AW Groupies/Chat Logs/AWGroupies-2008-10-07

[9:29] Morgaine Dinova: Hiya Zha
[9:29] Zha Ewry: settles in and waits for a quorum
[9:29] Saijanai: Kuhn: Hey Fearless Leader
[9:29] Saijanai: Kuhn: I done did my spam
[9:29] Morgaine Dinova: Would a doughnut do instead of a quorum?
[9:30] Deanna Trollop: Doughnuts will do instead of a great many things.
[9:30] Morgaine Dinova: Haha. Hiya Deanna
[9:31] Saijanai: Kuhn: do we have an agenda?
[9:31] Morgaine Dinova: Only doughnut worship so far
[9:32] Saijanai: Kuhn: is that anything like wingnuts?
[9:32] Latha Serevi: Starter topic, perhaps not for this meeting --- I've been sniffing around about the possibility of contributing to an open grid protocol spec that takes Zha's approach in spirit -- support multiple levels of interoperability, framework supports replaceable sub-parts, etc -- but is hosted at a neutral location and has some sort of community process. The hope would be to give friendly LLers an open process to join, but not control. So far, the right approach has not made itself obvious. Anyone have any input?
[9:32] Morgaine Dinova: Dunno. Are those like right wing nuts?
[9:33] Deanna Trollop: Only in girth contribution, using Limbaugh as the prototype
[9:33] Morgaine Dinova: That sounds interesting, Latha
[9:33] Deanna Trollop: Doughnuts, that is.
[9:34] Lillie Yifu: hihi can I ask that we skip th rl politics and stick t the OGP things... there's enough drama in waht we need to do anyway...
[9:34] Zha Ewry: From my end, the challange, would be having a charter and a set of rules which encouraged sensible IP handling
[9:34] Zha Ewry: and a clear sense of how governance would be done
[9:34] Zha Ewry: digs for the coprorpate master's comments about that
[9:36] Zha Ewry: [1]
[9:36] Latha Serevi: Anyone have other topics for today? Thanks for any help on the neutral-location-open-grid-spec topic, but perhaps we can better spend time on some detail stuff.
[9:36] Zha Ewry: quotes the company line, which, is very applicable in this context
[9:37] Zha Ewry: Naked UUIDs
[9:37] Morgaine Dinova: I agree, Lillie. I never understood why techies have any time at all for politics, since no politician ever created anything. I guess it's low self-esteem.
[9:37] FWord Utorid: those are big words. ibm writes a lot
[9:37] Zha Ewry: Pretty much says "Be well run be transparent, be public"
[9:37] Morgaine Dinova: Is that in answer to the ISO mess?
[9:38] Zha Ewry: shrugs
[9:38] Latha Serevi: Sounds OK, Zha. How does the current LL OGP process fare by those standards?
[9:38] Saijanai: Kuhn: bah, AL Gore created teh internet Vint Cerf says so
[9:38] Latha Serevi: Sai--
[9:38] Zha Ewry: I think, with no certainty? That it is in response to the ISO stuff, some stuff at OASISS, some messes at OMG a few years ago
[9:38] Zha Ewry: And...
[9:38] Morgaine Dinova: And Stephens hooked up the tubes
[9:38] Zha Ewry: Linden does *pretty* well at transparent
[9:38] Zha Ewry: But
[9:39] Zha Ewry: governance?
[9:39] Zha Ewry: Not so much
[9:39] Zha Ewry: And sometrimes, it's just "Oh, we did this"
[9:39] Zha Ewry: but to be fair
[9:39] Zha Ewry: they aren't calling it a standards process
[9:39] Teravus Ousley: apparently the FED doesn't do so well at governance either..
[9:39] Teravus Ousley: .. sorry.. just had to add that in there. :D
[9:40] Lillie Yifu: I get enough politics with my paid work..
[9:40] Zha Ewry: chuckles and banishes politics to the outer realms of the meteverse
[9:40] Morgaine Dinova: Why would the Federal Reserve have anything to do with governance. It's a private bank, not regulated by anyone.
[9:40] Latha Serevi: Morgaine, for shame.
[9:40] Saijanai Kuhn: starts snoring. safer.
[9:40] FWord Utorid: we need politics interop, where politics connects to other politics, across protocols. oh, wait... we're doing that already
[9:41] Zha Ewry: chuckles and banishes politics to the outer realms of the meteverse
[9:41] Morgaine Dinova: It's like expecting a vampire to be interested in governance.
[9:41] Latha Serevi: Topics please?
[9:41] Lillie Yifu: moves teh previous question.
[9:41] Zha Ewry: Naked UUIDs and when they are OK
[9:41] Teravus Ousley: err.. addressing space
[9:41] Zha Ewry: I've been trying to see if we can't keep that simple
[9:41] Morgaine Dinova: Good topic, Zha
[9:41] Zha Ewry: I think we can
[9:42] Lillie Yifu: So why would they ever be OK?
[9:42] Zha Ewry: Roughtly, I'd propose that any service is welcome to hand you a naked UUID
[9:42] Zha Ewry: if, and only if, they can hand your a URL to it, given the UUID, for the lifetime of the UUID
[9:42] Zha Ewry: So, if you want to toss one on a baked ave skin, and hand it to the client
[9:43] Zha Ewry: go to town
[9:43] Zha Ewry: But
[9:43] Zha Ewry: anywhere else
[9:43] Zha Ewry: where you can't meet that sort of "Its just for now" requirmeet
[9:43] Zha Ewry: you hand out a URL
[9:43] Zha Ewry: There are a *lot* of places where we only need to identify something in a narrow context
[9:44] Lillie Yifu: so can we say that a UUID implies a cap? How do I know when the "lifetiem of the UIID" is, if it is anaked uuid?
[9:44] Zha Ewry: But.. you're deeply in trouble if you let them escxape
[9:44] Zha Ewry: I think a UUID implies, there is a well konwn way to resolve them
[9:44] Zha Ewry: If its naked, you're saying I expect you to konw this object already, or be able to fetch it from my well konwn UUID resolver
[9:45] Lalinda Lovell: oo this pose isnt lady like
[9:45] Morgaine Dinova: Was there a specific danger that brought this issue of naked UUIDs to the fore?
[9:45] Dale Innis: is still hoping someone will put up a description of how caps in general are expected to work somwhere.
[9:45] Zha Ewry: Most beanbags are safe
[9:45] Zha Ewry: Ones without people in them ;-)
[9:45] Lalinda Lovell: wont let me sit
[9:45] FWord Utorid: here
[9:45] Goldie Katsu: Yeah I thought I'd rezzed all the way....
[9:46] Lalinda Lovell: sorry, nearly sat on your face fword
[9:46] FWord Utorid: did i complain about that?
[9:46] Zha Ewry: Caps.. heh.
[9:46] Lalinda Lovell: :P
[9:46] Zha Ewry: So... The spec says
[9:46] Zha Ewry: You get them from a seed cap, and they are stable
[9:46] Zha Ewry: Not much else
[9:47] Dale Innis: :) yeah
[9:47] Zha Ewry: I think *some* sane handling of lifetime would be intresting
[9:47] Lalinda Lovell: can i ask a timely question?
[9:47] Zha Ewry: For 90% of what we do, tho, it's either sort of forever (assets(
[9:47] Zha Ewry: or "While this thing exists in world"
[9:47] Zha Ewry: Sure Lalinda
[9:48] FWord Utorid: dude, elanthius isn't fat any more
[9:48] FWord Utorid: what aletdown
[9:48] Elanthius Flagstaff: Sorry guys
[9:48] Gwen Hermit: a yoi yoi hoy
[9:48] Lalinda Lovell: with world banks toppling and financial crises, are virtual worlds going to be hit too?
[9:48] Gwen Hermit: how late am i?
[9:48] Lalinda Lovell: opinions
[9:48] Anthonys Republic: yes
[9:48] FWord Utorid: everything will be hit
[9:48] Latha Serevi: Not at this meeting, please, Lalinda.
[9:48] Anthonys Republic: its already being hit
[9:48] Dale Innis: sure everything's going to be hit to some extent. Not really our job here t ho.
[9:48] Anthonys Republic: sry
[9:49] Gwen Hermit: banks and the government are interop'ing too much ;)
[9:49] FWord Utorid: eventually we will all turn into paste. but, the naked UUID, that's what I'm talking about.
[9:49] Lalinda Lovell: will people sell up and sell out
[9:49] Gwen Hermit: so.... what's teh OGPv4 gossip?
[9:49] Saijanai: Kuhn: TP clarification is all I know about
[9:49] FWord Utorid: zha was trying to talk, and all of you landed on my head and destroyed the tranquility
[9:50] FWord Utorid: go, zha, don't let these interlopers dissuade you, your mission MUST CONTINUE.
[9:50] Rex Cronon: greetings
[9:50] Morgaine Dinova: SL economics are much simpler. There's no Federal Reserve to loan money to LL, they can just print it directly. :-) But on the positive side, there's no Fractional Reserve Banking to manufacture 9 x the printed amount out of thin air.
[9:51] FWord Utorid: Morgaine, SL economics are largely people's entertainment budget, which gets hit in time of crisis. And now, Back to Zha
[9:51] Zha Ewry: Trying to sort out if the economy willi mpact virtual worlds is pretty far out of scope, for the AWG. My guess is "of course, there will be some impact, less spare cash can't be wondrous for any non essential business"
[9:51] Zha Ewry: nods
[9:51] Saijanai: Kuhn: The banking system is like musical chairs, nobody loses until the music stops
[9:51] Zha Ewry: what FW said
[9:51] Zha Ewry: So...
[9:51] Zha Ewry: In general
[9:51] Zha Ewry: if you get handed a naked UUID?
[9:51] Zha Ewry: Passing it onward is not cool
[9:52] FWord Utorid: (queue morgaine's counter argument)
[9:52] Zha Ewry: Unless you want to take on the resolver role (and you don't)
[9:52] Morgaine Dinova: Zha: more spare cash with nothing behind it is just a liability, so let's be grateful if there's less spare cash. :-)
[9:52] FWord Utorid: (zha attempts to return to topic)
[9:53] Zha Ewry: My basic sense, after thinking about UUIDs for a chunk of time, is that, you don't ever pass one on, becasuse you are implicitly saying "I know how to find a URL to this UUID" when you do so
[9:53] FWord Utorid: (fword mentions something obscene to pass the time)
[9:53] Gwen Hermit: i hear the topic is naked UUIDs, very simple to figure this one out
[9:53] Gwen Hermit: you don't use UUIDs as the initial locator
[9:53] Gwen Hermit: you use a URL
[9:53] Zha Ewry: and thereafter
[9:54] Zha Ewry: you can pass one along the pipe, as long as you're willing to handle resolving it
[9:54] Zha Ewry: You have to assume, that the recpient, always discarded the original URL if you handed themone, so.. you have to provide a resolver anyway
[9:54] Morgaine Dinova: Zha: if you never pass a UUID, are you saying that they are purely internal sim values, never used in s-s nor s-v communication?
[9:55] Gwen Hermit: i'd say internal sim values or caches are good
[9:55] Zha Ewry: Not quite
[9:55] Zha Ewry: You're saying, they are only usable, between the person creating the UUID, and its direct peers, and then, only as a shorthand
[9:55] Morgaine Dinova: Or maybe they're purely asset d/b keys.
[9:55] FWord Utorid: who just beamed me
[9:56] Zha Ewry: They really can't be, Morgain, once you have multiple
[9:56] Zha Ewry: places an asset could come from
[9:56] Morgaine Dinova: Zha: good point
[9:56] Zha Ewry: they arentt useful, in general
[9:57] Zha Ewry: But..
[9:57] Zha Ewry: They are wicked good in a pipeline between peers
[9:57] Gwen Hermit: once a URL has been used, you cache the UUID on the server it's used at
[9:57] Gwen Hermit: problem solved
[9:57] Zha Ewry: Not quite, Gewn
[9:57] FWord Utorid: notes Gewn as a nickname for Gwen
[9:58] Zha Ewry: you never konw if it got there reliably
[9:58] Zha Ewry: must be a bad UDP packet between my atomic self and SL self
[9:58] Gwen Hermit: Zha - checksums and other such things for reliability
[9:58] Zha Ewry: like gewn?
[9:58] Zha Ewry: sure
[9:58] FWord Utorid: I agree with Gewn, we need more numbers
[9:58] Morgaine Dinova: Ah, I think Zha is saying that naked UUIDs end up purely as shorthand identifiers that denote particular URIs only between communicating peers who know which URIs they're actually talking about.
[9:58] Zha Ewry: But, you never know if someoen got, ot kept a cached URL
[9:59] FWord Utorid: there should be a number 5 every time things are bad. it makes the 2012 people happy.
[9:59] Teravus Ousley: (hippos)
[10:00] Zha Ewry: (Aardvarks)
[10:00] Lalinda Lovell: aardvark was murdered - zebea did it
[10:00] Morgaine Dinova: So an UUID just ends up as a RLE-type symbol to optimize comms between peers. The symbol is interpreted by the peers as really meaning the URI.
[10:00] Lalinda Lovell: zebra*
[10:00] FWord Utorid: (cue random segment)
[10:01] Goldie Katsu: (zehrva?)
[10:01] Zha Ewry: Pretty much, Morgaine, I think
[10:01] Zha Ewry: What I want to be able to do, is for us to all make sure that is what we want, then push on it a it, and bake it in, and start laying out the path from "lots of UUIDs" to
[10:01] Zha Ewry: "We only use naked UUIDs when we can safely"
[10:02] FWord Utorid: we should just have double uuids
[10:02] Morgaine Dinova: So handing out a naked UUID loses data. The recipient then has to scratch her head and ask "With respect to what namespace?"
[10:02] Gwen Hermit: fword - quit trolling :)
[10:02] FWord Utorid: one uuid is the originator / pwner, the other is the actual uuid
[10:02] FWord Utorid: [derisive comments directed at Gewn]
[10:02] Zha Ewry: At that point, I think, most of thr time, yo ucould have the URL
[10:03] FWord Utorid: I come to all of these meetings, Gewn. You're just an outsider, here to hit on women and say techno-sounding things.
[10:03] Lalinda Lovell: gewn n00b
[10:03] Gwen Hermit: fword - you're talking yourself out of a region ;)
[10:03] FWord Utorid: but the question comes down to, if there is an identifier for the originator of a uuid, then one wouldn't have to worry about collisions
[10:04] Gwen Hermit: besides the trolling, that "2 UUIDs" thing could make sense.... though prefixes would be even better
[10:04] Zha Ewry: You really don't need to much, anyway, if you follow the spec
[10:04] Dale Innis: Okay so, a naked UUID is bad, because it's only a unique identifier within a particular namespace.
[10:04] Gwen Hermit: a standard /64 prefix
[10:04] Teravus Ousley: I suppose that's where the URL comes into play
[10:04] Dale Innis: URLs are good, because they're unique.
[10:04] Gwen Hermit: dale - they can also change
[10:04] FWord Utorid: Ahah, so Gwen does concur with my concept in spite of my having masked it inside of a nonsense parfait
[10:04] Teravus Ousley: UUID collision chances are small if you trust everyone to play nice
[10:05] Lalinda Lovell: ter - its not collisions thats the problem
[10:05] Dale Innis: Change as in, point (or fail to point) to different things at different times?
[10:05] Zha Ewry: And, fatal if someone doesn't.
[10:05] Lalinda Lovell: its identifing where it belongs
[10:05] Gwen Hermit: IANA to the rescue!
[10:05] Gwen Hermit: cut the address space up
[10:05] Teravus Ousley: Nuuuuuuu!
[10:05] Gwen Hermit: heh, we already have a system for that
[10:05] Gwen Hermit: IPv6
[10:06] Goldie Katsu: Would a UUID collision be a problem if its reference is mapped through a URL?
[10:06] Lalinda Lovell: yes
[10:06] Goldie Katsu: I guess the question is where is the naked UUID used.
[10:06] FWord Utorid: teravus, it's possible that if we invent a quantum fissue, we could have infinite numbers of UUID collisions. Also, Gwen was being mean to me
[10:06] Teravus Ousley: no, that's what we're pointing out really.. you could keep a copy of the actual URL on the server.. and send the client a random UUID
[10:06] FWord Utorid: *fissure
[10:06] Tao Takashi: Hi
[10:07] Teravus Ousley: Hi Mr Tao!
[10:07] Rex Cronon: hi
[10:07] Lalinda Lovell: tao your title is spam
[10:07] Zha Ewry: Shouldn't work
[10:07] FWord Utorid: if there was a wormhole, every possible UUID could collide. But the question is, what would happen if there was a UUID collision, actually?
[10:07] Tao Takashi: sure, I like spam :)
[10:07] Morgaine Dinova: Dale: except for extremely rare collisions, UUIDs actually are unique, but uniqueness isn't the issue. The point is that just being unique doesn't make it usable as an address, because there is no universal index of UUIDs to values (nor can there be). So it's the lookupability of a UUID in a given namespace that's the real issue with naked UUIDs.
[10:07] Gwen Hermit: my name is also spam ;)
[10:07] Zha Ewry: exactly, Morgaine
[10:07] FWord Utorid: I have no last name any more
[10:07] Tao Takashi: as is this program's title
[10:07] Gwen Hermit: morgaine - i like the idea of using a DNS-like system
[10:08] Teravus Ousley: and.. one solution to that is to take the UUID in context of an address
[10:08] Gwen Hermit: someone RFC this bitch and get an IANA allocation
[10:08] FWord Utorid: but the core question didn't get addressed, what if there are two UUID clones? does it matter? if so, how?
[10:08] Gwen Hermit: or..... rather than prefixes, just a UUID for the domain
[10:08] Teravus Ousley: isn't too keen on having a central registry for UUIDs
[10:08] Zha Ewry: Central registrries don't web scale in general
[10:08] Gwen Hermit: zha - DNS works
[10:08] Goldie Katsu: It shouldn't if they aren't on the same domain, right? (As in they aren't naked in the same place)
[10:09] Tao Takashi: just a quick summary: what do we need naked UUIds for? :)
[10:09] Gwen Hermit: ok, how about a simple /64 split?
[10:09] Zha Ewry: Well
[10:09] Tao Takashi: would otherwise simply propose to forget about UUIDs and use URLs instead
[10:09] Zha Ewry: one) we want to be as webish as possible
[10:09] Gwen Hermit: the first /64 identifies where the f**k it lives, the last /64 identifies wtf it is
[10:09] FWord Utorid: tao you are standing on my boobs
[10:09] Tao Takashi: UUIDs mighth be sort of an implementation detail then
[10:09] FWord Utorid: dale, satan stole my last name
[10:09] Zha Ewry: URLS are the way we do IDS, not UUIDs
[10:09] Tao Takashi: I am sorry
[10:10] Latha Serevi: UUIDs = painful form of data compression?
[10:10] FWord Utorid: it's ok, tao. i forgive you
[10:10] Gwen Hermit: i think we need UUIDs and caches etc
[10:10] Morgaine Dinova: Gwen: no semantics in UUIDs, that breaks the concept of randomness.
[10:10] Zha Ewry: Second, we *may* be able to justify that the shortcut is useful
[10:10] Gwen Hermit: morgaine - good point
[10:10] Dale Innis: Let's just do URLs, and see if we ever wish we had UUIDs.
[10:10] Goldie Katsu: wonders if that beanbag might benefit from being moved as it seems to be a popular lm that people tp in
[10:10] FWord Utorid: ok. so the double uuid is the win. much like the session id, and the agent id
[10:10] Gwen Hermit: so.... first time a remote asset is imported or accessed, UUID is cached
[10:10] Tao Takashi: we won't be webby in the end anyway.. ;-)
[10:11] Gwen Hermit: i read that "webby" as web.py :(
[10:11] FWord Utorid: tao, i disagree. in the end, everything will be webby
[10:11] Gwen Hermit: man web.py is a sexual framework
[10:11] Tao Takashi: well, then we have two worlds of webbiness ;-)
[10:11] Goldie Katsu: so how is the shortcut useful.
[10:11] Gwen Hermit: 1 - paste URL to texture into your build tool
[10:11] Gwen Hermit: 2 - simulator fetches
[10:12] Morgaine Dinova: OGP isn't webby. It just rides on top of a very efficient infrastructure used by the web.
[10:12] Gwen Hermit: 3 - simulator stores UUID and URL together, possibly caching the actual texture
[10:12] Gwen Hermit: 4 - ???
[10:12] Gwen Hermit: 5 - profit?
[10:12] Tao Takashi: what's the UUID good for then?
[10:12] Lalinda Lovell: 4 - make money
[10:12] FWord Utorid: OGP is 10 steps away from the WebWorldWonder we will all be absorbed by
[10:12] Gwen Hermit: Tao - to avoid sending the whole URL to clients, and getting clients to keep fetching it
[10:13] Gwen Hermit: the simulator or other services in the region domain acts as a proxy OR cache
[10:13] Tao Takashi: the sim can give the client another URL to fetch from
[10:13] Zha Ewry: The OGP stuff, is going to slowly become much more webby tho
[10:13] Goldie Katsu: I would guess that the UUID references a specific thing, but in most instances you need to know how to get to that thing
[10:13] Tao Takashi: if the original source should not be displayed
[10:13] Teravus Ousley: Is this not a vailed attempt to get around the fact that the viewer uses UUIDs only? :D
[10:13] Goldie Katsu: the url gives you directions
[10:13] Gwen Hermit: Goldie - say to the simulator or asset server "oy! gimme what this UUID refers to"
[10:13] FWord Utorid: [random madness segment]
[10:13] Zha Ewry: The thing that is kind of key, is that we have a ton of stuff to pass along the pipe to cloients
[10:14] Gwen Hermit: and asset server says "ok, here it is", or "here's the URL reidrect"
[10:14] Zha Ewry: We can alsways pass URLS
[10:14] Zha Ewry: (and compess)
[10:14] Goldie Katsu: which asset server
[10:14] Zha Ewry: But.. UUIDs aren't bad
[10:14] Zha Ewry: and WHICH is exactly right
[10:14] Goldie Katsu: LL or Goldiees awesome worlds
[10:14] Tao Takashi: well, the URL might be using a UUID but I see this as implementation specific
[10:14] Gwen Hermit: Goldie - your agent/region domain's local asset server
[10:14] FWord Utorid: eventually http servers will be asset servers too
[10:14] Gwen Hermit: fword - you mean like now?
[10:14] FWord Utorid: and they will run on corn
[10:14] Gwen Hermit: ;)
[10:14] Goldie Katsu: but how does the sim know I'm rezzing from my asset server or ll asset server when I rez something?
[10:14] Zha Ewry: I don't know that there is one, at all, one single place to look
[10:15] Tao Takashi: I wish I could use my http server to serve textures for SL :)
[10:15] Goldie Katsu: I could have a unified view in my inventory that has both.
[10:15] Gwen Hermit: goldie - it checks it's table of known UUIDs
[10:15] FWord Utorid: tao, you can, if you rewrite everything
[10:15] Gwen Hermit: if you rez something you give it a URL
[10:15] Goldie Katsu: exactly
[10:15] Goldie Katsu: that's why a URL
[10:15] Tao Takashi: FWord: I would like more LL to implement an URL field for my texture tab :)
[10:15] Gwen Hermit: the UUID becomes in this case simpl a means to preserve the same content if the URL changes
[10:15] Gwen Hermit: if i send you to www.garethnelson.com/stuff/blabla today
[10:16] Morgaine Dinova: Zha: for AD+RD -> client comms, UUIDs are nowhere near an efficient "mutual shortcut" though.
[10:16] Gwen Hermit: and blabla is currently a small orange
[10:16] Gwen Hermit: tomorrow i might delete blabla
[10:16] FWord Utorid: tao, that would represent a great thing to do to really mess up people's viewers
[10:16] Gwen Hermit: and replace blabla with a pear
[10:16] Gwen Hermit: but it wouldn't matter if the orange had it's own UUID
[10:16] Gwen Hermit: the pear and the orange are unique
[10:16] FWord Utorid: if you could url link anything and make it load as a texture, you could probably crash sl
[10:16] Tao Takashi: why do URLs change?
[10:16] Gwen Hermit: tao - it happens
[10:16] Tao Takashi: sure, but that might mean that the AD moved to another host or so
[10:17] Zha Ewry: Keep in mind, you never know whether someone has kept your copy of a UUID/URL mapping,or. lost it, and you may not assume state is held
[10:17] Tao Takashi: then the UUID would also be needed to get fetched from the changed AD's URL.. (the object it refers to I mean)
[10:17] Goldie Katsu: I know why you want a UUID, I'm just suggesting that the URL is useful because it allows further clarification on where to get the asset on the initial fetch - rather than just passing a uuid
[10:17] Object: Hello,: Avatar!
[10:17] Gwen Hermit: both URL and UUID are useful
[10:17] Tao Takashi: as for the http images as textures: My browser does usually handle it quite well ;-)
[10:17] Dale Innis: Well, this is an important question; does the UUID always promise to always point to the "same thing"?
[10:17] Zha Ewry: I'marguing that, in general, a UUID naked, is not something safe to pass around
[10:18] Gwen Hermit: we've got a chance to KILL BROKEN LINKS AND SLAUGHTER THEIR FAMILIES here
[10:18] Zha Ewry: And.. UUIDs do promise that
[10:18] Dale Innis: but URLs don't?
[10:18] Zha Ewry: Not as such
[10:18] Tao Takashi: what do UUIDs promise?
[10:18] Zha Ewry: UUIDs don't change under you
[10:18] Zha Ewry: If you get a UUID to a texture, or a prim
[10:18] Gwen Hermit: but URLs might :) hence my point
[10:18] Zha Ewry: it always gets you that texture
[10:18] Morgaine Dinova: Tao actually has a point. URIs should never change, because by REST you're always talking to a proxy ("connector"), so what happens behind the scenes in the resource itself is hidden from the access by URI. Or at least it should be so.
[10:18] Latha Serevi: Zha - is your idea O
[10:18] Latha Serevi: typo
[10:18] Gwen Hermit: use URLs to pass around, use UUIDs to tell clients what to download
[10:19] Tao Takashi: well, a UUID is useless without a context so this context might still change
[10:19] Tao Takashi: and probably that context is identified by an URL
[10:19] Zha Ewry: A little more broadly than that, Gwen
[10:19] Gwen Hermit: client says "gimme <UUID>", simulator or asset server says what it knows
[10:19] Dale Innis: So a UUID in a given context always means the same thing?
[10:19] Gwen Hermit: Dale - indeed
[10:19] Zha Ewry: Actually, Dale
[10:19] Zha Ewry: Globally
[10:19] Teravus Ousley: actually client gets prim referencing UUID
[10:19] Zha Ewry: New assets get new UUIDs
[10:19] Tao Takashi: if a URL change make it return a 301 or 302
[10:19] Teravus Ousley: .. client asks Sim for UUID
[10:20] Teravus Ousley: so.. if you change the UUID that the prim is referencing.. the client will ask for a different one
[10:20] Tao Takashi: my point is simply, if you put context and UUID into one (because you need to in order to retrieve it) then you have a URL
[10:20] Teravus Ousley: .. but it'll keep the old one in cache under the first UUID.. and you'll have two image copies in cache.
[10:20] Zha Ewry: Well, the thing, is once you have the item in your cache
[10:21] Zha Ewry: a UUID is fine
[10:21] Zha Ewry: A sim can pass the UUID at will
[10:21] Tao Takashi: a URL as well
[10:21] Zha Ewry: and if the client has discarded it
[10:21] Zha Ewry: the sim has to respond to a "Get me the URL for this UUID"
[10:21] Zha Ewry: But.. UUIDs are somewhat terser
[10:21] Zha Ewry: I am.. 90% happy with pure URLs, but.. I htink
[10:21] FWord Utorid: ok, so there IS an identifier for the originator of the UUID in the works, I was right, even if Gwen is trying to blackmail me to behave
[10:22] Zha Ewry: that it's sort of icky
[10:22] Tao Takashi: I think using pure URLs makes it more understandable for people as it's one concept less to care about
[10:22] Tao Takashi: and people are used to URLs
[10:22] Zha Ewry: Tao, people don't need toundertsand this, mostly coders
[10:22] Tao Takashi: right, and every new concept is a problem
[10:22] Zha Ewry: We have tons of optimizaions in computer science for this reason
[10:22] Tao Takashi: I am talking about coders
[10:22] Tao Takashi: and they are usually lazy people ;-)
[10:22] Tao Takashi: except myself of course ;-)
[10:23] FWord Utorid: i vote for tao to do all the work
[10:23] Tao Takashi: but I am unreliable ;-)
[10:23] Morgaine Dinova: Does OGP currently support both URIs and UUIDs as object references, or just URIs?
[10:23] FWord Utorid: i vote for someone to fix tao's problems
[10:23] Tao Takashi: does OGP at this stage actually reference objects?
[10:23] Dale Innis: I don't thikn OGP *has* anyhting about object references in it yet.
[10:24] Zha Ewry: OGP barely touches on this yet
[10:24] Zha Ewry: except
[10:24] FWord Utorid: ok. so what is the best way to make this happen?
[10:24] Zha Ewry: ythere are lots of UUIDs in the teleport spec
[10:24] Dale Innis: write things on the WIki! :)
[10:24] Morgaine Dinova: Zha: are you involved in working that part of OGP out?
[10:24] Zha Ewry: I am
[10:24] Zha Ewry: currently asking
[10:24] Morgaine Dinova: Cool
[10:24] Zha Ewry: Infinity
[10:24] Zha Ewry: and Zero
[10:24] Zha Ewry: and everyone else
[10:24] Zha Ewry: questoins
[10:25] Zha Ewry: trying to pin down a possible sane story
[10:25] Zha Ewry: and a path from
[10:25] Zha Ewry: where we are today
[10:25] Zha Ewry: (100% UUIDs)
[10:25] Zha Ewry: to step by step
[10:25] Zha Ewry: only naked UUIDS in some cases
[10:25] FWord Utorid: so everything becomes a cap to get the actual UUID when needed?
[10:25] Gwen Hermit: [10:22
[10:22] Takashi: except: myself of course ;-) < i could mention a certain person who likes to do mad things in python...... and, do we want UUIDs at all when doing cross-grid stuff?
[10:26] Zha Ewry: We do not Tao, and..
[10:26] Zha Ewry: We have them in the OGP spec
[10:26] Teravus Ousley: UUIDS should be clothed in addresses.. not nikkid
[10:26] Tao Takashi: I see this standard not really easy to understand for web people .. caps and UUIDs and our own IM system etc.
[10:26] Zha Ewry: Excet possibly, when the context is onbvious
[10:27] Zha Ewry: Well, its not a standard, or close to yet, Tao
[10:27] Zha Ewry: and, the Caps, are the easy bit
[10:27] Tao Takashi: well, once it's finished of course
[10:27] Gwen Hermit: i really think we could do fine with UUIDs simply not existing in OGP outside of single simulators
[10:27] Tao Takashi: the caps are the hard bit ;-)
[10:27] Zha Ewry: And, I agrree
[10:27] Teravus Ousley: I don't think there should even ben the special case 'obvious'
[10:27] Lillie Yifu: that's like asking a math major what "trivial" means.
[10:27] Tao Takashi: to explain at least
[10:27] Morgaine Dinova: Tao: hehehe. Well as I keep saying, it's NOT the web. :-)
[10:27] Catherine Pfeffer: zha : I'm having a hard time with caps
[10:27] FWord Utorid: whatever gets us paid the fastest. I still haven't heard WHY UUIDs are bad and what would happen if there were collisions
[10:27] Tao Takashi: and to implement with normal web frameworks as well
[10:27] Teravus Ousley: special cases = more code
[10:27] Zha Ewry: nods
[10:27] Zha Ewry: If we realy think that, we shoudl make the case
[10:27] Dale Innis: There'sw nothing about our own IM system in the OGP spec I don't think.
[10:27] Tao Takashi: it's not the web but it would be nice if it would become one with the web (sort of)
[10:27] Zha Ewry: "NO NAKED UUIDs eever"
[10:27] Zha Ewry: ut
[10:27] Zha Ewry: *but
[10:27] FWord Utorid: does the database crash with an evil clone UUID?
[10:28] Zha Ewry: I think then, we will need to make the case carefully.
[10:28] Zha Ewry: And
[10:28] Zha Ewry: we will ikely have a migratory path anyway
[10:28] FWord Utorid: does the viewer crash with a clone UUID?
[10:28] Latha Serevi: Zha, I like your approach as a practical step. Move towards "you must promise to resolve your UUIDs on request". UUIDs are a relatively useful legacy data-compression technique that it would be painful to disallow.
[10:28] Tao Takashi: Dale: from what I hear from Zero it sounds like it will be the OGP IM system though because of all the integration with object transfers etc.
[10:28] FWord Utorid: do people steal things with a clone UUID?
[10:28] Tao Takashi: if I understood that right
[10:28] Zha Ewry: Zero, was sort of being unclear there
[10:28] Teravus Ousley: FWORD, no.. but it moves prims around with duplicate UUIDS.. I've seen it.
[10:28] Zha Ewry: He ponted out that its not just text to move
[10:28] Zha Ewry: but.. that is only a encapsulatoin issue
[10:29] Tao Takashi: right and I would move the rest by different web services
[10:29] Morgaine Dinova: FWord: it's nothing to do with collisions (although they would add trouble). It's to do with what can be used to look things up in a multi-world network. UUIDs can't be used for that, even when they're unique.
[10:29] Tao Takashi: if possible that is
[10:29] Zha Ewry: Well, you want to be able to refernce things in IMs
[10:29] Tao Takashi: he knows more about all this then I do :)
[10:29] Zha Ewry: "Hre is an asset for you"
[10:29] Gwen Hermit: clone UUIDs are an interesting one
[10:29] Gwen Hermit: basically spoofing
[10:29] Goldie Katsu: I guess I'd ask where are the naked UUID's used - or will be used in the future. On the SIM, in the Asset database and ...?
[10:29] Gwen Hermit: only solution is crypto and signing
[10:29] Teravus Ousley: Nuuuuuuu!
[10:30] Tao Takashi: well, I can still use Tao's web service to give Tao a box
[10:30] Dale Innis: yeah, there's "naked" as in "without context" and "naked" as in "unencryup0ted".
[10:30] Zha Ewry: But, how do you tell tao's web service which box?
[10:30] Zha Ewry: URL in the text?
[10:30] Tao Takashi: I really need to redo that agent domain implementation to have something to experiment with
[10:30] Tao Takashi: why does it need to be in IM?
[10:30] Zha Ewry: So I can say 'Here is a box, << Special box>>
[10:31] Morgaine Dinova: IM will eventually carry "rich content", that's why.
[10:31] Zha Ewry: which we do all the time
[10:31] Teravus Ousley: they've already said they were going to decouple inventory transfers from IM for OGP
[10:31] Zha Ewry: Right
[10:31] Zha Ewry: so you don't move the object that way
[10:31] Zha Ewry: but you want to be able to refer to it
[10:31] FWord Utorid: [ now is the time on sprockets where we dance ]
[10:31] Morgaine Dinova: Yep
[10:31] Zha Ewry: And tich context in general
[10:31] Tao Takashi: well, you might use a URL ;-)
[10:31] Teravus Ousley: previously they were using the 'Binary Bucket' field of IMS
[10:32] Zha Ewry: You might well use a URL
[10:32] Zha Ewry: but you would like to do so in context, with some support in the messaging
[10:32] Zha Ewry: But again
[10:32] Zha Ewry: that is just encapsulation
[10:32] Tao Takashi: yes, the client might need to understand it and display it appropriately
[10:32] Zha Ewry: so I am not 100% Zero is right to put that on IMs
[10:32] Saijanai: Kuhn: that's a text-to-speech generator, Teravus?
[10:32] Zha Ewry: Since they just transport stuff
[10:32] FWord Utorid: lol i asked teravus that last time
[10:32] FWord Utorid: i think teravus sounds like a robot irl
[10:32] Morgaine Dinova: Just like right now, IM delivers a "pseudo object" when it carries URLs (although it's a cheat, because the client just interprets the text and forms a link). Later, it'll carry real references to any VW entity.
[10:32] FWord Utorid: or IS a robot
[10:33] Bartholomew Kleiber: LOL stop it, laughing me butt off
[10:33] Tao Takashi: so should I have time I would like to get this AD going again (maybe with opensim backend support) and some example client to experiment with such things
[10:33] Gwen Hermit: IMs can carry binary buckets and set mime types
[10:33] Tao Takashi: that cheat is ok, IRC does nothing different if you paste a URL, it does not send the whole web page to you ;-)
[10:33] FWord Utorid: that's a lot of extra weight for something not needed
[10:33] FWord Utorid: 99.9% of IMs don't need the bucket
[10:34] Gwen Hermit: if it's not needed, it's not sent
[10:34] Tao Takashi: and if I am able to just install an IRC server or ejabberd instead of implementing my own when I want to write an AD I think that's a plus
[10:34] Saijanai: Kuhn: the extra zero isn't too big a deal, though the added complexity has been a pain to deal with, reverese-engineering-wise
[10:34] Zha Ewry: nods
[10:34] Zha Ewry: As I said, encapsulatoin
[10:34] Tao Takashi: maybe write some integration code for handling auth
[10:35] Saijanai: Kuhn: all GUI I/Oin the client is hardwired to the specific servers its dealing with: asset, IM, whatever
[10:35] Zha Ewry: SOmeone is using sound which will of ocurse, not be in the transcript
[10:35] Teravus Ousley: I think the client doesn't have the support for it built in right now.. just like it doesn't have inventory support built in right now
[10:35] Tao Takashi: I would like a rendering lib btw, which I can reuse in Python :)
[10:36] Morgaine Dinova: LL probably won't, because it does nothing it doesn't want to (like scaling regions). But third parties will have object references delivered in IM, guaranteed, because they've proved so useful in the chat of a zillion online games.
[10:36] Teravus Ousley: Tao, Poke Dahlia, she made a python version
[10:36] FWord Utorid: for those of you who just joined us, Teravus Ousley is using voice to say supportive, nurturing things to us. Now, back to our program.
[10:36] Saijanai: Kuhn: Teravus, the client has direct hooks to the various server/packets for both chat and inventory
[10:36] Saijanai: Kuhn: in the GUI
[10:36] Bartholomew Kleiber: lol
[10:36] Teravus Ousley: Saij: Actually.. it's using the legacy stuff..
[10:37] Zha Ewry: In general, toh, it's realy a bad way to handle it
[10:37] Saijanai: Kuhn: Dahlia did? I thought Gareth had put a Python wrapper around llvolume.cpp
[10:37] Teravus Ousley: Saij: the legacy stuff isn't running through the agaent domain
[10:37] Bartholomew Kleiber: no, she has a c# and python version
[10:37] Zha Ewry: You're dead right that it would be handy to pass a better fromed thing, encapsulated than replicate that
[10:37] Bartholomew Kleiber: of the meshmerizer
[10:37] Saijanai: Kuhn: TEravus, yes, but the client's GUI expects to see specific packets, regardless of how they are sent
[10:37] Gwen Hermit: sai - the LLVolume wrapper i was working on is tiny and pathetic
[10:38] FWord Utorid: so far it is the only llvolume wrapper ever mentioned
[10:38] Gwen Hermit: it's not suitable for much beyond rezing a cloud right now
[10:38] Teravus Ousley: Saij: My point is the agent domain currently does not get in the middle of IM
[10:38] FWord Utorid: in the history of sports
[10:38] Zha Ewry: So, btw, we have an added bit on the table, which is how to refer to assets in the large, without UUIDs
[10:38] Gwen Hermit: in fact, it's not suitable for ANYTHING beyond rezzing a cloud
[10:38] Tao Takashi: cool, will ask her :)
[10:38] Saijanai: Kuhn: gwen, yeah, but if it can handle at least some of the workload in native C++ that makes Morgaine's multi-process client possible. Even gets around the GPL/Apache licensing issue
[10:39] Morgaine Dinova: The IM interop interface between connected systems is going to need a content mapper, if LL refuses to let OGP carry structured IM content.
[10:39] Gwen Hermit: no it doesn't - it's tightly integrated with the viewer :)
[10:39] Saijanai Kuhn: goes back to obsessing about implementing legacy gorup IM as a test case for OGP
[10:39] FWord Utorid: [ begin heated debate segment ]
[10:40] Tao Takashi: probably goes back home now ;-)
[10:40] Zha Ewry: OK
[10:40] Morgaine Dinova: Hehe Sai. It's a good test case, useful too
[10:40] Tao Takashi: as I need some fooood
[10:40] Zha Ewry: Anyone have topics for next week?
[10:40] Saijanai: Kuhn: food?
[10:40] Saijanai: Kuhn: doughnuts
[10:40] Lalinda Lovell: can we chat slightly off topic now?
[10:40] Zha Ewry: thinks someone shoudl airship Saij some dunkin donutsgoodness
[10:40] Saijanai: Kuhn: later Tao. I'll try to finish the pyogp GUI and upload it to google

AW Groupies/Chat Logs/AWGroupies-2008-10-07

Navigation menu

Search