User:Which Linden/Office Hours/2009 May 28

[11:03] Morgaine Dinova: Making coffee
[11:04] Dahlia Trimble: something is wrong with this seat
[11:04] Which Linden: guten morgen!
[11:05] Dahlia Trimble: hola
[11:05] Which Linden: ha ha yeah maybe there's an AO conflict or something
[11:06] Which Linden: or you got hydraulics
[11:06] Dahlia Trimble: lol
[11:07] Dahlia Trimble: I'm attempring to animate my avatar from my irc client ;)
[11:07] Morgaine Dinova: Hiya Which!
[11:07] Which Linden: good morning!
[11:07] Morgaine Dinova: Hi Sai!
[11:08] Morgaine Dinova: Dahlia, I assume you added the av controls to Idealist, right?
[11:08] Dahlia Trimble: no I'm using gridproxy and the 1.23 client
[11:08] Morgaine Dinova: Ah, injecting into the proxy, clever
[11:08] Which Linden: that proxy is cool stuff
[11:09] Dahlia Trimble: <3 gridproxy
[11:09] ATechwolf Foxclaw: returns
[11:09] ATechwolf Foxclaw: I need to find more info on that. Re:gridproxy and injection.
[11:09] Which Linden: so I don't have a topic for today, anyone want to suggest anything?
[11:09] Saijanai Kuhn: mutters something about pyogp GUI for injection and goes back to sleep
[11:10] Morgaine Dinova: Wow, it's Eddy. Nice to see you again :-)
[11:10] Eddy Stryker: what is going on?
[11:10] Saijanai Kuhn: hey Teach
[11:10] Eddy Stryker: did i login to a meeting?
[11:10] Saijanai Kuhn: Which is pontificating about something or another
[11:10] Morgaine Dinova: Which is just starting OH, no topic yet.
[11:10] Which Linden: we're just getting started with my office hours
[11:10] Which Linden: taking topic suggestions
[11:10] ATechwolf Foxclaw: Saijani, that would be nice. Handy for testing and hacking the av, like the new alpha textures.
[11:10] Dahlia Trimble: oh oh it's the gridproxy master samarai
[11:11] Morgaine Dinova: Well we could make the topic Cable Beach and how SL is going to factor out asset services :-)))
[11:11] Morgaine Dinova: ducks
[11:11] Eddy Stryker: :)
[11:11] Dahlia Trimble: lol
[11:11] ATechwolf Foxclaw: Cable Beach?
[11:11] Which Linden: what's cable beach?
[11:11] Eddy Stryker: which: are you going to metaverse U tomorrow?
[11:11] Dahlia Trimble: Eddy's asset server
[11:11] Morgaine Dinova: Over to Eddy :-)
[11:11] ATechwolf Foxclaw: groans as he get the bad ppun.
[11:12] Morgaine Dinova: Describe Cable Beach in a paragraph, Eddy:-)))
[11:13] Eddy Stryker: i'm going to give a short talk on cable beach at metaverse u tomorrow. but the quick summary is a service-based architecture for virtual worlds, where you define each component (3d simulation, messaging, assets, etc) as a service and each service can live in a separate trust domain
[11:13] Morgaine Dinova: Cool, nice one-liner.
[11:13] Morgaine Dinova: Where is Metaverse U?
[11:13] Which Linden: interesting
[11:13] Eddy Stryker: cable beach is more about defining the interactions and potential trust relations between the services rather than the services themselves, although i'll probably do some service definition as well just to get the ball rolling
[11:13] Eddy Stryker: morgaine: stanford
[11:14] Morgaine Dinova: In-world presence here, or in Opensim?>
[11:14] Dahlia Trimble: its in sl
[11:15] Saijanai Kuhn: mutters something about x-world media interop
[11:15] Morgaine Dinova: Having trouble finding it in Search
[11:15] Morgaine Dinova: I'll check events
[11:16] Eddy Stryker: last note on cable beach: the current progress is that you can go to a world server, enter your openid, authenticate with your identity provider, get redirected to an inventory server and do an oauth confirmation, and land back at the world server with a secondlife:/// uri to click on that is preauthenticated
[11:16] Which Linden: cable beach doesn't sound like it's really an "asset server" though, it's more of a service registration framework
[11:16] Which Linden: cool!
[11:16] Dahlia Trimble: well it was once an asset server :/
[11:16] Saijanai Kuhn: It sounds like its a different way to handle the Agent Domain authentication
[11:17] Saijanai Kuhn: is it overlapping or does it replace it completely?
[11:17] Which Linden: it sounds like it's taking the agent domain idea to its logical conclusion
[11:17] Dahlia Trimble: [1]
[11:17] Morgaine Dinova: We're talking about authenticating for asset access, not for region access.
[11:17] Morgaine Dinova: (To Sai)
[11:18] Which Linden: i.e. why not have agent domain, inventory domain, asset domain, physics domain, etc, right?
[11:18] Eddy Stryker: saijanai: right now it's an independent implementation, although i demoed it to infinity at the LL office the other day and we're working on ways to bring some of the concepts together
[11:18] Dahlia Trimble: parts of it are in place on osgrid
[11:18] Morgaine Dinova: Which: that's the idea. But when I tried to bring that up, Zero refused to talk about such factoring.
[11:19] Saijanai Kuhn: OK. I guess what I was really wondering is whether it was meant to provide authentication for initial login
[11:19] Youri Ashton: hello
[11:19] Morgaine Dinova: Hi Youri
[11:19] Youri Ashton: sorry, was still having dinner :p
[11:19] Saijanai Kuhn: which is all AD does right now in OGP
[11:19] Which Linden: hi youri
[11:19] Youri Ashton: hey Which :)
[11:19] Which Linden: brb gotta wash maple syrup off my hands
[11:19] Youri Ashton: lol
[11:20] Dahlia Trimble: waves hi at youri
[11:20] Youri Ashton: hih i :)
[11:20] Youri Ashton: ugh, laggy :p
[11:20] Eddy Stryker: the main difference that we identified between the two right now is who is in charge of specifying services. in OGP, someone requests to auth with an agent domain and the AD says "ok, here are the services i've selected for use with this grid". in cable beach, the user can optionally attach preferred services to their identity which can be overridden by the world they are entering
[11:21] Morgaine Dinova: I assume that it's both an OpenID provider and consumer, so that if the AD used OpenID then it would be able to use that?
[11:21] Saijanai Kuhn: that doesn't sound like too big a deal, as long as there's trust that can be established at connection time
[11:21] Eddy Stryker: i've been trying to brainstorm different ways to make OGP slightly more user-centric instead of grid-centric, but i'm going to start by documenting everything i have for cable beach and see where things evolve from there
[11:21] Morgaine Dinova: Cool
[11:21] Saijanai Kuhn: big an issue*
[11:22] Youri Ashton: I just got online and I saw Torley and Maurice @ Blue's office, was lucky i guess...
[11:22] Which Linden: so what is the service breakdown?
[11:22] Saijanai Kuhn: the hypergrid assumes all services/worlds are trusted automatically, and the AD assumes they are NOT. CableBeach seems to compromise in a way
[11:23] Morgaine Dinova: Er ...
[11:23] Youri Ashton: *mutes sound* that tune starts to get on my nerves o_0
[11:23] Youri Ashton: hey Flimsey
[11:23] Flimsey Freenote: hi :)
[11:23] Saijanai Kuhn: appears to be deaf
[11:24] Youri Ashton: lol, your closer to it Sai
[11:24] Morgaine Dinova: Hiya Flim :-)
[11:24] Dahlia Trimble: Im using the linux client so sound isnt working :(
[11:24] Flimsey Freenote: waves at Morgaine
[11:24] Youri Ashton: lol, ty
[11:24] ATechwolf Foxclaw: Use openal soft.
[11:24] Which Linden: offending object removed
[11:24] Saijanai Kuhn: that's strange. All sound input enabled, but can't hear a thing except the wind and typing sounds
[11:24] Youri Ashton: ty for returning that :p
[11:25] Which Linden: mushroom mushroom!
[11:25] Morgaine Dinova: Sounds is working fine in Linux, using OpenAL. I haven't used FMOD for many months
[11:25] Eddy Stryker: saijanai: yes. the idea is that worlds can choose to use untrusted services supplied by a user, or they can override and provide their own. once you break things down as far as "assets", "inventory", etc. you might decide there really isn't any danger in using an "untrusted" asset service. no more dangerous than fetching content off the web
[11:25] Morgaine Dinova: Yep, agree with that.
[11:26] Saijanai Kuhn: as long as the client doesn't do something totally funky with the data, I don't see it as an issue. BUt I could see security viloations (AD perspective-wise) creeping in if the client does certain things
[11:27] Morgaine Dinova: I expect the *VAST* majority of assets to be totally free of restrictions, for the simple reason that people are not going to put up with their things disappearing when they visit some other world.
[11:27] Which Linden: eddy, what's the list of services? I see asset, inventory, identity so far
[11:27] Saijanai Kuhn: I forsee a rather strange period of disappearing/reappearing assets as everyone adjusts to the new world order
[11:27] Eddy Stryker: the other thing that i think everyone has been toying with a bit (infinity, people on my team, and people from the opensim community have independently mentioned it) is a way to aggregate trust. in CB we'll probably define a trust service that acts like a broker for trust. so if a random service is given at login, you can check with your trust service (a trusted third party service) that will tell you if they know about that service and how trusted it is
[11:28] Youri Ashton: if i may interrupt, what is the topic? o_0
[11:28] Morgaine Dinova: Yeah, like in the early days of the web, when we had broken images everywhere. Then people realized that that's unworkable.
[11:28] Which Linden: youri we're talking about cable beach: [2]
[11:28] Dahlia Trimble: the topic is new world order ;)
[11:29] Which Linden: Morgaine: and now everyone hosts images themselves to prevent broken images
[11:29] Eddy Stryker: which, sorry. there is world and simulation node which are in the same trust domain. so you get one world service and lots of simulation node services which makes up a 3d world concept. then assets, inventory, trust is being worked on, and that's all i have right now. messaging will need to exist before IM and things like that work
[11:29] Eddy Stryker: and the rest can probably be vendor-specific
[11:29] Youri Ashton: lol, no idea what that is Which. so far i understand, its something about the asset sever. correct me please if im wrong
[11:29] Youri Ashton: @Dahlia: new world order? hmm... you really wanna go there :p
[11:30] Dahlia Trimble: giggles
[11:30] Which Linden: youri: check out the page, it's got a decent overview
[11:30] Morgaine Dinova: There also needs to be a much stronger user-emphasis, so that you can have you clothes on your PC and not in some grid's asset server, and be able to carry them into new worlds with you.
[11:30] Which Linden: hm yeh eddy, it seems that defining the services is a difficult problem in and of itself; I mean we could toss in messaging and group services, but those interact in complex ways
[11:31] Saijanai Kuhn: I can see a possible security issue. NOt sure if its valid though: if a client establishes 2-way communication with an outside server of somekind and funnels data back to them, it could be come a tool for asset theft ala copybots. Though, that can already be done just by hacking the GPL or openmv clients
[11:31] Youri Ashton: im slightly dislectic and i can not focus for long on large texts :p
[11:31] Youri Ashton: mostly if i need to rush read :p
[11:31] Eddy Stryker: morgaine: my gut feeling is that most people will be happy with storing their clothes on their [insert vendor here
[11:32] Which Linden: yeah definitely; I've been migrating from "Everything runs on my home server" to "colo" to "hosted" for my internet services
[11:32] Youri Ashton: hi latif
[11:32] Latif Khalifa: hello
[11:32] Eddy Stryker: which: the reason why service definition is so sparse is that you only really need to define services for things that interact across trust domains. maybe it's determined that it is too difficult to do a group service across trust domains, so no service is defined. each world just does their own thing behind the curtain and the client (and other worlds) are none the wiser
[11:32] Morgaine Dinova: Eddy: as long as the asset services (plural) selection mechanism is flexible, one can always have an additional asset service on one's ownworkstation.
[11:33] Youri Ashton: take a seat!
[11:33] Youri Ashton: and no, doesnt mean free chairs :p
[11:33] Eddy Stryker: which: you don't *need* to define any cable beach services to get a functional grid. it just means that everything runs LLGrid style, where the fun is behind the magic curtain of the simulator
[11:33] Which Linden: but aren't we tryingto move away from the simulator-as-arbiter-of-everything model?
[11:34] Which Linden: it has a lot of flaws
[11:34] Eddy Stryker: defining assets and inventory as services means pulling those things out from behind the curtain, sticking them on the web, and saying "hey amazon can run these things instead"
[11:34] Which Linden: not least of them that the poor simulator gets overloaded easily
[11:34] Eddy Stryker: yes definitely
[11:34] Zha Ewry: One hopes tha service deployment is seperate from service deployment
[11:34] Zha Ewry: defintiino
[11:34] Zha Ewry: let me try that in english
[11:34] Morgaine Dinova: Gotta agree Zha :-))))
[11:34] Eddy Stryker: CB is just working one piece at a time though. smaller pills to swallow
[11:34] Zha Ewry: "On hopes that servcie design is different from servcie deployment"
[11:35] Morgaine Dinova: watches the logic explode
[11:35] Eddy Stryker: yes
[11:35] Zha Ewry: The fact that it sits on the sim today as a deployment endpoint, should be 99% irrelevent
[11:35] Morgaine Dinova: Well that's pretty simple: service design needs to be a superset of service deployment.
[11:35] Zha Ewry: mutters some nasty words about how service to client messages get handled int he large in that case, and listen more
[11:36] Which Linden: heh
[11:36] Saijanai Kuhn: I thought that OGP was meant to searpate the services out anyway, at least fom the sim. Cable beach appears to be a meta service provider in a sense
[11:37] Saijanai Kuhn: an alternative to the Agent DOmain
[11:37] Morgaine Dinova: So, how are we going to get a list of asset services that can be selected into the AD?
[11:37] Morgaine Dinova: Asset and inventory services.
[11:37] Dahlia Trimble: I think it would be cool if multiple services could be combined and chosen by each user so several asset servers could be simultaneously supplying a single simulation
[11:38] Which Linden: several simulators, you mean?
[11:38] Zha Ewry: It alsmost has to end that way, sooner or later, Dahlia
[11:38] Zha Ewry: Both, actually
[11:38] Saijanai Kuhn: seems to me that that is a problem that the OGP model has to solve anyway. Cable Beach just brings it to the front, instead of deferring it until LL decides what to do next
[11:38] Dahlia Trimble: and several simulators too
[11:38] Eddy Stryker: morgaine: i suggested two ideas. 1) the AD knows about multiple services that can fill the same role, and it can send a list of services to the client and give the client a choice. 2) in the initial exchange, the client can send a list of suggested services and the AD can decide to use those or ignore the list
[11:38] Zha Ewry: You really want to be able to fetch assets from a range o possibls stores
[11:38] Morgaine Dinova: Indeed. A user needs to be able to tell the AD on login that she wants to use SL's asset service, and (say) OSgrid's, and (say) the asset service running on her own PC.
[11:39] Dahlia Trimble: IE, I could have a private asset server for some regions on osgrid, and also allow access to the public asset servers at the same time
[11:39] Zha Ewry: And more generally, you really want to not tie any type of item to a singel service provider
[11:39] Morgaine Dinova: All 3
[11:39] Morgaine Dinova: And in general, N
[11:39] Zha Ewry: (ie you really have to have URIs not UUIDs for most of the non local stuff)
[11:40] Saijanai Kuhn: I can see a potential for confusion though: what if you decide to use Service C to obtain an asset A, and someone malicious has defined that asset to be dfiferent than what the client expects?
[11:40] Eddy Stryker: that's one thing i punted on just so i could get code and documentation written instead of staying in the planning phase forever. but support for multiple (simultaneous) services absolutely has to happen
[11:40] Zha Ewry: As soon as you say "Oh, its one of these 4 services which will handle this UUID"
[11:40] Zha Ewry: you're broken any notion of scalability and adressability
[11:41] Morgaine Dinova: Yep, needs to be an arbitrary set
[11:41] Youri Ashton: so reading this all... is it right that LL is trying to get SL more open for everyone? asin more opensim style? like letting people create their own sims and/or offline worlds?
[11:41] Zha Ewry: Caps, get you about 1/3 of the way there...
[11:41] Which Linden: yeah....we have to nuke bare uuids in the long run, unless the uuids are in a context such that the full url is completely derivable
[11:41] Zha Ewry: But the current binding of UUID to implcit service providers isn't wondrous
[11:42] Zha Ewry: right, its fine for a Servcie to accept UUIDs, it issued, for short term use
[11:42] Zha Ewry: But.. any long term holding of a naked UUID is just... asking for pain
[11:42] Dahlia Trimble: say well the public asset server on osgrid is somewhat insecure in that it allows anyone to access it, but I'd like to set up regions that can display more valuable asssets that I would like to store on a private server, while still allowing visitors to access their inventories
[11:42] Morgaine Dinova: So, same question again ..... how are we going to get the ability to use multiple asset services into the AD?
[11:42] Zha Ewry: (Well if Inventory is nothign more than a set of URIs, then the ones the sim can't fetch simply get marked as such wen you use them
[11:42] Youri Ashton: hmm... so my offline sandbox idea can be a reality soon :D
[11:42] Which Linden: presumably the idea is that your inventory server contains links to multiple asset servers
[11:43] Youri Ashton: good news then :p
[11:43] Which Linden: what zha said
[11:43] Zha Ewry: Mind you
[11:43] Zha Ewry: There is a gotcha lurking in there
[11:43] Dahlia Trimble: why not still have a UUID and have a giant look-up table on the internet somewhere?
[11:43] Zha Ewry: which is at the moment.. permisinos sit in a funy place
[11:44] Zha Ewry: offers to sell Dahlia a supercomputer cloud to host the table
[11:44] Dahlia Trimble: I was thinking of using google for free ;)
[11:44] Mojito Sorbet: In the High-Availability world, they will often define multiple "instances" of a service, any one of which can handle a request for that service. Does that apply?
[11:44] Dahlia Trimble: and you can sell a cloud to google
[11:44] Zha Ewry: if you do URIs, to an extent
[11:44] Morgaine Dinova: Zha: do you have some sort of framework for allowing the AD to handle multiple asset and inventory providers? It's not in OGP, and when I tried to get Zero to talk about it, he refused. So how are you going to get that in there?
[11:44] Eddy Stryker: dahlia: i don't think it needs to be "on the internet somewhere". the simulator should have all of the knowledge it needs to map from UUID to URI
[11:45] Thoys Pan: uh oops
[11:45] Eddy Stryker: mojito: that's closer to load balancing. in this setup, each service is speaking the same protocol but serving up different sets of content
[11:45] Dahlia Trimble: well for multiple asset servers it may not be practical for the region to know all that
[11:46] Eddy Stryker: dahlia: why not?
[11:46] Zha Ewry: If you get away from UUIDs ro URIs, its explicit
[11:46] Dahlia Trimble: hypergrid visitors can come from anywhere
[11:46] Zha Ewry: or... and here is another trick you can play
[11:46] Eddy Stryker: zha: you going to rewrite LLUDP to use URIs in place of UUIDs?
[11:46] Mojito Sorbet: Well, somehow part of a UUID has to indicate the service that created it.
[11:46] Morgaine Dinova: The region HAS to know about it, since it has to take the assets from all over and distribute them to the people in the region.
[11:47] Zha Ewry: If the inventoory is unified, the inventory host can keep track of the asset server
[11:47] Dahlia Trimble: I dont think the region knows now, it asks the asset server
[11:47] Saijanai Kuhn: but that means that each asset server needs to be registered with the inventory host
[11:47] Eddy Stryker: regions already track tons of metadata attached to objects. just attach another uri to them that acts as the base uri when resolving content for that object. problem solved
[11:47] Which Linden: Eddy: I think the long term plan is to migrate away from UDP for the inventory-related services
[11:47] Zha Ewry: and.. If you don't movbe from UUID to URI, you're always stuck with odd implicit couplling
[11:47] Mojito Sorbet: Yes, registered.
[11:48] Mojito Sorbet: Putting URIs into UUIDs makes themn kinda big
[11:48] Eddy Stryker: which: i'm not talking about inventory though
[11:48] Morgaine Dinova: Dahlia: sure, the region doesn't know now, but we're thinking ahead. :-)
[11:48] Eddy Stryker: inventory is easy
[11:48] Latif Khalifa: Eddy, but you could have a single object requiring multiple asset servers... ie different textures on different faces
[11:48] Morgaine Dinova: Inventory can't be unified, unless you mean on a person's own PC.
[11:48] Dahlia Trimble: say the region has to track too much already
[11:48] Which Linden: Eddy: oh....then what else makes major use of uuids?
[11:48] Eddy Stryker: latif: is that a good idea?
[11:49] Eddy Stryker: if it's a good idea then ok, back to the drawing board. but i'm not sure that's a good idea
[11:49] Zha Ewry: I think its an inevtiable idea
[11:49] Eddy Stryker: which: the objects in-world right now
[11:49] Mojito Sorbet: 5 sides of a picture frame come from one place, the painting on the front comes from another
[11:49] Latif Khalifa: well i could acquire textures from different places... in my inv... so when creating an object i should not be restrained to texturing it from a single assset server
[11:49] Zha Ewry: So I can put a texture on a surface which is served up out of Akami
[11:49] Zha Ewry: while the prim isn't.
[11:50] Dahlia Trimble: say and who serves the "missing texture" texture?
[11:50] Zha Ewry: Hopefully the client.
[11:50] Saijanai Kuhn: And even if you DO want to keep assets in inventory tracked, if you decorate something with a new texture and bring it back into inventory, the asset server and inventory host needs to be able to handle the arbitrary handoffs
[11:50] Mojito Sorbet: Built into client?
[11:50] jessicalee Shamrock: hey
[11:50] Morgaine Dinova: Latif: yep, that's what we're saying ... each asset can be anywhere, and what clients will be doing is creating a mashup of elements from different places, proxied (or no!) via the region.
[11:50] Mojito Sorbet: That sounds good.
[11:51] Which Linden: Eddy: I see....I'd say that those uuids are, like Zha says, temporary implicit references
[11:51] Eddy Stryker: makes sense
[11:51] Mojito Sorbet: Have to be really careful not to store these temporary UUIDs for too long
[11:51] Zha Ewry: One very helpful thing, is that Regions dont' get naked UUIDs in this model as inputs
[11:51] Which Linden: presumably they woulldn't last past a session
[11:51] Zha Ewry: so.. if they ever got the asset, they ought to have a way of knowing whence it came from
[11:51] Saijanai Kuhn: the rule of thumb right now is that if it goes into inventory, its no longer got a temp reference
[11:52] Zha Ewry: Which means, if we sort inventory
[11:52] Zha Ewry: 90% of it gets hidden
[11:52] Saijanai Kuhn: i.e. the asset server knows about it
[11:52] Eddy Stryker: sorry to take off early but i need to keep working
[11:52] Mojito Sorbet: Ok, so the Region maps the UUID to the full internal representation, which includes a URI?
[11:52] Eddy Stryker: bye everyone
[11:52] Morgaine Dinova: Cya Eddy
[11:52] Saijanai Kuhn: take care Eddy
[11:52] Which Linden: ok eddy thanks for popping in!
[11:52] Dahlia Trimble: laterz Eddy :)
[11:52] Youri Ashton: bye eddy
[11:52] Flimsey Freenote: awesome stuff Eddy
[11:52] Latif Khalifa: see you eddy
[11:53] Youri Ashton: Thoys, like sitting on Zha's lap? ;)
[11:53] Zha Ewry: Fast departing eddy
[11:53] Saijanai Kuhn: I've heard of him, I think
[11:53] Zha Ewry: laughs
[11:53] Morgaine Dinova: So, what's the answer to my question? How do we get asset service diversity into the AD?
[11:53] Youri Ashton: i guess you have a fan, Zha ;)
[11:53] Latif Khalifa: lol
[11:53] Latif Khalifa: thoys how rude!
[11:53] Zha Ewry: Mostly into the Inventory servcie, actually Morgaine, I think. which lets the rest happen implciityly
[11:53] Which Linden: Morgaine: well, I feel like if the inventory service is just a list of links then it's pretty implicit
[11:54] Which Linden: jinx zha
[11:54] Thoys Pan: what?
[11:54] Youri Ashton: lol
[11:54] Zha Ewry: I expect I wasn't rezzed to Thorys
[11:54] Thoys Pan: i didnt see her yet
[11:54] Youri Ashton: you were sitting on Zha's lap, Thoys
[11:54] Thoys Pan: same for john
[11:54] Thoys Pan: first i sat on john
[11:54] Mojito Sorbet: I always stand to one side until all green dots are accounted for
[11:54] Youri Ashton: lol
[11:54] Youri Ashton: atleast you wont fall off ;)
[11:54] Youri Ashton: haha
[11:55] Which Linden: man, I remember when people couldn't sit on each other....then we introduced that ability either as a bug or a feature, can't remember which
[11:55] Dahlia Trimble: if it is a uuid instead of a url, then it could move to a different asset server
[11:55] jessicalee Shamrock: whats up
[11:55] Zha Ewry: For about 90% of the early going, if you get inventory to allow you to point to mujltiple asset servers, the world is helped a lot
[11:55] ATechwolf Foxclaw: I like to have folk be able to set on an avatar saddle.
[11:55] Youri Ashton: hey jessicalee
[11:55] Which Linden: it could move to a different server if it's a url as well
[11:55] Morgaine Dinova: Pity the server doesn't enforce freezing an av until its surroundings have appeared.
[11:55] Mojito Sorbet: Then we would all pile up at the TP poiint, waiting for rezzing to finish.
[11:56] Latif Khalifa: they would risk being frozed forever ;)
[11:56] Mojito Sorbet: Which can take looong time sometimes
[11:56] Which Linden: Morgaine: SL would be a pretty frozen experience if it did that
[11:56] Zha Ewry: gently suggests URI vs URL in 90% of the spec level cases, even tho many willbe URLs
[11:56] Dahlia Trimble: I meant a different domain
[11:56] Zha Ewry: So, to copy an asset, you'd end up with a new URL, possibly not a new URI
[11:56] Morgaine Dinova: That would be a good thing, because people would scream blue murder and something would be done about it .... like getting the cache working :P
[11:56] Which Linden: Dahlia: but how would that change be communicated? You'd have to change some field somewhere to point to the new domain, might as well just be the hostpart of at url
[11:57] Zha Ewry: You can fall onto classic Http and do a reidrect
[11:57] Mojito Sorbet: Hmm, copying an asset from one server to another? I think there be security problems with that.
[11:57] Zha Ewry: Well, if you hae permsions, as always Mojitor
[11:58] Dahlia Trimble: dunno, but I could see moving my assets from one provider to another if one went under or changed their business practices
[11:58] Mojito Sorbet: The case mentioned earlier, where you do not trust one grids sewrver with your valuable stuff
[11:58] Morgaine Dinova: Jeez, let's get stuff working before we spend years thinking about how to stop it working.
[11:58] Mojito Sorbet: The multiple asset servers may be entirely ionvisible to the end user
[11:58] Zha Ewry: The asset server, being a web service can decline to give up a copy to other services if it doesn't want to..
[11:59] Mojito Sorbet: That is ok
[11:59] Which Linden: Morgaine: hm....the risk is that you end up with something like SMTP
[11:59] Which Linden: it works, but it's so easily abused that it doesn't work
[11:59] Youri Ashton: you could just add in a feature that tells you where the server is located. official LL server, opensim, or what ever. then you can decide to proceed with the TP or stop. or for that matter even filter out non-LL servers
[11:59] Which Linden: if you ignore security concerns during the design phase
[11:59] Morgaine Dinova: Which: or something like the web, which just works despite the occasional abuse.
[11:59] Youri Ashton: isnt that better then just letting the thing run wild?
[12:00] Youri Ashton: see it as a extra protection
[12:00] Dahlia Trimble: has rl to attend to... bye all, interesting conversation :)
[12:00] Mojito Sorbet: Is some asset servers do not enforce copyright protections (imagine one in China), then I may not want copies of my creations ending up there
[12:01] Morgaine Dinova: The Internet is a good example of letting things run wild, and vindicates the model. People are amazingly good at self-organizing
[12:01] Youri Ashton: bye dahila
[12:01] Morgaine Dinova: Get it working for unencumbered assets, so there is no issue.
[12:01] Zha Ewry: later dahila
[12:01] Saijanai Kuhn: Mojito, the initial services would be for Creative Commons style assets Allowing SL assets intot he wild requires all sorts of legal/scoial issues to be solved first
[12:01] Mojito Sorbet: ok
[12:02] Saijanai Kuhn: jonx Morgaine...
[12:02] Morgaine Dinova: Hehe
[12:02] Which Linden: ok I should run!
[12:02] Morgaine Dinova: If we don't take that approach, we'll be here forever trying to solve a near-impossible problem.
[12:02] Youri Ashton: kk which! thank you for this meeting!
[12:03] Mojito Sorbet: yes
[12:03] Morgaine Dinova: Thanks Which, take care :-)
[12:03] Saijanai Kuhn: SO, [aside to Prokofy
[12:03] Saijanai Kuhn: take care Which
[12:03] Which Linden: ha ha h, thanks everyone, great discussion today
[12:03] Skills Hak: D: too late
[12:03] Skills Hak: seeya which

User:Which Linden/Office Hours/2009 May 28

Navigation menu

Search