User:Zero Linden/Office Hours/2008 September 30

From Second Life Wiki
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.
  • [13:02] Saijanai Kuhn: hey INfinity
  • [13:02] Infinity Linden: hola
  • [13:02] Morgaine Dinova: Hiya Inifinity!
  • [13:02] Infinity Linden: you want i should tap Zero on the shoulder
  • [13:03] Infinity Linden: and remind him it's 1PM SLT
  • [13:03] Infinity Linden:  ?
  • [13:03] Rex Cronon: hello everybody
  • [13:03] Infinity Linden: brb
  • [13:03] Saijanai Kuhn: sure
  • [13:03] Wrapp Seiling: hello Rex ...
  • [13:03] Rex Cronon: hi
  • [13:03] Gwen Hermit: hi all
  • [13:03] Infinity Linden: okay... ladies and gentlemen
  • [13:04] Infinity Linden: i'm gonna be the warm-up act while we're waiting for Zero
  • [13:04] Infinity Linden: Agenda items?
  • [13:04] Gwen Hermit: zero's returned?
  • [13:04] Gwen Hermit: yay
  • [13:04] Gwen Hermit: one agenda item i've literally just thought up now:
  • [13:04] Gwen Hermit: RFC for LLSD
  • [13:04] Infinity Linden: yup. we'll put that on the agenda
  • [13:04] Gwen Hermit: as well as other relevant RFCs
  • [13:04] Infinity Linden: there has been movement
  • [13:04] Gwen Hermit: RFC-ing in general
  • [13:04] Infinity Linden: in that space
  • [13:05] Infinity Linden: sure.. anyone going to Minneapolis for the IETF meeting?
  • [13:05] Gwen Hermit: if someone pays for a plane ticket
  • [13:05] Gwen Hermit: lol
  • [13:05] Infinity Linden: yeah... i couldn't justify LL letting me go this time
  • [13:05] Kael Platt: Hello everyone.
  • [13:05] Rex Cronon: hi
  • [13:05] Infinity Linden: too much OGP stuff needs getting done
  • [13:06] Dahlia Trimble: Hi :)
  • [13:06] Hirokii Hyun: OGP?
  • [13:06] Infinity Linden: Open Grid Protocol
  • [13:06] Morgaine Dinova: OGP isn't close to requiring IETF involvement :-)
  • [13:06] Infinity Linden: +1 Morgaine
  • [13:06] Gwen Hermit: OGP needs IETF involvement at some point eventually
  • [13:06] Gwen Hermit: yo zero
  • [13:06] Morgaine Dinova: Hiya Zero
  • [13:06] Zero Linden: Heeeeeeeeeeeeeeeeeeeeeeeeeeeeeeere's Zero!
  • [13:06] Kael Platt: hey o/
  • [13:06] Rex Cronon: hello zero
  • [13:07] Gwen Hermit: no coffee :O
  • [13:07] Infinity Linden: Gwen brought up the LLSD as an RFC item as an agenda point
  • [13:07] Gwen Hermit: and the more general subject of RFCing stuff
  • [13:07] Zero Linden: who ever built that victorian shower/tub object did a wonderful job ---- only it should have auto shut off!
  • [13:07] Hirokii Hyun: Infinity, you might know this. While I know this is about OGP for the most part; What's going on with the other projects at LL like Windlight?
  • Developement on that seems to have grinded to a halt.
  • [13:07] Hirokii Hyun: No details needed, just wondering if Windlight is still being worked on or if its on a backburner for now
  • [13:08] Zero Linden: the working drain plug is awesome
  • [13:08] Gwen Hermit: other departments handle that thing, right?
  • [13:08] Hirokii Hyun: They do.
  • [13:08] Infinity Linden: Hirokii... i'm definitely the wrong person to talk about WRT WindLight
  • [13:08] Zero Linden: yes - they do - and I think, other than bugs, I don't think there is active development on it right now
  • [13:08] Rex Cronon: try salami:)
  • [13:08] Infinity Linden: i'm a server person mostly
  • [13:08] Hirokii Hyun: Ah, thanks, Zero
  • [13:08] Hirokii Hyun: That's what I wanted to know.
  • [13:09] Infinity Linden: i look at a GPU and say "hey! a generic SIMD compute engine!"
  • [13:09] Hirokii Hyun: Lawl.
  • [13:09] Zero Linden: there - coffee
  • [13:09] Hirokii Hyun: Zero; Know if any of the WL project staff run office hours?
  • [13:09] Hirokii Hyun: I can't remember the people involved in WL off the top of my head.
  • [13:09] Hirokii Hyun: Bigpappi is one, isn't he?
  • [13:09] Zero Linden: I don't know, sorry
  • [13:09] Zero Linden: yes - he would be in the WL loop
  • [13:10] Hirokii Hyun: Got it.
  • [13:10] Morgaine Dinova: If we're not overwhelmed with agenda items, could you give us some idea of how server-side is tested at LL? And the state of the QA team in
  • general, operationally.
  • [13:10] Zero Linden: So - welcome all to my Office hours
  • [13:10] Zero Linden: as is well known, but bears repeating - the transcripts are published in the wiki (thanks Tree!)
  • [13:11] Zero Linden: so - speak freely - speak openly
  • [13:11] Zero Linden: So - I hear agenda item #1
  • [13:11] Zero Linden: 1) LLSD RFC
  • [13:11] Zero Linden: others
  • [13:11] Saijanai Kuhn: group IM in OGP (not new design, old design)
  • [13:11] Morgaine Dinova: Good one Sai
  • [13:11] Zha Ewry: As is less well known, latecomers shoudl bring thier own bean bag chairs.
  • [13:12] Teravus Ousley: One that I was going to bring up.. but eventually found a solution to.. is the client only receives the 'No events for you' message in
  • EventQueueGet if it's a HTTP/1.0 response.
  • [13:12] BJ Mayer: that was a bad crash
  • [13:12] Teravus Ousley: but the client doesn't specifically say it wants a 1.0 response.. so many servers say .. it didn't specify.. so we're sending it HTTP/1.1
  • [13:12] Infinity Linden: @Terevus... did you report that one?
  • [13:12] Zero Linden: Teravus - more like if it isn't a KeepAlive
  • [13:13] Teravus Ousley: Well.. I worked for a long time over last weekend trying all sorts of crazy ways to get it to respond.. and the only way it works with with
  • HTTP/1.0 502 Upstream error:
  • [13:13] surface Easterwood: hey . where can a person get a chair?
  • [13:13] Morgaine Dinova: Oh, Whump's here (hiya :P). What was that question we had last time, which was answered with "You should really wait until Whump is here for
  • that." ?
  • [13:13] Zero Linden: The floor has a nice rug....
  • [13:14] Zero Linden: (at least I like it... in fact it is one of the SL things that I see and think - I'd like that in RL!)
  • [13:14] surface Easterwood: yea it is. trade ya?
  • [13:14] Zero Linden: Okay -
  • [13:14] Zero Linden: so let's begin, shall we
  • [13:14] Zero Linden: 1) LLSD RFC
  • [13:15] Zero Linden: Infinity has been very hard at work on getting a draft ready
  • [13:15] Dahlia Trimble: was one published?
  • [13:15] Zha Ewry: Was
  • [13:15] Zero Linden: Not yet.....
  • [13:15] Zha Ewry: I thought I saw one pointed to last night, or night before
  • [13:15] Zha Ewry: skims mail
  • [13:15] Gwen Hermit: there's a very decent specs document on the wiki for LLSD
  • [13:15] Gwen Hermit: would not take much to RFC-alise it
  • [13:16] Gwen Hermit: though i'd like to see other things, like this fancy new LLIDL, all RFC-alised
  • [13:16] Infinity Linden: yup. the LLSD RFC will essentially be all the things on the WIKI and in the SPEC rolled up into one document
  • [13:16] Gwen Hermit: and eventually (after this long R&D phase) OGP
  • [13:16] Zha Ewry: [1]
  • [13:16] MistressHolly Lemon: lol sorry
  • [13:16] Infinity Linden: Gwen.. i think you will see that "soon"
  • [13:16] Infinity Linden: the definition of "soon" is left as an exercise to the reader
  • [13:17] Gwen Hermit: infinity - how soon? OGP is still under heavy development........
  • [13:17] Gwen Hermit: that would not be my version of soon for one
  • [13:17] MistressHolly Lemon: i have a question as to wether somthing is possible
  • [13:17] Gwen Hermit: the other thing while on RFCs: IANA registration of port numbers - ever going to happen?
  • [13:17] Infinity Linden: Zero... can we do one more round of internal review on the ID and then convert the ID to be the LLSD section of the OGP spec?
  • [13:17] Zero Linden: it will have three parts: 1) type system and data abstraction, 2) serialization formats (XML, JSON & Binary), 3) Description langauge
  • [13:17] MistressHolly Lemon: would holding assets on an opensim and using them on this grid possible?
  • [13:18] Zero Linden: Infinity - let's target later this week for that
  • [13:18] Infinity Linden: 'k
  • [13:18] Teravus Ousley: :D
  • [13:18] Zero Linden: Gwen - this portion will go out sooner than the other parts
  • [13:18] MistressHolly Lemon: i been thinking about the whole 3d import thing
  • [13:18] Infinity Linden: and Gwen... are you talking about registering port numbers for simulators?
  • [13:18] MistressHolly Lemon: its possible on open sim so why not get open sim to send the finished prim over?
  • [13:18] Zero Linden: [Notes second Agenda item: 2) asset location]
  • [13:19] surface Easterwood: have anyone taken notice that there a giant bird in the room?
  • [13:19] Infinity Linden: and people are not wearing enough hats
  • [13:19] Morgaine Dinova: We're not specist here
  • [13:19] Gwen Hermit: infinity - yes
  • [13:19] Gwen Hermit: and for the CAPS stuff too
  • [13:19] Dahlia Trimble: its ok as long as it isnt flying overhead
  • [13:19] Zero Linden: Usually, committees ignore the giant elephant in the room --- I suppose it applies to birds too
  • [13:20] Teravus Ousley: yep.. or the purple dragon
  • [13:20] Zero Linden: adjusts his hat
  • [13:20] MistressHolly Lemon: anyone here know how asset transport would be handled?
  • [13:20] Zha Ewry: I worry at least a smuch as poepl egetting burned by whump
  • [13:20] Zha Ewry: eyes the steam dischage
  • [13:20] Whump Linden: Zha, the warning label would be larger than my avatar.
  • [13:20] Dahlia Trimble: that's CO2 gas ;0
  • [13:20] Morgaine Dinova: He's brewing my tea, leave him alone
  • [13:20] Infinity Linden: hold up MsHolly... let's finish the RFC discussion
  • [13:21] MistressHolly Lemon: ok
  • [13:21] Zero Linden: So - has anyone looked over the IDL stuff yet?
  • [13:21] Infinity Linden: Gwen... can we take that offline... i'd love to get more discussion amongst OpenSim, PyOGP
  • [13:22] Saijanai Kuhn: is in irc all day
  • [13:22] Gwen Hermit: infinity - we can
  • [13:22] Infinity Linden: i just wanna make sure we know what all the requirements are
  • [13:22] Gwen Hermit: is Gareth everywhere btw - see you on IRC or IM later
  • [13:24] Zha Ewry: The IDL stuff, looks promising.. I was actually wondering if anyone has looked at how hard it owuld be do to a mapping between it and some of the
  • other IDLs
  • [13:24] Gwen Hermit: are there pre-existing IDLs for web services?
  • [13:24] surface Easterwood: maybe we should cut back on the development til the bailout is finalized> you think?
  • [13:24] Infinity Linden: if you mention CORBA IDL, i'll scream
  • [13:24] Saijanai Kuhn: the bailout will never be finalized. IMHO
  • [13:24] Zha Ewry: There are a bunch of them, Gewn, none wonderful
  • [13:25] Gwen Hermit: zha - any of them feasible for our purposes?
  • [13:25] Gwen Hermit: they need not be "wonderful", just "sufficiently clean and STANDARD"
  • [13:25] Zha Ewry: chuckles.
  • [13:25] Zero Linden: points to some soapboxes a few sims away for folks wishing to discuss the bailout....
  • [13:25] Infinity Linden: btw... we developed our own DL 'cause it was the minimum for what we needed
  • [13:25] Saijanai Kuhn: shuts his trap
  • [13:25] Gwen Hermit: i'll go with infinity on that....
  • [13:25] Infinity Linden: hopefully this will make it easier to map to other type systems
  • [13:26] Zha Ewry: Yeha, I was re-reading it
  • [13:26] Infinity Linden: going from LLIDL to CORBA, ASN.1, etc. should be "trivial"
  • [13:26] Zha Ewry: and thinking I wonder if thi sis a proper subset of most.
  • [13:26] Zha Ewry: Yeah, it should
  • [13:26] Zero Linden: Zha - I looked at a few dozen --- there aren't many that are general enough to handle REST, yet not so enormous that we wouldn't all run screaming
  • [13:26] Gwen Hermit: one thing that could be cool is a script to take LLIDL and produce stub code for something like mulib
  • [13:26] Infinity Linden: the other way would be easy, as long as descriptions use only bits that are also in LLIDL
  • [13:26] Zero Linden: most of the smaller ones, are all firmly entrenched in RPC
  • [13:26] Gwen Hermit: probably something that belongs in pyogp
  • [13:27] Zero Linden: as the idiom of description
  • [13:27] Morgaine Dinova: And hence unRESTful
  • [13:27] Infinity Linden: @Gwen... yeah... that's what we're shooting for
  • [13:27] Zero Linden: or Classes and local callable APIs
  • [13:27] Infinity Linden: though
  • [13:27] Zha Ewry: I think.. that we might want to show the mapping up, in a few cases
  • [13:27] Zero Linden: WSDL is somewhat more flexible.... but..... Infinity and Zha will both run screaming....
  • [13:27] Zha Ewry: (and prove we really can, so we shut up the people who say we should use a bigger one)
  • [13:27] Zha Ewry: sighs
  • [13:27] Infinity Linden: internally... we have a preference in LL for using dynamic languages
  • [13:27] Gwen Hermit: if we could extend it with proper typing, perhaps default values etc it could be quite powerful for automating new implementations fast
  • [13:27] Gwen Hermit: infinity - i've noticed ;)
  • [13:27] Zha Ewry: I will walk away quitely crying. Much more dignified.
  • [13:27] Infinity Linden: so you don't have to explicitly define types or structures for each message
  • [13:28] Infinity Linden: which makes parsers less of an issue
  • [13:28] Infinity Linden: but validators... yeah...
  • [13:28] Infinity Linden: that's good
  • [13:28] Zha Ewry: I don't htink anyone is going to suggest WSDL, or BPEL, or... anything that heavy
  • [13:28] Zero Linden: I think, our major reasons for wanting an IDL were two-fold
  • [13:28] Infinity Linden: yay! no one's recommending WSDL at Zero's Office Hours!
  • [13:28] Zero Linden: 1) Clear, unambiguous way to specify resource classes in the OGP spec
  • [13:29] Morgaine Dinova: 2) To make Infi feel at home :P
  • [13:29] Zero Linden: 2) Ability to generate good diagnostic information for programmers
  • [13:29] Infinity Linden: lol
  • [13:29] Saijanai Kuhn: drinks to that having spent hours trying to disambiguate things for TP
  • [13:30] Teravus Ousley: <boolean>1</boolean>.... <boolean>true</boolean>
  • [13:30] Zero Linden: Most IDLs are hitting something like point 1 --- but they often have "generate stub code" and/or "dynamically bind" as other reasons
  • [13:30] Infinity Linden: @Teravus... in theory, you should never see <boolean>1</boolean> in the XML serialization
  • [13:30] Gwen Hermit: teravus - i ran into that one
  • [13:30] Gwen Hermit: opensim does 1
  • [13:31] Teravus Ousley: apparently, so does the OGP code..
  • [13:31] Infinity Linden: ugh
  • [13:31] Dahlia Trimble: lol
  • [13:31] Zha Ewry: Its painfully common, Inf.
  • [13:31] Gwen Hermit: code i had to use to get the output of llsd.py recogonised by aditi:
  • [13:31] Gwen Hermit: resp = resp.replace('<?xml version="1.0" ?>',).replace('true','1')
  • [13:31] Zero Linden: you can shoot Phoenix and me for that... we probably should never have allowed both
  • [13:31] MistressHolly Lemon: im guessin that aint lsl
  • [13:31] Zero Linden: you had to remove the XML decl? why?
  • [13:31] Gwen Hermit: zero - it didn't like it
  • [13:32] Zero Linden:  !
  • [13:32] Zero Linden: well - that is f**dup
  • [13:32] Infinity Linden: hmm... i'm thinking we should get the protocol verifier written and use it to identify where our code is messing up
  • [13:32] Zero Linden: not the least is XML well-formedness (not - not validated, just well-formed)
  • [13:33] Zha Ewry: We should
  • [13:33] Zero Linden: (there is little point in running the XML through a validating parser if you are going to run it into a LLSD parser -- since that will essentially
  • validate it as it parses)
  • [13:33] Zha Ewry: it would be nice to be able to say, "Here, it has to meet this, if it doesn't we've goofed"
  • [13:33] Whump Linden: Gwen, would you write up a JIRA and link it to [2]
  • [13:33] Gwen Hermit: whump - remind me with a PM on IRC
  • [13:33] Whump Linden: kk
  • [13:33] Zero Linden: Zha - it is a little tricky with something as mutable as LLSD - and our goal for very ressiliant structures --
  • [13:34] Zha Ewry: nods
  • [13:34] Infinity Linden: no no... validate the morphology of the message and the lexical goodiness of the LLSD (not the XLM serializatinon)
  • [13:34] Zero Linden: but we will be able to say things like "hmmmm... permissions should have been a map, not an integer" or some such thing
  • [13:34] Zha Ewry: I think it's a litle tricky.. but not unsuable.
  • [13:34] Zha Ewry: *undoable (sheesh)(
  • [13:34] Zero Linden: I agree totally
  • [13:34] MistressHolly Lemon: what u goin on about
  • [13:34] MistressHolly Lemon: lol
  • [13:34] Gwen Hermit: you talking about an LLSD validator as in something which validates messages against LLIDL?
  • [13:34] Infinity Linden: yes
  • [13:34] Gwen Hermit: or a more boring XML DTD validation?
  • [13:34] Zero Linden: the former
  • [13:34] Gwen Hermit: the former sounds cool
  • [13:35] Infinity Linden: for the "morphology" of the "sentence"
  • [13:35] MistressHolly Lemon: ok head hurts now
  • [13:35] Zha Ewry: The former
  • [13:35] Zero Linden: the dream is this
  • [13:35] MistressHolly Lemon: lol sounds like some kinda xml sl uses
  • [13:35] Zero Linden: you take the XML spec, use XSLT to extract all the sections marked role="LLIDL"
  • [13:35] Zero Linden: you catatenate those together
  • [13:35] Infinity Linden: i.e. - the agent_login message DOES contain an identifier and an authenticator, and the authenticator type is one of: 'hash', 'challenge' and
  • 'pkcs5pbkdf2'
  • [13:35] Infinity Linden: that sort of thing
  • [13:35] Zero Linden: you put that in a file called: OGP.llidl
  • [13:36] Infinity Linden: a step up teh stack frmo the XML de-serialization
  • [13:36] Zero Linden: then put that in some config dir and turn on message validation
  • [13:36] Zha Ewry: That is what you'd like to be able to do.. Drive it as much as possible that way.
  • [13:36] Zero Linden: then ... bang ... the code reads the OGP.llidl file, and for each resource class that has an LLIDL entry, checks the morphology
  • [13:36] Zero Linden: and reports errors to a log
  • [13:37] Zero Linden: or - even better - in headers in the responses!
  • [13:37] Zero Linden: well - both
  • [13:38] Zha Ewry: One thing we noticed in the early OGP debugging is how hard it was to get some of it right, you'd just get 50X errros. it woudl rock, if the
  • validtaoin could
  • [13:38] Gwen Hermit: fatal errors should give a different HTTP response code
  • [13:38] Zero Linden: "X-LLIDL-Error: request entry foo.bar[3].position needs to be an array of three reals, not a strng"
  • [13:38] Zha Ewry: tell you "Hey, that needes a map of 3 integrets here"
  • [13:38] Zha Ewry: +127.78 Zero.
  • [13:38] Gwen Hermit: zha and zero: telepathically joined since 2008
  • [13:38] Zha Ewry: laughs
  • [13:38] Saijanai Kuhn: shudders
  • [13:38] Infinity Linden: @gwen... we're working on that one
  • [13:38] Zero Linden: "X-Points: Zha;+1
  • [13:39] Dahlia Trimble: +127.70 Gwen
  • [13:39] Zero Linden: really - we are in different time zones - and don't have a back channel on AIM going
  • [13:39] Gwen Hermit: map(present_avatars, lambda x: x++)
  • [13:39] Zero Linden: +10 to the first person to write that a monad
  • [13:39] Infinity Linden: application/llsd+xml ; charset=UTF-8; randommessage="Zero++; Zha++; Gwen++"
  • [13:39] Zha Ewry: wimpers at Infinity
  • [13:40] Infinity Linden: uh oh.. i think i might have gone over 72 characters on that line
  • [13:40] Zero Linden: okay - looks like we are ready to hit Agenda Item 2
  • [13:40] Zero Linden: Asset locations
  • [13:40] Morgaine Dinova: Brb, I'll just slip into a continuation
  • [13:40] Zero Linden: MistressHolly - did you want to restate the topic question?
  • [13:40] Infinity Linden: refuses to make lisp jokes though she really wants to
  • [13:40] MistressHolly Lemon: yes
  • [13:40] Zero Linden: Monad in Bernaise sauce, anyone?
  • [13:40] MistressHolly Lemon: i had some suggestions towards the 3d import
  • [13:41] MistressHolly Lemon: cross grid
  • [13:41] MistressHolly Lemon: could it be imported on opensim then a finished product be sent to this grid
  • [13:41] Gwen Hermit: i'd think the idea of cross-grid import dies if we have proper URLs for assets rather than IDs
  • [13:42] MistressHolly Lemon: my idea is to stream the data like you would media
  • [13:42] MistressHolly Lemon: but from grid to grid
  • [13:42] Zero Linden: Or to put it another way, we've been generally thinking of assets as URL addressable bundles of bits --- and hence they aren't so much imported
  • [13:42] Zero Linden: as they are referenced
  • [13:42] Tree Kyomoon: would be interesting to have a professional "content developer" grid or client you could post finished content to
  • [13:42] MistressHolly Lemon: like obj 3ds
  • [13:42] Gwen Hermit: zero++
  • [13:42] Morgaine Dinova: Infi: [3]
  • [13:42] MistressHolly Lemon: its all just verts
  • [13:42] Infinity Linden: but how does that map to "objcet portability"
  • [13:42] Zha Ewry: And you need to be vert careful about
  • [13:42] Zero Linden: Once they have URLs, assuming they are HTTP or HTTPS URLs then they are fetched via HTTP
  • [13:42] Zha Ewry: what you are speaking of
  • [13:42] Infinity Linden: when you want to move frmo one agent domain to another?
  • [13:42] Gwen Hermit: morgaine - everyone knows the universe is coded in cython ;)
  • [13:42] Zha Ewry: Asset offline, asset in a sim, asset on an ave, etc.
  • [13:42] Dahlia Trimble: SL™ doesnt work well with verts
  • [13:43] MistressHolly Lemon: the open sim can handle imporing
  • [13:43] MistressHolly Lemon: could that interpreted data then be sent to this grid
  • [13:43] Zero Linden: groans that the utter truth to that xkcd
  • [13:43] Gwen Hermit: mistressholly - if you have the right asset fetch protocol
  • [13:44] surface Easterwood: ok i im you all keep me informed of the progress. nice to see you all hard at work .goodby , and have a nice day
  • [13:44] Gwen Hermit: quantum mechanics runs on lisp for certain........
  • [13:44] MistressHolly Lemon: i see so open sim and linden would need to work very closely for it to be a reality
  • [13:44] Zero Linden: later, Surface
  • [13:44] Gwen Hermit: heh, OGS's OpenAsset already uses a nice REST protocol
  • [13:44] Gwen Hermit: only thing that sucks is the serialisation used
  • [13:44] Gwen Hermit: it's an XML file with a binary blob embedded as base64
  • [13:44] Zero Linden: ouch
  • [13:45] Teravus Ousley: well, maybe we could write a web service for importing 'content'.. but.... we still have not identified all of the 'trust' factors yet.
  • [13:45] Morgaine Dinova: Aye, that xkcd is not just a good joke, but a sad reflection on the state of computing today.
  • [13:45] Zha Ewry: The issues, have far more to do with getting permisions and trust and policy right than moving the prims
  • [13:45] MistressHolly Lemon: hmmm what if it were possible to convert it to a format sl understands
  • [13:45] Latha Serevi: Do we have, or are we considering, any way for two assets, at URLs on two different servers, to be "the same" or "derived" objects in any way? Or
  • do we assume that a copy has no relation to the original?
  • [13:45] MistressHolly Lemon: the normal prims r xml correct?
  • [13:45] Zero Linden: right - we can talk of moving the prims / textures / animations / sounds/ etc.... just as we do today on the web
  • [13:45] Zero Linden: the trust issues come in with deciding to give someone access to those bits or no
  • [13:45] Gwen Hermit: [4]
  • [13:45] Gwen Hermit: an example
  • [13:46] Gwen Hermit: that URL works in a regular browser
  • [13:46] Dahlia Trimble: Latha, a copy is usually a reference until it's been rezzed
  • [13:46] Gwen Hermit: (btw, that one isn't actually opensim, it's pygrid - but same format)
  • [13:46] Latha Serevi: Dahlia, that seems unrelated to my question.
  • [13:46] Infinity Linden: Morgaine, we did not hack SL up in perl. we used python
  • [13:46] Zero Linden: wonders how those XMLSchema namespaces made it into that document....
  • [13:46] MistressHolly Lemon: hmmm so qarl uses maya?
  • [13:47] Morgaine Dinova: Infi: yeah, I noticed, or the startup time would be faster ;P
  • [13:47] Gwen Hermit: zero - ask Guido
  • [13:48] Zero Linden: Latha - I think it would be best if we can make the assumption that once copied - at the protocol level, they are distinct
  • [13:48] Gwen Hermit: question - will we still have asset UUIDs in the great future?
  • [13:48] Gwen Hermit: or will we just use URLs everywhere?
  • [13:48] Zha Ewry: REST would storngly imply thy are different URLs
  • [13:48] Zero Linden: Well - URLs everywhere is lovely and general
  • [13:48] Zha Ewry: You've created a new resource, it gets a new URL
  • [13:49] Morgaine Dinova: I did some handcoding and benchmarking on Parrot yesterday, was fun. It's disastrously slow, even in optimized mode. It must be jam packed
  • with debug or something.
  • [13:49] Gwen Hermit: i'd be concerned that URLs can have the content behind them change while assets today are immutable i believe?
  • [13:49] Zero Linden: BUT for some uses, UUIDs are worth the compactness ... but I think in those situations, it is like the UUID has a "base URL" that they are off of
  • [13:49] Infinity Linden: the benefit of UUID based addressing is that if you add a registration, you can have object portability
  • [13:49] Infinity Linden: but the benefit of URL based addressing is that it doesn't require registration
  • [13:49] Zha Ewry: the problem is naked UUIDs kill you
  • [13:49] Zero Linden: Gwen - assets being immutable under REST is equivalent to saying you can cache it forever
  • [13:49] Gwen Hermit: i like the base URL idea for some uses, but should it be enforced on the protocol level?
  • [13:50] Dahlia Trimble: "naked" URLs
  • [13:50] Gwen Hermit: for legacy things, that makes sense
  • [13:50] Dahlia Trimble:  ?
  • [13:50] Gwen Hermit: Zero - a TTL?
  • [13:50] Infinity Linden: Zha... assuming there was a great registry in ths sky where you could resolve the URL for a UUID
  • [13:50] Gwen Hermit: for most assets the TTL could be 0 (forever)
  • [13:50] Infinity Linden: would it still kill you?
  • [13:50] Zha Ewry: sighs
  • [13:50] Zha Ewry: Yeah, I think it would
  • [13:50] Zero Linden: I don't think anything in the system today relies on the property that, once forgotten, if you refetch the asset you'll get the same bits
  • [13:50] Gwen Hermit: that great registry would of course be LL ;)
  • [13:50] Zero Linden: that is a subtle difference, but an important one
  • [13:50] Tree Kyomoon: controlled by diebold of course
  • [13:50] Gwen Hermit: you just saved your business model ;)
  • [13:50] Gwen Hermit: become ICANN for assets
  • [13:51] Teravus Ousley: diebold.. haha
  • [13:51] MistressHolly Lemon: eeeer did anyone get what i was trying to tell you?
  • [13:51] Infinity Linden: we could say things are URIs
  • [13:51] MistressHolly Lemon: the obj format
  • [13:51] MistressHolly Lemon: is just numbers of what should go where
  • [13:51] Gwen Hermit: mistressholly - the cross-grid thing and importing?
  • [13:51] Morgaine Dinova: Well, it may be controlled by Verisign ultimately, because of certs, and they are every bit as clueless as Diebold.
  • [13:51] Infinity Linden: then if they have the form ['http://somwhere.com/asset/something'] then it's interpreted as a URL and accessed
  • [13:51] Dahlia Trimble: MistressHolly, nothing is SL™ is compatable with .obj format
  • [13:51] Infinity Linden: ahem
  • [13:52] Infinity Linden: may we stop critisizing VeriSign?
  • [13:52] MistressHolly Lemon: no but its possible to take that data and change it?
  • [13:52] Zero Linden: Yes, currently object format is a collection of prim parameters, texturing information, and contents (which are references to other assets, with
  • permission controls on 'em)
  • [13:52] Gwen Hermit: sculpties and other mess
  • [13:52] Rex Cronon: into sculpties
  • [13:52] Morgaine Dinova: Infi: no, it's fun, and accurate :P
  • [13:52] Gwen Hermit: i like the "if it's a UUID, just ask your local region domain for it, if it's a URL, go fishing"
  • [13:52] Infinity Linden: it's also a system i am very proud of
  • [13:53] MistressHolly Lemon: well what if you could map it if that makes sense
  • [13:53] Gwen Hermit: where does UUID resolution belong?
  • [13:53] Zero Linden: As in, access it directly off the original domain?
  • [13:53] MistressHolly Lemon: take the xml normal prims use and allow extra faces
  • [13:53] Zha Ewry: If you have contextt, you're OK
  • [13:53] Zha Ewry: its the naked UUID that kills you
  • [13:53] MistressHolly Lemon: sub deviding
  • [13:53] Gwen Hermit: winks at morgaine
  • [13:53] Infinity Linden: oh oh.. .Zha.. .i see what you're saying
  • [13:53] Infinity Linden: but URLs in this system aer naked as well
  • [13:54] Infinity Linden: without context
  • [13:54] Gwen Hermit: zha - the naked UUID can be resolved against your local simulator
  • [13:54] Tree Kyomoon: could we get those UPC guys to handle UUIDs?
  • [13:54] Infinity Linden: unless you allow introspection for all caps
  • [13:54] Zha Ewry: That's context
  • [13:54] Gwen Hermit: the local simulator could even give a UUID to cached assets which were originally URLs
  • [13:54] Zero Linden: URLs are not naked..... so long as they are not relative-URLs
  • [13:54] Infinity Linden: @Tree.. lol
  • [13:54] Morgaine Dinova: Infi: you can't be proud of their DNS error hijacking, nor of the lack of training in their front desk, surely :-)
  • [13:54] Infinity Linden: i'll have to get a tattoo of the bar code of my UUID
  • [13:54] Tree Kyomoon: in your RFID chip of course
  • [13:54] Zha Ewry: The point, being that you need to be very careful to never let naked UUIDs escaep from thier context
  • [13:54] Gwen Hermit: on the right hand
  • [13:55] Teravus Ousley: .. yep.. those new ones that 'can't be duplicated...... yet'
  • [13:55] Gwen Hermit: hey.... in SL we all have a number which controls whether we can buy or sell!
  • [13:55] Infinity Linden: Morgaine. .let's take this offline. this is not the forum for you to describe problems you've heard other people have with VeriSign
  • [13:55] Zero Linden: Uhm, as amusing as it is to discuss the failings of other companies... let's leave that for a session conducted over beer
  • [13:55] Tree Kyomoon: mmm naked UUID's
  • [13:56] Gwen Hermit: not a byte left to the imagination!
  • [13:56] Infinity Linden: it's oky... we can say "naked UUID" here... it's a mature region
  • [13:56] Gwen Hermit: can we BE naked UUIDs?
  • [13:56] Gwen Hermit: gets a large transparent prim with nothing but a UUID on it and wears it.....
  • [13:56] Morgaine Dinova: Good idea Zero .... although what we've just heard is "interesting"
  • [13:56] Saijanai Kuhn: MistressHolly seems to me that you need to check what folks have been saying in the jiras about handling.obj format vs new tools for content creation
  • [13:56] Gwen Hermit: i don't think there's going to be a clean way to map OBJs straight to SL
  • [13:56] Gwen Hermit: better to export in a higher level foramt
  • [13:56] MistressHolly Lemon: maybe not obj
  • [13:56] Gwen Hermit: *format
  • [13:57] Saijanai Kuhn: [5]
  • [13:57] MistressHolly Lemon: but obj is easily converted
  • [13:57] Gwen Hermit: VRML?
  • [13:57] Gwen Hermit: lol
  • [13:57] Zha Ewry: I'm compelled to gently mention Collada edge format
  • [13:57] Rex Cronon: obj is possible
  • [13:57] Teravus Ousley: well, the UDP protocol doesn't define anything for obj
  • [13:57] Teravus Ousley: therefore.. you don't see them..
  • [13:57] Saijanai Kuhn: From Qarl: hey all - i'm trying to gather resident feedback for an internal LL debate. the question is - do residents want LL to spend more resources on
  • A) allowing mesh import or B) improving existing modeling tools. sorry for the vagueness here - mainly i'm trying for a general feel for resident sentiment.
  • [13:57] Infinity Linden: i don't think there's going to be much X3D stuff ehre
  • [13:58] Dahlia Trimble: obj files can be huge and would need to be segmented before transferred
  • [13:58] Saijanai Kuhn: [6]
  • [13:58] Rex Cronon: udp doesn't define things for lots of things, but does it matter?
  • [13:58] Gwen Hermit: sai - your sid is showing ;)
  • [13:58] Gwen Hermit: i think meshes should be assets
  • [13:58] Morgaine Dinova: Surely it would be better to work on full support for extensible types, and then anyone can make mesh formats or whatever. Having LL work on
  • one specific new types is just time wasted.
  • [13:58] Gwen Hermit: a new asset type
  • [13:58] Infinity Linden: okay... if Zha gets to mention Collada... i get to mention the idea that we store everything as nurbs
  • [13:58] Zero Linden: security breech!
  • [13:58] Gwen Hermit: and let you assign a mesh to a prim
  • [13:58] MistressHolly Lemon: mesh prims
  • [13:59] Dahlia Trimble: obj can do nurbs
  • [13:59] Tree Kyomoon: yay nurbs
  • [13:59] Gwen Hermit: sai - unless JIRA is sane enough to use IP tieing, logout
  • [13:59] Rex Cronon: not all obj r huge
  • [13:59] Zha Ewry: laughs
  • [13:59] Saijanai Kuhn: I c licked on the link in the jira
  • [13:59] Infinity Linden: @Morgaine.. yes... if you have no installed base you can say that
  • [13:59] Zha Ewry: Not general Collada, just.. the edge format
  • [13:59] Gwen Hermit: you'd be surprised how many places DON'T tie sessions to IPs.....
  • [13:59] Tree Kyomoon: wants bezier control handles on all the points in SL
  • [13:59] Zero Linden: Gwen - it appears to be sane enough
  • [13:59] Gwen Hermit:  :)
  • [13:59] Gwen Hermit: mesh prims would be a rather sexy idea
  • [14:00] Dahlia Trimble: wants bezier handles on her avatar ;)
  • [14:00] Zha Ewry: Next logical step from scuplties, in many ways
  • [14:00] Infinity Linden: Tree and i are going to run off and make nurbs prims
  • [14:00] Zero Linden: Gwen - you can't always - some people are behind HTTP proxies that use blocks of IPs in rotation and don't have "stickiness"
  • [14:00] Gwen Hermit: i'd get myself (literally - the actual me) scanned in 3D and then make the phrase "i got a virtual sex change" MEAN something
  • [14:00] MistressHolly Lemon: that is the exact thing i have been trying to talk about gwen
  • [14:00] Tree Kyomoon: and procedural prims. mmm procedural prims...and textures..mmmm...
  • [14:00] Gwen Hermit: mistress - you weren't clear on it
  • [14:00] MistressHolly Lemon: taking the idea of sculpties and getting rid of the restrictions
  • [14:00] Saijanai Kuhn: so anyway, guess my topic won't be raised. title said it all anyway
  • [14:00] Gwen Hermit: replace sculpty textures with meshes
  • [14:00] Zero Linden: wonders where the compute power for a region with 1000 procedural prims is going to come from....
  • [14:01] Gwen Hermit: zero - limit the number of meshes per region
  • [14:01] Latha Serevi: I'm thinking of trying to ask my object-identity question again a different way. Or should I save that for some other time?
  • [14:01] Morgaine Dinova: Infi: well there is no installed base for a mesh type that doesn't exist. And we have to have extensible primitives in OGP or else we'll be
  • stuck in the current backwater forever instead of making a protocol for the future.
  • [14:01] MistressHolly Lemon: could a sculpt just as easily be called from a file rather than an image?
  • [14:01] Dahlia Trimble: its a viewer problem ;)
  • [14:01] Gwen Hermit: physics
  • [14:01] Zha Ewry: offers to sell you a rack of bluegenes
  • [14:01] Tree Kyomoon: nanotubes and graphenes of course!
  • [14:01] Rex Cronon: i think that obj is the bnext logical step:)
  • [14:01] Gwen Hermit: physics!
  • [14:01] Infinity Linden: @Zero.. .GPGPU
  • [14:01] MistressHolly Lemon: obj isnt complex its just vertex locations
  • [14:01] MistressHolly Lemon: alot like a sculptys map
  • [14:01] Whump Linden: Saij, we can hold it for tomorrows' OO, and Morgaine's question too.
  • [14:01] Zha Ewry: idly calculates the commision
  • [14:01] Dahlia Trimble: obj is a lot more than vertices
  • [14:01] Rex Cronon: and facets
  • [14:01] Saijanai Kuhn: morgaine did you see what Avi Bar-Zeev saidabout the original primitives design he worked o for SL?
  • [14:02] Morgaine Dinova: Got a link Sai?
  • [14:02] Gwen Hermit: the physics would be the chokepoint on mesh prism
  • [14:02] Gwen Hermit: *prims
  • [14:02] MistressHolly Lemon: yes but it will take no more time than 50 sculpts
  • [14:02] Tree Kyomoon: gives people a reason to upgrade their video cards anyway
  • [14:02] Gwen Hermit: hmm.... how about PhysX cards on the LL sims?
  • [14:02] Gwen Hermit: would that help any?
  • [14:02] Gwen Hermit: offload from the CPU
  • [14:02] Saijanai Kuhn: [7]
  • [14:02] Zha Ewry: Not so much, Gwn, unles syou get Havok to really support that
  • [14:02] Morgaine Dinova: Ta
  • [14:02] Gwen Hermit: or are these hardware physics accelerators really that crap
  • [14:02] MistressHolly Lemon: hmmmmm what about streaming it rather than uploading
  • [14:02] Gwen Hermit: Zha - i'm talking about replacing havok
  • [14:03] Infinity Linden: @Gwen... we would need a version of havoc that used the NVidia GPGPU interface
  • [14:03] Gwen Hermit: ok, granted - that's rather a bigger task
  • [14:03] Tree Kyomoon: I thought we were going plug and play with the physics engine?
  • [14:03] Dahlia Trimble: hits up rl for a change... bye all :)
  • [14:03] Gwen Hermit: bye
  • [14:03] Rex Cronon: bye dahlia
  • [14:03] Saijanai Kuhn: won't Havok more likely support Intel's new design than NVIdia?
  • [14:03] Infinity Linden: and changing physics engines right now is probably not somethign that's gonna geta lot of traction
  • [14:03] Zero Linden: wellllllll...
  • [14:03] Zero Linden: it is that time!
  • [14:03] Gwen Hermit: yeah, indeed
  • [14:03] Zero Linden: thanks for all coming
  • [14:03] Gwen Hermit: it's over?
  • [14:03] Infinity Linden: aren't PhysX cards from nvidia?
  • [14:03] Rex Cronon: bye zero
  • [14:03] Gwen Hermit: nnnnnooooooooooooo
  • [14:03] Tree Kyomoon: can someone pass me a transcript from the beginning?