Linden: (anyone old enough to get the reference?)
[8:19] Tree Kyomoon: lol
[8:19] Tree Kyomoon: madge
[8:19] Zero Linden: Bingo - give that man a manicure!
[8:19] Sifu Moraga: (probably not versed enough in amaericana)
[8:19] Zero Linden: We do this because we want the server to be able to make REST requests to the viewer
[8:19] Zha Ewry: nails, soacking in dish washing detergent
[8:19] Zero Linden: BUT
[8:20] Tree Kyomoon: pololive
[8:20] Zero Linden: we obviously can't open up a web server on the viewer and connect into it from the server
[8:20] Tree Kyomoon: pomolives
[8:20] Zero Linden: So, instead, we queue up the messages in the sim -
[8:20] Zero Linden: when the viewer "polls" - if there is anything waiting, we send it immediatly
[8:21] Zero Linden: if not, we .... sort .... of ... take ... a .... long .... time .... to ... say .... nothing ... here
[8:21] Sifu Moraga: In the ideal world you'd be using SCTP
[8:21] Zero Linden: and if anything comes in while we're taking our sweet time, we just say "oh, here"
[8:21] Zha Ewry: Heh. The joys of a deeply asymettriocal world
[8:22] Tree Kyomoon: asymmetry is good, gives us all time to think before we speak!
[8:22] Zero Linden: well - perhaps, but at this stage of the game, tools for HTTP over TCP are easy, readily available - and understood
[8:22] Zero Linden: So - the long poll is doable now, in world if you want ot replace XML-RPC
[8:22] Sifu Moraga: of course, I wasn't seriously suggesting it
[8:22] Gus Plisskin: What's a nominal time between polls?
[8:22] Zha Ewry: Oddball question. The client clearly loses synch from time to time, anyway we cuold have a command option to say "Hey, try and e-synch with home?"
[8:22] Zero Linden: 30 seconds
[8:23] Zero Linden: Zha - sync? actually, I think it looses all communication!
[8:23] Gus Plisskin: kinda re-log w/o relogging
[8:23] Zero Linden: Yeah - that would be nice...
[8:23] Zha Ewry: I see a lot of cases, where I konw that we'e lost our pipe
[8:23] Zha Ewry: but it could probably be re-set
[8:23] Tree Kyomoon: amen to that
[8:23] Zero Linden: I think, down the road, logging in will be only a small part of the state
[8:23] Zha Ewry: without havign to go and log.rre-log
[8:23] Gus Plisskin: yes, silent disconnects can be troublesomne
[8:23] Zero Linden: so, reconnecting to a region w/o relogging will be quite possible
[8:24] Tree Kyomoon: particularly when you keep doing building and scripting
[8:24] Zero Linden: Right now, if you lose connection with the sim, or get too far out of sync in our UDP stream, you are essentially logged out as there is no other connection to fall back on
[8:25] Zha Ewry: Right, but generally, the sim doesn't actually know.
[8:25] Zha Ewry: Most of the time, when I lose synch, when I re-log, that triggers the log-out process
[8:25] Sifu Moraga: Everyone: did I miss the llHTTPRequest restrictions discussion? I was afk for a while back a few minutes
[8:25] Tree Kyomoon: thats me too
[8:25] Zero Linden: Correct - but if you had a connection, perhaps something more reliable and less saturated, like a connection to an Agent server
[8:25] Tree Kyomoon: aha!
[8:25] Zero Linden: you could use that connection to reestablish your simulator connection
[8:26] Zero Linden: (where you = viewer in this case!)
[8:26] Zha Ewry: Which would be sweet, since we save the sim re-loading the client state
[8:26] Zha Ewry: and the cavatar losing a lot of state
[8:26] Zha Ewry: avatar
[8:26] Vitis Obviate senses a grand scheme here
[8:26] Zero Linden: Yes
[8:26] Tree Kyomoon: mmm cavatar
[8:26] Zha Ewry: Sifu, you should mention your chaining desire,
[8:27] Zero Linden: We did talk about http a bit, where there specific limits you wanted lifted, Sifu?
[8:27] Zha Ewry: or erlse I'll get it from out offliner, and pas it to zero
[8:27] Zha Ewry: pass.
[8:27] Zha Ewry: Sheesh.
[8:27] Sifu Moraga: Which ever way you want
[8:27] Zero Linden: Offline is fine too - you can either
[8:27] Zha Ewry: Give him the outline, then we cn pass the details
[8:27] Zero Linden: 1) e-mail me at
[8:27] Zero Linden: or
[8:27] Zero Linden: 2) hang it off the wiki - perhaps in my talk page
[8:28] Vitis Obviate: btw - just sent you notes on html on a prim requested last week
[8:28] Sifu Moraga: my main problem is that I need to send more data than can fit into a string
[8:28] Zero Linden: or maybe we should make a page for Zero Office Hours discussion
[8:28] Zha Ewry: oh, that last would be lovely
[8:28] Zero Linden: Vitis - thanks
[8:28] Sifu Moraga: So we wanted to chunk the data, but chunking isn't suppoerted from llHTTPRequest
[8:28] Tree Kyomoon: Zero have you had a chance to read and perhaps comment on some of the HTML on a prim notes yet?
[8:29] Sifu Moraga: OK, I'll post it to the wiki
[8:29] Sifu Moraga: (I'm also suffering serious lag from Zurich)
[8:30] Tree Kyomoon: yes, second hand scripts cause lag
[8:30] Zero Linden: er - here's the link to my page in the public wiki:
[8:30] Zero Linden: Tree, no, I spoke to the engineer doing the HTML work just yesterday
[8:30] Zero Linden: we're going to go over the stuff this week or next
[8:30] Tree Kyomoon: schweet
[8:30] Zero Linden: and he is interested in parhpas making a borader appeal for input
[8:30] Zha Ewry: Ah. I'll try and get you my use cases over the weekend then
[8:31] Zero Linden: Excellent - as I said, I loves dem use cases
[8:31] Gus Plisskin: Will we have an llClearCookies? :)
[8:31] Zha Ewry: btw, zero, this is clearly shaping up to be a nice channel for us. Thanks for the time and effort.
[8:31] Tree Kyomoon: how bout llReadCookies
[8:31] Zero Linden: we aren't planning on managing your cookies for you....
[8:31] Vitis Obviate: hear hear
[8:31] Tree Kyomoon: aye
[8:32] Zero Linden: I'm pretty sure there are some other ways of handling that ....
[8:32] Zero Linden: but - I'm all ears for use cases
[8:32] Sifu Moraga: OK, got the wiki page, tx (took a while to load for some reason)
[8:32] Zero Linden: also - I'm going to put up the latest silo.php script into the wiki
[8:32] Zero Linden: and I have other examples of advanced http usage (mostly from my personal avatar's project, the notary)
[8:33] Zha Ewry: Ohh.
[8:33] Zero Linden: would people like that stuff? thinks like doing challenge authentication, etc...
[8:33] Zha Ewry: Yes, very much
[8:33] Tree Kyomoon: dam straight
[8:33] Vitis Obviate nods
[8:33] Zha Ewry: Was having a serious discussion of how to do real "contracts" in world two nights ago
[8:33] Zha Ewry: proof of signature was a bg part of that
[8:34] Sifu Moraga: do tell, zha (perhaps backchannel, though)
[8:34] Gus Plisskin: One can send emails, PDF, or FedEX... :)
[8:34] Vitis Obviate perks up his ears bigtime - this is an important topic
[8:35] Tree Kyomoon: well the ability to keep your external auth system external, that is not rely on in world authentication
[8:35] Tree Kyomoon: so wee need access to those headers
[8:35] Zero Linden: okay - I'm going to have to sign off - now - my hour is up and I have to go code! (or my fingers atrophy and fall off)
[8:35] Sifu Moraga: Yes, well, that would be exactly what I'm doing
[8:35] Tree Kyomoon: lol
[8:35] Gus Plisskin: Back to scriptin' for me then
[8:36] Zha Ewry: Super, Zero, I think now that we have the smaller group (yoru first office hour was a zoo) this is really nice. Thanks again for the time
[8:36] Vitis Obviate: ty again Zero
[8:36] Gus Plisskin: thx Zero
[8:36] Sifu Moraga: Authentication externally, that is.
[8:36] Sifu Moraga: tx
[8:36] Zero Linden: welcome all -
[8:36] Tree Kyomoon: thanks again
[8:36] Zero Linden: I'll post this transcript in just a bit
[8:36] Sifu Moraga: cu
[8:36] Zero Linden: bye