User:Zero Linden/Office Hours/2007 Nov 20

From Second Life Wiki

Second Life Wiki > Zero Linden/Office Hours/2007 Nov 20
Jump to: navigation, search
  • [13:12] Zero Linden: Thursday is Thanksgiving here in the U.S.A. and I'll be
  • [13:12] Zero Linden: baking pies
  • [13:12] Alexa Linden: woot!
  • [13:12] Kooky Jetaime: whoo hoo, what kinda
  • [13:12] Kooky Jetaime:  ?
  • [13:12] Zero Linden: So - no office hours
  • [13:12] Saijanai Kuhn: wants pecan
  • [13:12] Kooky Jetaime: I might just make a cake
  • [13:12] Kooky Jetaime:  :D
  • [13:12] Alexa Linden: wants carrot cake
  • [13:12] Tree Kyomoon: /whoa! fully bully!
  • [13:12] Tao Takashi: I make nothing ;-)
  • [13:13] Tao Takashi: probably I simply go to work
  • [13:13] Kooky Jetaime: Chocolate Cake, w/ Strawberries on top
  • [13:13] Zero Linden: pumpkin and probably apple/pear/ginger tart
  • [13:13] McPhenius Swain: oh, are we all invited?
  • [13:13] Saijanai Kuhn: welll, those are ok too
  • [13:13] Alexa Linden: supplies are limited... please order NOW
  • [13:13] Zero Linden: But the week after I want to move the Thursday meeting
  • [13:13] Morgaine Dinova: I'll bake a pie for synergy with you :-)
  • [13:13] Theodore Folsom: ^^
  • [13:13] Hypatia Callisto: No Thanksgiving where I'm at, but a happy thanksgiving to you all in the states :)
  • [13:13] Tree Kyomoon: I knew a ginger tart in highschool :)
  • [13:13] Zero Linden: your choices are: Wed @ 7:30am or Thur @ 8:30
  • [13:13] Zero Linden: I can't do Thur @ 7:30am any more
  • [13:13] Zero Linden: SO -
  • [13:13] Alexa Linden: Booos at Tree
  • [13:13] Zero Linden: vote on the wiki
  • [13:14] Rex Cronon: so, choice of food is now a important tenchnical discussion:)
  • [13:14] Zero Linden: [1]
  • [13:14] Saijanai Kuhn: "now"
  • [13:14] Squirrel Wood: both are too early for me. I'll be still at work or on me way home
  • [13:14] Saijanai Kuhn:  ?
  • [13:14] Zha Ewry: Ooooh How democratic
  • [13:14] Zero Linden: Okay- moving on
  • [13:14] Zha Ewry: of course, we need a REST style wiki ;-)
  • [13:15] Zero Linden: groans
  • [13:15] j3rry Paine: give it a rest
  • [13:15] Tao Takashi: we still need a faster wiki ;-)
  • [13:15] Morgaine Dinova: We have one already. It RESTs a lot ;-)
  • [13:15] Strife Onizuka: <.< it's not my fault
  • [13:16] Saijanai Kuhn: zero is waiting patiently...
  • [13:16] Zero Linden: If you don't all GET off the puns, I'll DELETE this whole structure, PUT you off my parcel, and write a cranky blog POST as I HEAD off to the pub
  • [13:16] Zero Linden:  :-)
  • [13:16] Morgaine Dinova: Hehehe
  • [13:16] Alexa Linden: LOL
  • [13:17] Alexa Linden: tyrant!
  • [13:17] Saijanai Kuhn: or not-so pateintly
  • [13:17] Ahzzmandius Werribee: shakes her head in pain.
  • [13:17] Zero Linden: That's me!
  • [13:17] j3rry Paine: zero should be shot
  • [13:17] Zha Ewry: winces
  • [13:17] McPhenius Swain: I'm still on the pub part :)
  • [13:17] Wyn Galbraith: voted.
  • [13:17] Ahzzmandius Werribee: please, no more punishment!
  • [13:17] Zero Linden: So, I'd like to discuss a "deliverable"
  • [13:17] Alexa Linden: ack
  • [13:18] Zha Ewry: Deliverable. Cool.
  • [13:18] Zero Linden: I think that we should aim at being able to demonstrate functionality, even if the protocols are highly incomplete, and still subject to change
  • [13:19] Zero Linden: That way we can test our designs as they are being done
  • [13:19] Ahzzmandius Werribee: that would be a good idea.
  • [13:19] Zero Linden: Here is something that I think would make a fantastic milestone
  • [13:19] Zha Ewry: Aye
  • [13:20] Rex Cronon: have the protocols even been started?
  • [13:20] Strife Onizuka: gave you Snapshot : Zero's Ground, Grasmere (181, 120, 26).
  • [13:20] Rex Cronon: there is no protocol even for log-in, log-out
  • [13:20] Zero Linden: Take an avatar that is hosted on a LL agent domain and log into an OPenSim region
  • [13:20] Hypatia Callisto: perks
  • [13:20] Zero Linden: Now, I expect that at that poing that is all it does:
  • [13:21] Zero Linden: I don't have my inventory
  • [13:21] Zero Linden: I may not even have my AV shape
  • [13:21] Zero Linden: or clothes
  • [13:21] Zero Linden: But I can appear in a region run by, say, OpenSim
  • [13:21] j3rry Paine: beauty is only skin deep
  • [13:21] Zero Linden: and walk around - with my "correct" name
  • [13:21] Tao Takashi: Protocol ideas have started ;-) I am still wondering though if I should get with my own caps server or if the LL one is released soon.. but hardly time anyway.. (sadly)
  • [13:22] Zero Linden: Oh- feh - thanks for reminding me-
  • [13:22] Finrod Meriman: is anyone here able to talk for opensim... are they ready for this?
  • [13:22] Zero Linden: I'll kick off that process (open sourcing the caps server) today internally
  • [13:22] Baba Yamamoto: they are ready
  • [13:22] j3rry Paine: hot sweaty and *totally* ready
  • [13:22] Zha Ewry: Defintely ready
  • [13:22] Zero Linden: And, I'm happy for it to be other folks besides OpenSim, if anyone else is ready
  • [13:22] Zha Ewry: Tho.. even taht. I think.. Is ilkely to run into UUID name space issues
  • [13:23] Zero Linden: Now - I imagine that we get login working, and NOTHING else - just enough to get into the region, get the currently existing protocol pieces and off you go
  • [13:23] Tao Takashi: great, Zero :-)
  • [13:23] Zero Linden: But that would be a big win
  • [13:23] Tao Takashi: sounds like a good plan :-)
  • [13:23] Zero Linden: I want to get a dev grid here at LL set up and committed to this,
  • [13:24] Tao Takashi: for a start I would go with my own agent domain though ;-)
  • [13:24] Zero Linden: Well- I expect that OpenSim would implement both sides - and use their own for development, as exactly would LL
  • [13:24] Zero Linden: but the proof is in the interoperability
  • [13:25] Tao Takashi: well, should I start working on the protocol implementation-wise I would probably do it on my own first on a non-graphics basis
  • [13:25] Tao Takashi: to get a feeling for it
  • [13:25] Tao Takashi: so starting off with a basic client and a basic agent domain
  • [13:25] Tao Takashi: you need to implement your agent domain yourself anyway I guess
  • [13:27] Zero Linden: yes, we do
  • [13:27] Zero Linden: though we can build on the existing agent inventory databases and our internal web services work for it
  • [13:27] Arawn Spitteler: wonders how, without other asset servers, the Agent would have so much as a name, and how redundancy is avoided with UUIDs.
  • [13:27] Tao Takashi: mine would start as dummy inventory probably
  • [13:28] Zha Ewry: You are going to habve UUID issues fromthe get go
  • [13:28] Zero Linden: well, name isn't actually an asset
  • [13:28] Ahzzmandius Werribee: couldn't UUIDs be assignable in a similar fashion to how ethernet MAC addresses are assigned?
  • [13:28] Zero Linden: and UUID collision is a non-issue
  • [13:28] Zha Ewry: Hmm?
  • [13:28] Zero Linden: UUIDs are (at least in so far asSL uses 'em) random 128 numbers
  • [13:28] Rob Linden: UUIDs traditionally use ethernet MACs as part of the generation algorithm
  • [13:29] Zero Linden: if done right - which isn't that hard - one needn't EVERY worry about collisions
  • [13:29] Rob Linden: odds of collision are EXTREMELY low
  • [13:29] Wyn Galbraith: checks that issue off the list.
  • [13:29] Ahzzmandius Werribee: well, you know what they said about 512k and memory. ;)
  • [13:29] Zero Linden: as in many times longer than heat death of the universe
  • [13:29] Arawn Spitteler: Extremely low equals eventual certainty.
  • [13:29] Baba Yamamoto: heh... the address space is VERY large.. i just wanted to use caps
  • [13:29] Tao Takashi: btw, I put that video of David Recordon up where he talks about openid and oauth
  • [13:29] Tao Takashi: oauth is indeed more like caps
  • [13:30] Strife Onizuka: microsoft had a collision of CLSIDs once
  • [13:30] Zero Linden: Indeed, oauth is SO much like caps, I'm not sure how it would fit in to AWG stuff
  • [13:30] Zero Linden: ah, but MS's CLSIDs are not the random form of UUIDs
  • [13:30] Tao Takashi: you might be logged in to two services with e.g. your openid or however and then you can use oauth to get a limited feature set without giving your password
  • [13:30] Strife Onizuka: nods
  • [13:31] Zero Linden: And, to be honest, we had a case where there were duplicate UUIDs being generated - but it turned to be a clear flaw in a python UUID library
  • [13:31] Tao Takashi: here is the video: [2]
  • [13:31] Ahzzmandius Werribee: well, either way, there are solutions to the UUID potential colision issue.
  • [13:31] Zero Linden: Tao - which is sort of exactly what Caps does
  • [13:31] Tao Takashi: yes, maybe we don't need oauth
  • [13:31] Tao Takashi: or maybe we don't need caps ;-)
  • [13:31] Ahzzmandius Werribee: it can be revisited later if necessary.
  • [13:31] Zero Linden: only I worry that lots of oauth is left to "agreement between parties"
  • [13:32] Tao Takashi: I guess we go with caps anyway as LL uses them already
  • [13:32] Zero Linden: which leaves it looking like a standard with too many options
  • [13:32] Zha Ewry: With two seperate systems generating UUIds aren't you likely to sooner or later have some issues?
  • [13:32] Tao Takashi: maybe, need to dig into it deeper. I definitely might need it for other projects
  • [13:32] Ahzzmandius Werribee: zha, no more than just one system with the same number of UUIDs generated.
  • [13:33] Ahzzmandius Werribee: assuming that the generator is truly random.
  • [13:33] Zero Linden: Zha - there is an RFC on their generation, and as long as you are all doing the same type (type 5, if I'm not mistaken)
  • [13:33] Zero Linden: then you are safe
  • [13:33] Zha Ewry: Depending on that.. across many systems.. strikes me as hinky
  • [13:33] Baba Yamamoto: Zha, even if every server in the world was generating a new uuid every milisecond it wouldn't be a problem
  • [13:34] Zero Linden: However, there might be cases where one would want to be super hyper paranoid - not against collision, but against errors in the UUID generator
  • [13:34] Teravus Ousley: notes that the possibility of that happening is very slim.. and in addition.. .. the main thing you'll see 'with assets in world' is .. them move on you when it's duplicate sends out an update.
  • [13:34] Ahzzmandius Werribee: zha, that was the idea behind the creation of UUID, a number that everyone can randomly pick and be reasonably assured that it's unique.
  • [13:34] Zha Ewry: I beileve, that well implemented ones will be ok
  • [13:34] Zha Ewry: I'm less convinced that arbitrary ones will be
  • [13:34] Zero Linden: So, for example, when creating the UUID of an identity - sure, perhaps that is vetted on a global scale
  • [13:34] Zero Linden: but UUIDs for caps - no point, as they are scoped by the cap server anyway
  • [13:34] Baba Yamamoto: zha, uuid uniqueness should be a local problem rather than a global one
  • [13:35] Teravus Ousley: has seen it in openSim.. restoring an exact copy of a region.. UUIDs and all to multiple regions..
  • [13:35] Morgaine Dinova: We're riding on the shoulders of mathematicians here --- do NOT invent your own randomness generators. You'll get it wrong.
  • [13:35] Zero Linden: So, yes, one must be sure to use UUIDs correctly -
  • [13:35] Zero Linden: but you needn't worry about collisions from generation
  • [13:36] Zero Linden: For that kind of case
  • [13:36] Zero Linden: our server code has an option to generate all new UUIDs for objects on load -
  • [13:38] Gibson Willis: Doesn't the first U stand for Universally?
  • [13:38] Zero Linden: Yes - it does
  • [13:38] Zha Ewry: OK
  • [13:38] Zero Linden: [3]
  • [13:38] Zha Ewry: I'm willing to take that as given, then. Yeah.
  • [13:38] Zha Ewry: Of course.. you'll still have to be insanlye careful abotu what to do with a UUID, once it enters someone elses domain, until you have it nicely marked as where it came from
  • [13:39] Zero Linden: "In other words, only after generating 1 billion UUIDs every second for the next 100 years, the probability of creating just one duplicate would be about 50%."
  • [13:39] Wyn Galbraith: would hit that.
  • [13:39] Arawn Spitteler: would have many such days
  • [13:40] Rob Linden: hopes we don't spend the rest of the time on UUIDs
  • [13:40] Ahzzmandius Werribee: wait til we hit 1 billion internet users. :-)
  • [13:40] Morgaine Dinova: Zha: no you won't. By definition your cap from one system will be unusable anywhere else ... the population of objects in VWs is inifinitely sparse, from a UUID's PoV
  • [13:40] Kooky Jetaime: naa, when you run the risk of duplicates, you just add a bit
  • [13:40] Zha Ewry: UUID no
  • [13:40] Kooky Jetaime: or a byte
  • [13:40] Zha Ewry: Cap yes
  • [13:40] Zha Ewry: A cap.. has the URL to tel l you where to go with it
  • [13:40] Kooky Jetaime: can be a source byte
  • [13:40] Zha Ewry: a bare UUID.. you don't know what to do with
  • [13:41] Zero Linden: True -
  • [13:41] Baba Yamamoto: ahzzmandius, not every UUID will be a persistent reference
  • [13:41] Morgaine Dinova: The bare UUID is unusable anywhere else ... it will be a capability to nothing.
  • [13:41] Ahzzmandius Werribee: true
  • [13:41] Zero Linden: there must always be a context with a bare UUID (whcih is why I like URLs better)
  • [13:41] Arawn Spitteler: A bare UUID would require a centralized data server; sure recipe for overload
  • [13:41] Zero Linden: but when needing to move info. about 100s of objects, UUIDs are a decent trade off between compactness and compactness
  • [13:42] Baba Yamamoto:  ;0
  • [13:42] Zero Linden: with needing only a minimal amount of context (which domain theycome from)
  • [13:42] Baba Yamamoto: compactness and compactness?
  • [13:42] Zero Linden: ha ha
  • [13:42] Zha Ewry: I'm not sure I see that.. Because..
  • [13:42] Kooky Jetaime: thinks that Zero needs this holiday vacatino
  • [13:42] Zero Linden: compactness and ease of identification
  • [13:42] Zha Ewry: domain alone.. implies that you can easily map from uuid to capabitlity.. when the URL is encoding a lot more than that
  • [13:42] Zero Linden: I do - with pies
  • [13:43] Wyn Galbraith: Mincemeat for me please ;)
  • [13:43] Tree Kyomoon: mmm pie decimation duty
  • [13:43] j3rry Paine: hair pie here
  • [13:43] Zha Ewry: unless all your caps take the form of https:/consistent-path/uuid/....
  • [13:43] Ahzzmandius Werribee: ok, so remind me again, what is the question that we are trying to answer here...
  • [13:44] Zero Linden: No, I think it is important tht clients never assume anything about the structure of a cap
  • [13:44] Zha Ewry: you can't konw from uuid and the fact that it came from lineden labs.. where to take the uuid to use it
  • [13:44] Zuffy Frua: scratchies his bum
  • [13:44] Zero Linden: so that some domains might make them all [4]
  • [13:44] Zero Linden: and others will be all [5]
  • [13:44] Squirrel Wood: pieology - the art of pie creation ^^
  • [13:45] Baba Yamamoto: is uuid unique enough or do we have to worry about collisions... the answer is.. not in the near future (100 years)
  • [13:45] Zero Linden: Right - so when we say that a UUID is used in a protocol
  • [13:45] Wyn Galbraith: LOL Zuffy has fleas.
  • [13:45] Saijanai Kuhn: the y2.1k error
  • [13:45] Zuffy Frua: i does not!!!
  • [13:45] Zero Linden: we need to make it clear how that UUID names an object
  • [13:45] Zuffy Frua: scratchies his bum
  • [13:45] Zuffy Frua: just an itch
  • [13:45] Zha Ewry: nods
  • [13:45] Zuffy Frua: whinnies
  • [13:45] Wyn Galbraith: teases.
  • [13:46] Zero Linden: So - we could say these messages all have agent UUIDs in them
  • [13:46] Zero Linden: some other protocol has established
  • [13:46] Zero Linden: but then we'd need to state that
  • [13:46] Zero Linden: the mapping of agent UUIDs to domains that they come from
  • [13:46] Zero Linden: though I could imagine, for example
  • [13:46] Zero Linden: a region giving a viewer updates about the position of various agents
  • [13:46] Zero Linden: identifying them by UUID
  • [13:47] Zha Ewry: You'd also be imposing some structure on all agent domains
  • [13:47] j3rry Paine: thie singularity is near. we don't have to worry about antyhng pas 2030
  • [13:47] Zero Linden: and only relaying the domain information upon request
  • [13:47] Zero Linden: yes, but after the singularity, computes will be able to re-write all our software to use 512 bit UUIDs - and make the cost of memory cheap enough to use 'em
  • [13:48] Zero Linden:  :-)
  • [13:48] Zuffy Frua: singulariwhatity?
  • [13:48] j3rry Paine: kurzweil
  • [13:48] Zero Linden: Zha - Well, I've assumed that if we know that an agent domain is, say, avies-are-us.com
  • [13:49] Tao Takashi: since I had dinner with David Orban I know what it means ;-)
  • [13:49] Wyn Galbraith: myavatarspace.com
  • [13:49] Zero Linden: that we can assume a set of public web services at [6]
  • [13:49] Zero Linden: (or some such)
  • [13:49] Baba Yamamoto: i'm tired of uuids I want to use HappyIDs
  • [13:49] Baba Yamamoto: what's next?
  • [13:49] Tao Takashi: BabaIds
  • [13:49] Zero Linden: and so, if In know that UUID xyz is from avies-are-us.com
  • [13:49] Squirrel Wood: i-has-an-avat.ar ?
  • [13:49] Baba Yamamoto: armenia?
  • [13:50] j3rry Paine: we be toys
  • [13:50] Wyn Galbraith: LOL
  • [13:50] Zha Ewry: That, I think.. is problematic scale wise
  • [13:50] Wyn Galbraith: Pets.
  • [13:50] Zero Linden: I can do something like: GET to [7]
  • [13:50] Baba Yamamoto: i would not trust .ar addresses ;)
  • [13:50] Tao Takashi: why can't xyz be just the account name?
  • [13:51] Teravus Ousley: less to compare with :D
  • [13:51] Zha Ewry: You're at that twitchy spot whee URIs for namingf and URIs for resource binding get messy
  • [13:51] Zero Linden: For unathenticated access to domain data, like any web service on the net, it needs a well known access point -
  • [13:51] Zero Linden: usually a domain name
  • [13:52] Zha Ewry: Right. but a capability. is one level re-directed away fromt hat
  • [13:52] Morgaine Dinova: Tao: tying account names to the DNS spec wouldn't be a good idea :-)
  • [13:52] Baba Yamamoto: i think we need a proof of concept ;)
  • [13:52] Zero Linden: Tao - I think we should, in the public protocols, deal only with avatar ids - account numbers or names, are an artifact of my relationship with my agent domain provider
  • [13:52] Sammy Grigges: hello friends
  • [13:53] Strife Onizuka: nods
  • [13:53] Zero Linden: Zha - yes, but there are some services that you need access to to kick it off
  • [13:53] Rex Cronon: hi
  • [13:53] Zero Linden: like login
  • [13:53] Strife Onizuka: KISS
  • [13:53] Zha Ewry: ASye
  • [13:53] Zuffy Frua: stomps his hoofy!
  • [13:53] Ahzzmandius Werribee: zero, why not use a second number that identifies the source of the UUID. like 4:128 bits.
  • [13:53] Zha Ewry: Aye
  • [13:53] Zha Ewry: Thos.. ones.. sure. They need to be public and stable
  • [13:53] Zero Linden: and I'm pretty sure people want to be able to write code that talks to domains to do "stuff" with the domain data
  • [13:53] Strife Onizuka: who assigns those extra bits?
  • [13:53] Zha Ewry: But.. when we pass a uuuid around
  • [13:53] Ahzzmandius Werribee: leave resolution of the source up to DNS based on the source id.
  • [13:54] Zha Ewry: are we always talking abotu the public ones
  • [13:54] Zero Linden: such as scrape and search and other apps - that have to do with things when you are not logged in
  • [13:54] Ahzzmandius Werribee: well, how are domains assigned? or ssl certs?
  • [13:54] Ahzzmandius Werribee: there's a central authority
  • [13:54] Ahzzmandius Werribee: that's how.
  • [13:54] Zero Linden: Ahzz - well, a second number requires allocation (IANA?)
  • [13:54] Zero Linden: easier to just use DNS
  • [13:54] j3rry Paine: maybe more than one
  • [13:54] j3rry Paine: (CA)
  • [13:55] Tao Takashi: I think I just dislike unredable URLs ;-)
  • [13:55] Ahzzmandius Werribee: zero, huh? I don't follow you on requiring assignment.
  • [13:55] Ahzzmandius Werribee: oh, nm.
  • [13:55] Ahzzmandius Werribee: LL could manageme assignment initially.
  • [13:55] Zero Linden: Yes, we could
  • [13:55] Ahzzmandius Werribee: but that would provide context for the UUIDs.
  • [13:55] Ahzzmandius Werribee: while not interfering in the established UUID generation methods.
  • [13:55] j3rry Paine: ahzzi's point earlier re 512: nobody thought we would run out of ip addresses eitherm, rut now we have dyndns and ipv6
  • [13:56] Zuffy Frua: yes ipv6s
  • [13:56] Tao Takashi: and I wonder if there is any difference in the protocol regarding this issue.. the main thing should be that nothing is colliding but this is the same with email addresses
  • [13:56] Ahzzmandius Werribee: j3rry, this extra ID doubles as source identification and expansion.
  • [13:56] j3rry Paine: scratch your bum
  • [13:56] Zuffy Frua: awwww
  • [13:56] Ahzzmandius Werribee: my concern is taht by using domains we've lost the compactness
  • [13:56] Ahzzmandius Werribee: so instead, we use a unique source ID.
  • [13:57] Ahzzmandius Werribee: that can be mapped to services via a dns-like system.
  • [13:57] Baba Yamamoto: domains should only be needed when you're asking for remote information ;0
  • [13:57] j3rry Paine: more thought, less action. but i like the way you talk
  • [13:57] Zero Linden: side note: the issues wasn't running out of IP addresses - it was that router memory and speed required the class based allocation method which made terrible use of the available address space
  • [13:57] Zha Ewry: Well. be careful. if you want to allow lots of assets servers.. with lots of flexabliity in how they get distirbuted.. you need to have the full capabiilty, not the uuid, or you need some shorthand for doin the same
  • [13:58] Zero Linden: Ahz - I don't think we should worry about compactness until we have a piece of protocol that is going to clearly be affected
  • [13:58] Zero Linden: for example
  • [13:58] Ahzzmandius Werribee: that's why I suggested the 32:128 bit pairing.
  • [13:58] Zero Linden: passing 40 full big URLs for the 40 avatars here
  • [13:58] Zha Ewry: compressing down towards UUIDs is short cuircuiting the basic structure of urls
  • [13:58] Ahzzmandius Werribee: k
  • [13:58] Zero Linden: is not a big loss over some 128+x bit compact number
  • [13:59] Zero Linden: but, once those long URLs have been established to the viewer
  • [13:59] Ahzzmandius Werribee: i was thinking of the processing time required to manage hundreds of thousands of urls in a single sim.
  • [13:59] Zero Linden: then talking about agents in the protocol by UUID is a good compromise
  • [13:59] Ahzzmandius Werribee: since we can't assume anymore that all avies, assets, etc come from a singular source.
  • [13:59] j3rry Paine: a 166mhz 386 forks 100,000 times a second
  • [13:59] Zero Linden: they are certanly going to be unique
  • [13:59] Zero Linden: so there is no "management" issues with, say "8-bit local ids)
  • [14:00] Zero Linden: and the viewer still can relate it to the "full id", which is the URL
  • [14:00] Ahzzmandius Werribee: *nod*
  • [14:00] Zha Ewry: You could have a nice simlpe mapping so that you pass the head of each cap in once.. and then.. just use the uuid afterwards
  • [14:00] Morgaine Dinova: Has anyone estimated how many objects there might be in a supergrid with the project numbers? I'm guessing that that would still be infinitely sparse compared to 128-bit space ... but it's just a hunch. We should do better than that.
  • [14:01] Strife Onizuka: thought, why not use session UUID's at the protocall level instead of global UUID's, this way you won't be suseptible to cross domain UUID spoofing
  • [14:01] Zero Linden: 60M regions * 15k objects * 20 crazy-we-will-have-denser-places-mutiplier
  • [14:01] Ahzzmandius Werribee: I don't see why we can't do both, full text human readable CAPS URLs and 32:128 IDs.
  • [14:01] Strife Onizuka: *cross domain UUID agent spoofing
  • [14:01] Ahzzmandius Werribee: the use of either depends on the computing requirements.
  • [14:01] Zero Linden: = 18*10^12
  • [14:01] Morgaine Dinova: 15k objects per region? LOL, that's certainly not the future.
  • [14:02] Zha Ewry: Dual encodings on the protocol level.. almost always devolve to people chosing one or the other, anyway
  • [14:02] Zero Linden: that was my *20
  • [14:02] Ahzzmandius Werribee: ok, so things are identified by domain+UUID
  • [14:02] Zero Linden: but key, do * 2000
  • [14:02] Kooky Jetaime: thinks of that saying.. "Thats the great thing about standards, there are so many to choose from.."
  • [14:02] Baba Yamamoto: I expect our simple prim object will be obsolete in the next 10 years ;)
  • [14:02] Zero Linden: so you get ~ 2 * 10^15
  • [14:03] Zha Ewry: Not domain+uuid, but prefix+uuid once.. and then uuid thereafter
  • [14:03] Zha Ewry: You need the full URL at least once
  • [14:03] Zero Linden: hmmm - we should put up an identifier page.... I'll add that to mylist
  • [14:03] Ahzzmandius Werribee: k
  • [14:04] Zero Linden: all righty - it is now 2pm
  • [14:04] Zero Linden: and I'm sure we all need to move on
  • [14:04] Ahzzmandius Werribee: yep
  • [14:04] Zero Linden: I will see you next week
  • [14:04] Morgaine Dinova: And UUIDs have a 3 x 10^38 space. So, our object space will indeed be effectively empty
  • [14:04] Zero Linden: same bat-time, same bat-parcel
  • [14:04] Squirrel Wood: ^^
  • [14:05] Hypatia Callisto: bye :)) and happy holidays
  • [14:05] Rob Linden: bye all
  • [14:05] Rex Cronon: bye zero, rob
  • [14:05] Jason Swain: Thank you Zero
  • [14:05] Jason Swain: Bye Rob
  • [14:05] Zha Ewry: SUper Zero
  • [14:05] Masaw Umaga: bye
  • [14:05] Kerry Giha: Thanks
  • [14:05] Squirrel Wood: hoppy howlydays indeed!
  • [14:05] Zha Ewry: Many happy pies
  • [14:05] Ahzzmandius Werribee: laters. :-)
  • [14:05] Zero Linden: Greetings to all of your families
  • [14:05] Theodore Folsom: byes Zero
  • [14:05] Morgaine Dinova: See you Zero, happy pie making :-)
  • [14:05] Kerry Giha: It is good to know we won't run out of UUIDs for quite some time
  • [14:05] Kerry Giha: hehe
  • [14:05] Wyn Galbraith: Thank you Zero, have a good turkey day.
  • [14:05] Zero Linden: later
Personal tools