User:Zero Linden/Office Hours/2008 Jan 10

From Second Life Wiki
Jump to navigation Jump to search

Transcript of Zero Linden's office hours:

[8:36] Harleen Gretzky: He is typically late to his OH
[8:37] Rex Cronon: also, there is only one transcript for the whole month of december:(
[8:37] Leffard Lassard: =
[8:37] Zha Ewry: Ah.
[8:37] Zha Ewry: Zero's on
[8:37] Arawn Spitteler: Ask/Nag when he shows
[8:37] Zha Ewry: And there he is!
[8:37] Harleen Gretzky: Hi Zero :)
[8:38] Arawn Spitteler: All welcomes Zero
[8:38] Rex Cronon: he is close, and i think he can hear
[8:38] AC: Hi Zero !
[8:38] Zha Ewry: Morning Zero
[8:38] Zero Linden: Gooooood morning all
[8:38] Rex Cronon: hello zero
[8:38] Dr Scofield: hi all
[8:38] Rex Cronon: hi
[8:38] Tao Takashi: Good morning, Zero
[8:38] Arawn Spitteler wonders if Vineland is using Havoc 4, and if Zero is the person Zarf would want to ask about strange phenomena.
[8:40] Rex Cronon: zero, is there an ATA for the transcripts, for december, and the one from tuesday?
[8:40] Wyn Galbraith: Morning all.
[8:40] Rex Cronon: hi
[8:40] AC: hello Wyn
[8:40] Zero Linden: Vineland is not using Havoc 4 as far as I (or Zarf) knows
[8:40] Zero Linden: Rex... ETA?
[8:40] Rex Cronon: eta*
[8:41] Rex Cronon: sorry, seems like my fingers are a little slow today
[8:41] Zha Ewry: Zero.. As I cover my white board with grids about trusted/untrusted sim/server relationships, do you want to consdier
[8:41] Arawn Spitteler: Passive but Physical Object went off world, there.
[8:41] Zha Ewry: agent domains seperate from regionas, or assume they come a a pair?
[8:41] Dr Scofield: zha, why would they become a pair?
[8:42] Zha Ewry: For simplicity
[8:42] Zha Ewry: But. I assume they won't.
[8:42] Zha Ewry: Just makes more cases
[8:42] Zha Ewry: Raises questions like "Do we hand off agents from agent DomainX, to agent domainY"
[8:42] Zero Linden: Well, Tree K. was putting up the transcripts for me ---
[8:42] Tao Takashi: what does "hand off" mean?
[8:42] Arawn Spitteler wonders why replacement of one object would stop the Grid: The Grid is supposed to have some parrallel to The Internet, but the Internet was designed to recover ffrom a nuclear war.
[8:42] Zero Linden: I have the raw transcript for Dec., but (alas) without time stamps.... I'll put them up today
[8:43] Zero Linden: Zha - I've always assumed that they don't
[8:43] Zero Linden: come as pairs
[8:43] Tao Takashi: I did put one up I had for Novermber I think
[8:43] Zha Ewry nods
[8:43] Zha Ewry: Okies
[8:43] Zha Ewry: I'll do the full messy case then ;-)
[8:43] Zha Ewry mourns the last four sqaure inches of her whiteboard
[8:43] Tao Takashi: I think we need the full messy case ;-)
[8:43] Zero Linden: I don't like the idea of enforcing that every AV should need an identity-fragment on an Agent Domain associated with a Region Domain that they want to visit
[8:44] Zha Ewry: Nor do I.
[8:44] Zero Linden: smacks too much of to sliding in to the current situation on the internet today
[8:44] Zha Ewry: Just makes the picture a it middier
[8:44] Zha Ewry: *muddier
[8:44] Dr Scofield: eeek
[8:44] Zha Ewry: I'm about 1/3 of the way from having a usefuil layout of the basic problems
[8:44] Zha Ewry: So.. we can talk about what trust relationships we need in all the cases.
[8:44] Wyn Galbraith thinks Tree has been busy days as she is.
[8:44] Tao Takashi: I was trying to mention virtual worlds in the dataportability.org group btw ;-) so they think a bit further then just personal data but also content
[8:45] Zha Ewry: Also raises some interesting questoins about what ahppens if we move scripts to agent domains
[8:45] Zero Linden: I think, long term, that is inevitable
[8:45] Zha Ewry nods
[8:45] Zha Ewry: I agree
[8:45] Zha Ewry: Which.. means we'll dcoument some fun things
[8:45] Zero Linden: It is sort of like the inevitability of scripted IRC/Chat clients.....
[8:45] Tao Takashi: you mean running scripts on agent domains and just getting object updates etc. from the region domain?
[8:45] Zero Linden: at some point you are going to want to automate the way your avatar does some things....
[8:46] Tao Takashi: or are we talking about the client?
[8:46] Zha Ewry: Region sims allowing aves in, are gong to have to also allow the agent domain to see some tihings for the scripts, in a sesnible way
[8:46] Zero Linden: And again, I fear for the complexity, but I think that running attachment scripts in the AD will be a big win for usability
[8:46] Dr Scofield: zero, why not do it via the client?
[8:46] Zha Ewry: I tend to agree zero
[8:46] Dr Scofield: for automation
[8:46] Zha Ewry: And.. Actually, I think the complexity isn't that bad, because you shed a lot of complexity on the sim side
[8:47] Tao Takashi: and won't scripts in AD add a lot more of network traffic?
[8:47] Zero Linden: What advantage to on the client? Then your ability to do things depends you being on the right machine with teh right files?
[8:47] Zha Ewry: You get much smoother handling of the hand offs
[8:47] Zero Linden: or do we serve the scripts to the client?
[8:47] Zero Linden: it does remove the one thing safe from poaching: scripts - since they are never sent to the client now
[8:47] Dr Scofield: you could do that
[8:48] Tao Takashi: I think we need both.. some client specific scripting and some server side. They might serve different purposes
[8:48] Rex Cronon: the code, or the precompiled bytes?
[8:48] Dr Scofield: i was thinking about autmating your avatar
[8:48] Dr Scofield: not shipping the scripts for in-world objects, sorry
[8:48] Zero Linden: Well - I suppose if the RD grants capabilities to the AD for passing events and commands from attachement scripts, then the
[8:49] Zero Linden: AD could pass that capability on to the viewer along with the script and let the viewer run it....
[8:49] Tao Takashi: like clientside scripts might be like Firefox extensions
[8:49] Dr Scofield: didn't mean to confuse
[8:49] Tao Takashi: e.g. add IRC support as an example
[8:49] Tao Takashi: but that's different from scripted attachments
[8:49] Zero Linden: there is the issue of persisting the state of the script... especially when the connection between viewer and AD gets severed (say if the viewer crashes..... whcih I realize is a (cough)common(cough) event)
[8:49] Zha Ewry nods
[8:50] Zha Ewry: I think, in time, we'll have automatoin in both places
[8:50] Zha Ewry: We *can't* stop client side automaion
[8:50] Dr Scofield: yes...my comment related to automation of avatar via cleint
[8:50] Arawn Spitteler doesn't know AD from Client, but thinks straight Java could be used for Client Scripting.
[8:50] Zha Ewry: But.. we'll want AD automatoin
[8:50] Zero Linden: Sigh - probably - and probably in a multitude of languages and lib frameworks....
[8:50] Dr Scofield: not in-world objects/attachments
[8:50] Zero Linden: (AD = Agent Domain)
[8:51] Tao Takashi: I would hope for it not being just Java ;-)
[8:51] Tao Takashi: but anyway
[8:51] Dr Scofield: ;-)
[8:51] Zha Ewry: OK. So I'll try to habe a first draft tabele and summary in the wiki for Tuesday
[8:51] Kurt Stringer: so there would be different scripts for prims and avatars?
[8:51] Zha Ewry: *have even
[8:51] Kurt Stringer: agent vs region?
[8:51] Arawn Spitteler: Scale of Skill, Java for those expanding from Server Side LSL, and better for those who can actually program
[8:51] Zero Linden: Don't know...
[8:52] Rex Cronon: the problem will be than if the owner of the script trusts the creator
[8:52] Tao Takashi: so if scripts are run on he agent domain, what would actually needed to be transferred? I guess scanner commands and such are simply sent to the region and some result comes back then
[8:52] Dr Scofield: zero, could we discuss the secondlife:///app/login issue a bit? as neas pointed out on sldev this is the first cse where we actually run into issue re grid interop
[8:52] Zha Ewry: The sense informaoin is the obvious big one
[8:52] Tao Takashi: I can program but it's not better for me ;-)
[8:52] Tao Takashi: but that's a discussion for later anyway
[8:52] Kri Ayakashi: I believe that it won't be long before we'll see some custom clients with client side scripting... I'm actually suprised there aren't any out there yet
[8:52] Tao Takashi: so what else do attachments actually do?
[8:52] Zha Ewry: And.. this does, I think, imply, that scripts will be in at least two places
[8:53] Tao Takashi: HTTP probably but there it doesn't matter from where it gets triggered
[8:53] Rex Cronon: and java code can be decompiled quite easy, so can the creator trust the user?
[8:53] Zha Ewry: Because, you'll have scripts in rezzed objects, whichh don't belong to agents
[8:53] Zero Linden: hmmmm...... I was just thinking of generalizing the script interface so that one could have a script proxy on the sim for each attachemetn script (since they need to sense and interact with the sim environment)
[8:53] Zha Ewry: So.. one imagines those get to run out of the agent domain
[8:53] Zero Linden: but have all the events and calls proxyed to/from the script running on the Agent Domain (or Viwer)
[8:53] Zero Linden: this should be a generalized facilty
[8:53] Zha Ewry: (On the regions, or sepersate, doesn't much matter)
[8:53] Zha Ewry: and.. Yes.
[8:54] Zha Ewry: Because, if we decouple fully, we get ncier scaling
[8:54] Tao Takashi: would a region then have a general e.g. sense API which I eventually can use completely free (without a script in an attachment)?
[8:54] Zero Linden: it would allow you sell objects that had scripts where you didn't want the script to be out... and were willing to provide the compute power to run it
[8:54] Zero Linden: hmmmmmmm.
[8:54] Zero Linden: I'll talk to Babbage about this
[8:54] Tao Takashi: this would mean sensors without actual objects around
[8:54] Saijanai Kuhn exepcts to be reading a lot to catch up
[8:54] Zha Ewry: Erk.
[8:55] Zha Ewry: Interesting thing. The prims in the attachments are still goign to be on the region for people to see
[8:55] Zha Ewry: HUD ones, no
[8:55] Zha Ewry: HUD ones, are hidden
[8:55] Zha Ewry: But.. the prims rez in the regoin, don't they?
[8:55] Tao Takashi: well, do we need attachments then for these scripts if they don't show anything? ;-)
[8:55] Tao Takashi: like twitterbox
[8:55] Zero Linden: Tao - it might be a reasonable construction to enforce that anything that senses must be a script in an object (allows it to be governed by parcel controls, for example)
[8:55] Tao Takashi: well, it shows something but it wouldn't need to
[8:56] Zha Ewry: HUD attaches get interesting
[8:56] Tao Takashi: I am just thinking loud not necessarily saying that it's a good idea
[8:56] Tao Takashi: it might save prim count though
[8:56] Zero Linden: but it could be that a kind of script is just a generic proxy connection that sends events via HTTP and gets functions to invoke from HTTP
[8:56] Zha Ewry likes the thinking out loud
[8:56] Dr Scofield: HUd attaches would be a good cse for AD
[8:56] Saijanai Kuhn: so the case of attachments that are no permissions, but have scripts, will be handled differently than attachments WITH scripts...
[8:56] Saijanai Kuhn: than attachments WITHOUT* scripts
[8:56] Zha Ewry scribbles on her white board
[8:56] Tao Takashi: so what else do attachments do? maybe rez something out of their inventory
[8:57] Neas Bade: if the scripts were not too latency sensitive, it would be nice to allow the HTTP connections to be fully outside of grid
[8:57] Zha Ewry: Animations
[8:57] Zha Ewry: xmlrpc
[8:57] Zha Ewry: (tho, we hoope that goes away)
[8:57] Zha Ewry: chat
[8:57] Tao Takashi: so why not simply serve attachments from the AD, too?
[8:57] Rex Cronon: push
[8:57] Zero Linden: Neas - absolutely
[8:57] Tao Takashi: or are we saying this already? :)
[8:57] Kri Ayakashi: while on topic of transfer protocols wouldn't it be better to use something like XMPP instead of HTTP for this?
[8:57] Zero Linden: Tao - I think we are saying this
[8:57] Tao Takashi: ok
[8:57] Neas Bade: Kri, why XMPP?
[8:58] Zero Linden: ditto?
[8:58] Kri Ayakashi: for one latency and amount of data sent is smaller
[8:58] Kri Ayakashi: it's a two-way protocol
[8:58] Wyn Galbraith's work has switched to days so she has to go to work now.
[8:58] Neas Bade: kri, you've been using bad http servers then :)
[8:58] Saijanai Kuhn: Later Wyn
[8:58] Wyn Galbraith: B-bye!!! :D
[8:58] Zha Ewry: By Wynn
[8:58] Rex Cronon: bye wyn
[8:58] Dr Scofield: XMPP let's you deal better with async events
[8:59] Neas Bade: getting low http latency and good http scalability is a pretty well understood solved problem
[8:59] Zero Linden: Kri - it is true that XMPP is symetric, which is nice, but I don't think it's overhead is that much less than HTTP
[8:59] Zha Ewry: OH! Before I foget, Zero, you might want to update your "hours" sign on the easel there.
[8:59] Neas Bade: xmpp has some "fun" scaling issues that aren't all that well understood
[8:59] Neas Bade: if you look at the xmpp spec, the overhead really is more
[8:59] Dr Scofield: you only have to establish the connection once
[8:59] Zero Linden: My biggest concern with XMPP is that it doesn't really have the wide tool support
[8:59] Dr Scofield: then it's just XML messages
[8:59] Zha Ewry: you'd have to do HTTP, I think
[8:59] Zha Ewry: Becuase that's what eveything talks
[9:00] Neas Bade: dr scofield, connection overhead shouldn't be that big a deal
[9:00] Zero Linden: Well, the TCP handshake packets are a fart-in-a-whirlwind compared to the headers etc....
[9:00] Zha Ewry and the acres of angle brackets
[9:00] Zero Linden: XMPP doesn't play well with most XML tools (as it sends fragments of XML documents)
[9:00] Neas Bade: you can use keep alive and pipelining if you really think that it is causing issues
[9:00] Zha Ewry worries about the world wide angle bracket supply
[9:00] Dr Scofield: once you go for encrypted connection setup becomes expensive
[9:00] Neas Bade: yeh, HTTP is simple and nice
[9:00] Zero Linden: And there are many frameworks for serving XMPP
[9:01] Zero Linden: The big plus for me is the symetric connection no matter which way established
[9:01] Saijanai Kuhn notes that the avatar domain attachements thing seems to have died. Too many variables, not enough info?
[9:01] Tao Takashi: XMPP gets mentioned more and more it seems
[9:01] Tao Takashi: on the DiSo project they want to use that for contact detail updates
[9:01] Zha Ewry: Neas, am I right in recalling there has been some discussion in OpenSim land, about breaking scripting out into seperate spaces?
[9:01] Zero Linden: however, we've been doing some reasonable things with HTTP using the long poll technique
[9:01] Neas Bade: Tao, but typically not by folks that actually have had to implement an XMPP server :)
[9:02] Kri Ayakashi: you mean comet?
[9:02] Saijanai Kuhn: Zero if you mean EventQueueGet, that's only "reasonable" if you're sure what is going on
[9:02] Tao Takashi: Neas: They are actually implementing something for Wordpress right now
[9:02] Tao Takashi: but I don't know the details
[9:02] Zero Linden: sorry to open a fresh wound....
[9:02] Dr Scofield: neas, there are some decent XMPP servers around
[9:03] Neas Bade: we did that discussion in OpenSim land regarding region comms a few months ago, and a bunch of folks from Intel with significant XMPP experience said "you really don't want to go there. Beyond a certain scaling factor things fall over, and no one has really sovled them yet"
[9:03] Tao Takashi: well, comet also seems to grow slowly
[9:03] Neas Bade: there is a reason that HTTP is used everywhere at this point, it is really well proven for all kinds of tasks
[9:03] Dr Scofield: google seems to have
[9:03] Tao Takashi: Neas: is this readable under some URL? :)
[9:03] Neas Bade: check out the opensim-dev mailing list archives
[9:04] Tao Takashi: thanks, will do
[9:04] Neas Bade: http://opensimulator.org - follow links
[9:04] Neas Bade: it would have been Octoberish IIRC
[9:04] Dr Scofield: neas, that's just taking the "i've got a big hammer...and, oh, look, lots of nails here" approach
[9:04] Neas Bade: Dr Scofield, perhaps, but you can't underestimate the power of using technologies people already know
[9:05] Zha Ewry shrugs
[9:05] Zha Ewry: "Big swiss army knife,"
[9:05] Neas Bade: you want to bring in developers to interlock with these bits
[9:05] Neas Bade: and there are at least 100x more people that know HTTP than XMPP
[9:05] Saijanai Kuhn: properly designed, this stuff should be changeable down the road, if need be...
[9:06] Saijanai Kuhn: without HUGE hassles, even
[9:06] Kri Ayakashi: that's true but I dont think it's a good argument why to use something or why not to use something
[9:06] Zero Linden: Well, I'm equally concerned about things like - HTTP's failure modes are well understood - what happens when an XMPP connection drops?
[9:06] Neas Bade: yep
[9:06] Dr Scofield: true, i'm just concerned that we use HTTP for purposes it was never intended for
[9:06] Zero Linden: Caching is possible in HTTP, but not XMPP
[9:06] Tao Takashi: Well, I think I'd like HTTP more for that everybody knows it
[9:06] Neas Bade: HTTP was intended for reading physics papers
[9:07] Dr Scofield: pull
[9:07] Neas Bade: so we all use HTTP for things it was never intended for :)
[9:07] Tao Takashi: when I heard about XMPP I tried to read up upon it but stopped soon again as it seems a bit too read
[9:07] Dr Scofield: but not push
[9:07] Zha Ewry shrugs "90% of the HTTP uses theseb days are well beyond Mosaic"
[9:07] Tao Takashi: well, it depends on how many people will do these things with it and servers/clients will simply evolve
[9:07] Dr Scofield: ok, i'll shut up XMPP wise
[9:08] Zha Ewry: Zero.. how many singletones are there in SL at this point, which won't surivive wihout pluatlity in an open Gridded environment?
[9:08] Zero Linden: So - this has been tackled several times in HTTP: ModPubSub comes to mind and Rohit Khare's thesis ARREST - which is his pub/sub extention to REST
[9:08] Zero Linden: Hmmm... F-Sharp above middle C?
[9:08] Neas Bade adds to his reading list
[9:09] Tao Takashi: sooo, if the attachments are on the AD, how would they be rendered on the client?
[9:09] Dr Scofield: secondlife:///app/login?
[9:09] Tao Takashi: the client directly gets them from the AD? or via the RD?
[9:09] Neas Bade: yeh, the uri discussion is something I obviously have great interest in :)
[9:10] Zero Linden: Tao - good question... don't know
[9:10] Zero Linden: Dr - what are you asking?
[9:10] Neas Bade: Zero, have you kept up on the sldev thread regarding secondlife:// uris?
[9:10] Zero Linden: Zha - wouldn't it be easier to ask it the other way round?
[9:10] Tao Takashi: of we assume bad RDs then AD => client would probably be better
[9:10] Dr Scofield: the loginpage/web auth currently does not really return any information about the grid to connect to
[9:10] Zero Linden: Neas - no
[9:10] Neas Bade: it started from a discussion with Tess yesterday that Dr Scofield summarized
[9:10] Tao Takashi: the client will get these things anyway
[9:10] Zha Ewry: LOL
[9:11] Zha Ewry: Possibly
[9:11] Tao Takashi: and what about bad ADs? :)
[9:11] Zero Linden: "Yo - I'm one baaad AD, ya hear?!"
[9:11] Saijanai Kuhn: AD...?
[9:11] Zero Linden: Agent Domain
[9:11] Tao Takashi: I am not sure running scripts on ADs will solve the problem ;-)
[9:11] Tao Takashi: if it's not the AD the script creator choose to trust
[9:11] Dr Scofield: you can tell the latest clients to go your own loginpage, but all you can do in that is return the secondlife:/// URI
[9:12] Dr Scofield: we discussed with tess how we could add the grid information
[9:12] Tao Takashi: I wonder why the grid information cannot be simply in some config file on the FS
[9:12] Neas Bade: right, the issue is that secondlife:/// uri like things have implicit hostname
[9:12] Dr Scofield: (a) use a parameter like grid=.....
[9:12] Saijanai Kuhn: Was that after the Groupies meeting, or another talk?
[9:12] Tao Takashi: but then again I think I am not too much into the details of that problem
[9:13] Neas Bade: and we really need that to be an explicit hostname to make it work well with multiple grids
[9:13] Saijanai Kuhn: and did it get posted to the wiki?
[9:13] Dr Scofield: (b) use a proper URI: secondlife://grid.com/app/login?...
[9:13] Harleen Gretzky thought Tess was looking into adding -loginuri
[9:13] Dr Scofield: saijanai: i posted the chat log to the wiki
[9:13] Saijanai Kuhn: great
[9:13] Zero Linden: ah- is the concern over the integration of the correct URI syntax (secondlife:///) with the previous, tehcnically non-compliant (second://RegionFoo)
[9:13] Dr Scofield: and the summary to sldev
[9:13] Zero Linden: ?
[9:13] Dr Scofield: yes!
[9:13] Neas Bade: right, and more importantly getting to a point where
[9:13] Dr Scofield: secondlife:/// it turns out addresses the client
[9:14] Neas Bade: secondlife://osgrid.org/region/Wright%20Plaza/10/10/20 is a legitimate URI
[9:14] Saijanai Kuhn: https://wiki.secondlife.com/wiki/AWGroupies-2008-01-08
[9:14] Zero Linden: okay - so, since I'm the RFC-weenie inside LL that insisted on that third slash
[9:14] Zero Linden: let me explain
[9:14] Dr Scofield: whereas secondlife://region/X/Y/Z addresses a location in the grid
[9:14] Dr Scofield: and neas example makes a lot of sense
[9:14] Zero Linden: So - we've GOT an exsiting, ill-formed, non-compliant scheme
[9:14] Neas Bade: because SLURLs should really identify a unique network resource
[9:15] Zero Linden: secondlife://Grasmere/123/345
[9:15] Zero Linden: is Legal
[9:15] Zero Linden: really
[9:15] Dr Scofield: agree on that one, zero :-)
[9:15] Zero Linden: because the URI syntax requires that if your scheme starts with two slashes
[9:16] Neas Bade: yep, it's not that it isn't legal
[9:16] Zero Linden: THEN, the first segment MUST have the syntax for domain names
[9:16] Neas Bade: it's that by not including explicit hostname, there is an implied grid name
[9:16] Neas Bade: and to make all things equal, we should be explicit about the gridname
[9:16] Zero Linden: Well, the old syntax is what it is.... I can't fix that
[9:17] Neas Bade: right, but we could adjust over time
[9:17] Neas Bade: my last post to sldev gave a set of steps that could get you there
[9:17] Zero Linden: but let's make one thing clear, even if you put a host name there - nothing in the URI spec says that that has to be the host name you contact
[9:17] Chosen Raymaker: or introduce a new scheme name
[9:17] Zha Ewry: There is a fair question, about whether we can fix it now, so we don't have to revisit it in six to twelev months
[9:17] Neas Bade: correct
[9:17] Zero Linden: so -
[9:17] Saijanai Kuhn: or rather, so we might not have to revisit it...
[9:18] Zero Linden: secondlife: scheme is NOT intended to be the URLs of locations in world
[9:18] Zero Linden: really- that would be a poor choice
[9:18] Neas Bade: but, by making secondlife:// URIs act more like http:// URIs, it makes life a lot easier
[9:18] Saijanai Kuhn: the best laid plans of avatars...
[9:18] Zero Linden: I say, just use an HTTP name
[9:18] Zero Linden: http://slurl.com/secondlife/Grasmere/178/113/27
[9:18] Zero Linden: that is the URL of this location
[9:19] Neas Bade: ok, so where can I type that into the location browser in my SL client?
[9:20] Zero Linden: secondlife: is a scheme for invoking one's client to do something
[9:20] Neas Bade: I understand that is the way it is today
[9:20] Zero Linden: Neas - you can't 'cause our viewer has no address bar!
[9:20] Neas Bade: but it seems like it wouldn't be much to work to make secondlife:// uris actually be network resources in a sensible way
[9:21] Saijanai Kuhn: well, the map *could* be, but you'd need to have it parse and resolve slurls
[9:21] Zero Linden: It seems like an indirection that is generally uneeded
[9:21] Zero Linden: for example
[9:21] Neas Bade: and I really think that until that is make explicitly so, there is going to continue to be assumptions made in the client that don't work for alternate grids
[9:21] Zero Linden: let's take secondlife:///app/profile/Zero/Linden
[9:21] Neas Bade: it's social engineering as much as anything else
[9:22] Zero Linden: should that be *the* name of the resource that is my profile?
[9:22] Zero Linden: seems far less useful than
[9:22] Neas Bade: but your profile is only valid in the context of a specific grid
[9:22] Zero Linden: http://grid.secondlife.com/agents/Zero/Linden/profile
[9:23] Zero Linden: Ah - I see what you are driving out
[9:23] Neas Bade: ok, then get rid of secondilfe:// entirely, and just use http
[9:23] Zero Linden: driving at
[9:23] Saijanai Kuhn: actually with universal avatars...
[9:23] Zha Ewry listens closely
[9:23] Zero Linden: right - only the only way to get interaction between a web browser and a more complex UI (say, a viewer app) is this goofy scheme HACK
[9:23] Neas Bade: right
[9:23] Harleen Gretzky: Your profile is actually: secondlife:///app/agent/2e3e325b-2887-4a3c-aceb-c94227019b22/about already
[9:24] Neas Bade: and my point has been that instead of just doing the minimum for the hack using secondlife://, make the secondlife URIs explicitly contain grid infomation in them
[9:24] Neas Bade: which then makes the ideas of navigation between grids, and refernce points of the secondlife urls very explicity
[9:25] Saijanai Kuhn: /grid-"Second Life"
[9:25] Saijanai Kuhn: =*
[9:25] Zha Ewry: And.. sets us up nicely for the future
[9:25] Neas Bade: and starts to form a model where the secondlife scheme looks a lot more like www
[9:25] Neas Bade: zha: exactly
[9:25] Neas Bade: stop thinking of secondilfe:// as a hack to trick the client to do something, and think of those URIs as actual network resources that the client is asked to load
[9:26] Zero Linden: Neas - that then brings up the issue of "what identifies a grid"?
[9:26] Zero Linden: which domani name?
[9:26] Zero Linden: do we need to make the Agent Domain / Region Domain distinction now?
[9:26] Saijanai Kuhn would like to see a UUID for each grid but
[9:26] Neas Bade: hostname:port of the login server would be a reasonable starting place
[9:26] Neas Bade: as that is going to be the entry point
[9:26] Zha Ewry: As its an overlay construct over the www name space
[9:27] Zero Linden: hmmm... but then, here's the problem with that
[9:27] Zero Linden: if you use the login server and port
[9:27] Zha Ewry would hate to see it nind all thwe way to a single domain name
[9:27] Tao Takashi: btw, those profiles should contain microformats :)
[9:27] Zero Linden: BUT, that isn't the actual place the viwer goes to get the info it needs
[9:27] Zero Linden: THEN there is some form of transformation that the viewer needs to do to get to the right place
[9:27] Neas Bade: right, but google.com is 100k machines
[9:27] Tao Takashi: but then again there is not too much to markup with microformats anyway
[9:28] Neas Bade: but you can have a single name that gets you to 100k machines, and the http spreader figures out where you really go
[9:28] Zero Linden: well, that isn't quite the same, Neas
[9:28] Kri Ayakashi: btw, will SL follow login ideas like OpenID? meaning having an account on one grid lets you in to other grids ?
[9:28] Chosen Raymaker: Zero: An unrelated question before you close? Can you tell us why groups are limited at 25? If it is to limit server load, what traffic exactly is significant in this connection?
[9:28] Zha Ewry: In the one case, it's a http endpoint, in this case, we're nming an ovrlay
[9:28] Zero Linden: when you go to google.com - routers , level 2 swithces, and load balancers pick one of those 100k machines
[9:29] Zero Linden: but they are all equiavelent
[9:29] Zero Linden: for whatever you ask of google.com
[9:29] Nika Talaj: ? is a grid a service or a server
[9:29] Neas Bade: mail.google.com is pretty stateful
[9:29] Neas Bade: and that is a lot of machines
[9:29] Zero Linden: right - and they have load balancers that deal with that
[9:29] Zero Linden: BUT this is different
[9:29] Zero Linden: well
[9:29] Zero Linden: may be not
[9:29] Neas Bade: right, but those aren't l2 balancers
[9:30] Zero Linden: but what I'm saying is that if you see a resource identified by //google.com/ - you ask your question by connecting to google.com
[9:30] Neas Bade: to get mail.google.com working right, you are balancing much higher up, giving all the stateful ajax
[9:30] Neas Bade: docs.google.com is even more so the case, where your live editting is updated every 10 seconds
[9:30] Zero Linden: if you see something like secondlife://login.agni.lindenlab.com:443/app/profile/2e3e325b-2887-4a3c-aceb-c94227019b22/about
[9:30] Zha Ewry: def not l2 balances, no
[9:31] Neas Bade: anyway, the point is that having a name for a grid being a hostname shouldn't be a big deal
[9:31] Zero Linden: you need to connect to login.agni.lindenlab.com:443 to get the asnwer
[9:31] Zero Linden: and - it turns out, that that machine can't answer that question
[9:31] Neas Bade: then that URI shouldn't be returned
[9:31] Neas Bade: or if it is, we have 30x codes to bounce to the right location
[9:31] Zero Linden: Right now, the secondlife:/// scheme is intended to control the viewer, by asking it to do things with respect to the grid it is connected to now
[9:31] Kri Ayakashi: well it could use some rewrite rules based on the URL provided to bounce the request to other machine
[9:31] Zero Linden: The only excpetion is secondlife:///app/login
[9:31] Neas Bade: right, I do undestand what is is today
[9:31] Zero Linden: which needs to say which grid
[9:32] Zero Linden: I think you are absolutely right about how such things need to work in the future
[9:32] Neas Bade: but I really really really think that keeping with that design point is so fundamentally limitting that it has to change in the future
[9:32] Neas Bade: and the sooner the better
[9:32] Neas Bade: as it is a very clear early interop issue between multiple grids
[9:32] Zero Linden: I just think we should assume that there will be a new scheme than secondlife:///
[9:33] Zero Linden: well all
[9:33] Zero Linden: I'm now late for my next meeting
[9:33] Zero Linden: good stuff here
[9:33] Kri Ayakashi: well the old scheme could implictly assume linden grid as the default for some time to give everyone some space to adapt to new scheme
[9:33] Neas Bade: thanks zero, great talking with you :)
[9:33] Zero Linden: I guess I'll put this up in the wiki
[9:33] Rex Cronon: bye zero
[9:33] Zero Linden: thanks for coming
[9:33] Zero Linden: great discussions - all of 'em!
[9:33] Tao Takashi: thanks for hosting!
[9:33] Zero Linden: bye
[9:33] Tao Takashi: cya!
[9:33] Zha Ewry: Thanks zeor, lets docu ment that its choing to change, then
[9:33] Kri Ayakashi: bye Zero