User:Zero Linden/Office Hours/2008 August 07

From Second Life Wiki
< User:Zero Linden/Office Hours
Revision as of 09:48, 7 August 2008 by Saijanai Kuhn (talk | contribs) (New page: * [8:40] Zha Ewry: Morning, partly Rezed Zero * [8:40] G2 Proto: half a zero? * [8:40] Tao Takashi: I am in the normal viewer * ...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.
  • [8:40] Zha Ewry: Morning, partly Rezed Zero
  • [8:40] G2 Proto: half a zero?
  • [8:40] Tao Takashi: I am in the normal viewer
  • [8:40] Tao Takashi: old version though
  • [8:40] Tao Takashi: need to upgrade
  • [8:40] Kopilo Hallard: half a zero O.o
  • [8:40] G2 Proto: lol
  • [8:40] Bartholomew Kleiber: hmm
  • [8:41] Christophe003 Carter: A gray zero.
  • [8:41] Contagious Republic: hi zero
  • [8:41] Tao Takashi: 1.19.1
  • [8:41] Zero Linden: Hmmm.... I don't think even the Surreal numbers can represent half a zero
  • [8:41] Kopilo Hallard: :O breaks logic
  • [8:41] G2 Proto: haha
  • [8:41] Christophe003 Carter: Haha
  • [8:41] Zero Linden: It can represent half and infinity...
  • [8:41] Kopilo Hallard: hahahaha
  • [8:41] Brandon Lindman: hey you can't divide by zero
  • [8:41] Bartholomew Kleiber: half a Zero is U
  • [8:41] Zero Linden: Even the square root of infinity
  • [8:41] Bartholomew Kleiber: bottom part at least
  • [8:41] Zero Linden: Even the recipricol of infinity....
  • [8:41] Saijanai Kuhn: OK, who has brought my fps to .1?
  • [8:41] Kopilo Hallard: so half a zero is unity?
  • [8:41] Zero Linden: loves Surreal numbers
  • [8:41] Kopilo Hallard: sai, check your ram, memory leak may have occured
  • [8:41] Contagious Republic: loves funky math
  • [8:42] Widget Whiteberry: Zero Linden may be* a surreal number
  • [8:42] Saijanai Kuhn: 4.5 gigabytes
  • [8:42] Zero Linden: Uh h, Tao - what did that to your hair?!?!?!
  • [8:42] Zero Linden: OGP?
  • [8:42] Christophe003 Carter: lawl
  • [8:42] Zha Ewry: looks up
  • [8:42] Joette Bellic: What is this place exactly?
  • [8:43] Contagious Republic: zero, can a texture have clean edges between a 0% alpha zone and a 100% alpha zone? I've been trying to do it nonstop for 2 weeks and a half now.
  • [8:43] Tao Takashi: yes, OGP did that to me ;-)
  • [8:43] Zha Ewry: if (uuid == tao) {rez_overkill_hair() }
  • [8:43] Christophe003 Carter: no, textures all smoothen, contagious.
  • [8:43] Tao Takashi: too much trying to understand the SL protocol ;-)
  • [8:43] Zero Linden: Contagious?
  • [8:43] Kopilo Hallard: lol
  • [8:43] Contagious Republic: yes?
  • [8:43] Zero Linden: Oh! LOL
  • [8:44] Zha Ewry: if (uuid == Zero) {rez_massive_coffee_mug() }
  • [8:44] Christophe003 Carter: He didn't know it was your name, of course.
  • [8:44] Lucias Carnell: haha
  • [8:44] Zero Linden: I though it was a mis-type of "textures are all smoot, continuous"
  • [8:44] Waya Sion: is back from afk
  • [8:44] Zero Linden: Only in SL....
  • [8:44] Kopilo Hallard: they blur to some degree
  • [8:44] Zero Linden: Okay all
  • [8:44] Zero Linden: So I have one agenda item:
  • [8:44] Zero Linden: 1) OGP document updates
  • [8:44] Zero Linden: others?
  • [8:44] Zha Ewry: rHttp plans?
  • [8:45] Tao Takashi: OGP specification process
  • [8:45] Zero Linden: 2) rHTTP
  • [8:45] Christophe003 Carter: Mainland sims being invisible from outside, and unable to look out from?
  • [8:45] Zero Linden: Tao - can we take that with #1
  • [8:45] Rex Cronon: hello everybody
  • [8:45] Tao Takashi: and as we in pyogp have sometimes problems understanding the protocol I wanted to propose some bugtracker where these questions can be put in and answered by some competent Linden :)
  • [8:45] Zero Linden: C003 - you mean in the OGP beta, or here on the main grid?
  • [8:46] Christophe003 Carter: main grid
  • [8:46] Tao Takashi: with 1 is ok for me
  • [8:46] Zero Linden: so if it is a main grid issue, this probably isn't the right venue --- since my office hours are devoted to the future architecture of the rid
  • [8:47] Zero Linden: *grid
  • [8:47] Christophe003 Carter: k
  • [8:47] Zero Linden: 3) bugtracker
  • [8:47] Zero Linden: okay
  • [8:47] Zero Linden: #1) The OGP docs
  • [8:48] Zero Linden: so I wanted to report that Tess, Meadhbh, and I, with some help from Pheonix, have started in earnest in a full pass over the doc set
  • [8:48] Zero Linden: It will now be four documents (LLSD, OGP Base, OGP Authentication, OGP Teleport)
  • [8:48] Tao Takashi: greetings to Phoenix from me! :)
  • [8:49] Saijanai Kuhn: Meadhbh is Infinity?
  • [8:49] Oliver Szondi: shouts: will there be videos as well?
  • [8:49] Oliver Szondi: shouts: explaining whats happening
  • [8:49] Zero Linden: and each doc will document where we (the authors) believe what we (the AWG) have designed/proposed for OGP
  • [8:49] Saijanai Kuhn: sorry, usually no video at weekly meetings, Oliver
  • [8:49] Zero Linden: along with explicit notes on where the OGP Beta deviates from that
  • [8:49] Zero Linden: and how
  • [8:50] Oliver Szondi: Oh i was thinking of opensim docmentations
  • [8:50] Oliver Szondi: Sorry
  • [8:50] Saijanai Kuhn: OPenSIm meetngs are at WRight Plaza on Tuesdays and Saturdays, I believe
  • [8:50] Zero Linden: It may seem minor to most of you, but I'm really happy that we are making serious progress on the LLSD document -- especially as Phoenix wants to get
  • [8:50] Tao Takashi: yes, Meadhbh is Infinity
  • [8:50] Zero Linden: registrations for application/llsd+xml and other MIME types in soon
  • [8:51] Zero Linden: Ah yes, me bad
  • [8:51] Zero Linden: Infinity
  • [8:51] Tao Takashi: cool
  • [8:51] Zero Linden: One of the things Infinity and I finished on Tuesday is the design of the specification language for LLSD payloads
  • [8:52] Saijanai Kuhn: via caps you mean?
  • [8:52] Zero Linden: so that the rest of the specs can stop using that sort of casual syntax I'm always spouting: "{ agent_id: someUUID, favorites: [ url, ...
  • [8:52] Zero Linden: this stuff
  • [8:53] Cane Janick: ok
  • [8:53] Saijanai Kuhn: right, which is just sorta Pythonesque but not quite
  • [8:53] Kopilo Hallard: [
  • [8:53] Zero Linden: it is just that when we talk about what LLSD request or response a given resource has -- it has been informal
  • [8:53] Contagious Republic: lagged out - back.
  • [8:53] Zero Linden: and there is a need for it to be formal so that a) the spec. is precise
  • [8:53] Tao Takashi: looking forward to that, great stuff!
  • [8:53] Tao Takashi: indeed
  • [8:53] Zero Linden: and b) concise
  • [8:53] Saijanai Kuhn: and machine parsable?
  • [8:53] Zero Linden: another benefit is that it is mechanicially verifiable
  • [8:53] Zero Linden: Yes!
  • [8:54] Kopilo Hallard: and legiable
  • [8:54] Contagious Republic: so I was asking zero, is there a way to have non-smoothed alpha on textures?
  • [8:54] Saijanai Kuhn: LOcklainn will be happy to here this.
  • [8:54] Zero Linden: We've named it LLIDL, pronounced "little" --
  • [8:54] Saijanai Kuhn: Contagious this is about future designs of SL, not graphics questions. You'd want to check with Qarl on fridays
  • [8:54] Zero Linden: and it is a very simple description language -- and indeed we evaluated many others before choosing to build one of our won
  • [8:54] Zero Linden: our own
  • [8:55] Contagious Republic: okay
  • [8:55] Zero Linden: but we were pretty sure none of you would have liked ASN.1.... :-)
  • [8:55] Zero Linden: (aside to Contagious: All channels of all textures are in essence smoothed: They are Jpeg2000 encoded)
  • [8:56] Zero Linden: Now, Tao, by process did you mean going to a standards body? or just our process here?
  • [8:56] Tao Takashi: that sounds all very good, Zero :) When do you expect it to be published?
  • [8:56] Zero Linden: Two weeks I think
  • [8:56] Tao Takashi: process here, like how we can participate etc.
  • [8:56] Contagious Republic: doh, my sculptie-o-matic tool is useless then. 72 prims to a sculptie in world efficiently and quickly - made impossible by that smoothing
  • [8:56] Contagious Republic: oh well.
  • [8:56] Tao Takashi: as I noted earlier I would like to do more via some mailing list as I think it's for some stuff more productive
  • [8:56] Zero Linden: and we will get things out as we are done with them --- I think Authentication will be done sooner
  • [8:56] Zero Linden: As will teleport...
  • [8:56] Kopilo Hallard: you can get sharpe edges Contagious, but talk to qual about grahpical questions
  • [8:57] Zero Linden: Gridnauts?
  • [8:57] Tao Takashi: for instance
  • [8:57] Contagious Republic: k
  • [8:57] Tao Takashi: I think it would fit and maybe those bear emails will also stop at some point ;-)
  • [8:57] Zero Linden: sure, just perhaps put [spec
  • [8:57] Zero Linden: perhaps they should have [bear
  • [8:57] Tao Takashi: maybe they should be on a different list ;-)
  • [8:57] Cane Janick: lol
  • [8:57] Oliver Szondi: Hippos?
  • [8:57] Kopilo Hallard: [cuddly,bear]
  • [8:57] Oliver Szondi: X)
  • [8:57] Zero Linden: gridnaut-bear.....
  • [8:58] Bartholomew Kleiber: bearnaut
  • [8:58] Kopilo Hallard: xD
  • [8:58] Oliver Szondi: Hippos-Gridnaut
  • [8:58] Christophe003 Carter: lol
  • [8:58] Oliver Szondi: X)
  • [8:58] Kopilo Hallard: likes Bartholomew's
  • [8:58] Christophe003 Carter: sticks with bartholomew's
  • [8:58] Saijanai Kuhn: so will this format be used to describe payloads coming in through EQG?
  • [8:58] Zero Linden: If there is sufficient traffic on specs, then sure, we'll add another list
  • [8:58] Whichway Janus: Sorry, I think I started the lastes batch of those. BUt lots of folks are succeeding now.
  • [8:58] Kopilo Hallard: imagines a bear in a space suit
  • [8:58] Zero Linden: EQG?
  • [8:59] Saijanai Kuhn: EventQueueGet
  • [8:59] Tao Takashi: I sorta like the way how opensocial list works
  • [8:59] Tao Takashi: but only from having a glance at it
  • [8:59] Zero Linden: Yes, Sai, we'll be specing what comes through EQG with LLIDL
  • [8:59] Tao Takashi: it seems pretty effective and I might look a little deeper on how their process is
  • [8:59] Saijanai Kuhn: kool cause that's a big unkown right now
  • [9:00] Zero Linden: okay - please do and report back!
  • [9:00] Zero Linden: We will keep the practice of having each formal draft release as an immutable document so that we can have a solid reference point --- but we (LL) do need to get more frequent updates
  • [9:01] Saijanai Kuhn: also, you need a category listing for it
  • [9:01] Tao Takashi: maybe it's also good to have some initial discussions about new stuff before it gets written on the list?
  • [9:01] Zha Ewry: We really need at least three specs, at the moment
  • [9:01] Tao Takashi: so we can collect use cases and such for it and everybody has some common understanding what we are working on
  • [9:01] Zha Ewry: 1) Design as believed to be planned and idea
  • [9:01] steampunk Jetpack: (murga's build - kopilos script): All Go
  • [9:01] Zha Ewry: 2) What is currently suppsoed to be implemented
  • [9:02] Zha Ewry: 3) What is actually deployed
  • [9:02] Tao Takashi: and maybe this can be represented by some wiki structure
  • [9:02] Saijanai Kuhn: suggests [1] as well as Qrchitecture Workign group
  • [9:02] Saijanai Kuhn: Architecture*
  • [9:02] Zha Ewry: All three sometimes exist in parallel
  • [9:02] Saijanai Kuhn: so you can find those docs even if you don't know where to look for them
  • [9:03] Zero Linden: Zha - I wonder how often #1 and #2 will diverge
  • [9:03] Zha Ewry: Not a critique, btw, just reality
  • [9:03] Zha Ewry: Not very oftent, btu..t hey do from time to time
  • [9:03] Zero Linden: and keeping three docs uptodate might kill me... :-)
  • [9:03] Tao Takashi: right now I remember the redirect to the seed cap
  • [9:03] Saijanai Kuhn: or even Category:OGP_Documentation
  • [9:03] Tao Takashi: and the wrong header
  • [9:03] Zha Ewry: When we're discovring issues, mostly
  • [9:03] Tao Takashi: this would be between 2 and 3 I guess, I don't really get 1 ;-)
  • [9:04] Kopilo Hallard: maybe the first three steps of sdlc
  • [9:04] Zero Linden: well - I'd hoped we could do it all within one document
  • [9:04] Zero Linden: but that might be unreasonable
  • [9:04] Zero Linden: let's see how this draft goes....
  • [9:04] Saijanai Kuhn: could have one text doc and then lots of supporting docs and diagrams
  • [9:05] Zero Linden: It's the first one where there is a difference between #3 and #1/#2
  • [9:05] Zha Ewry: I'm mostl looking for us to be careful about knowing where the deltas are.
  • [9:05] Tao Takashi: maybe the deltas can be some list in jira
  • [9:05] Zero Linden: We've used an XML Spec element that formats as an inset red text box
  • [9:05] Zero Linden: the deltas stand out
  • [9:06] Tao Takashi: what are you using for writing that spec btw?
  • [9:06] Tao Takashi: just wondering after I sent out that email ;-)
  • [9:06] Zero Linden: the editor? oXygen
  • [9:06] Zero Linden: the schema? W3C's XML Spec
  • [9:06] Tao Takashi: and some custom xml format?
  • [9:07] Tao Takashi: ah, ok
  • [9:07] Zero Linden: W3C has XSLT that transforms into HTML and PDF, and I have slightly modified versions of those for this doc set
  • [9:07] Zero Linden: and I did XSLT that transforms into WikiText
  • [9:08] Tao Takashi: I wonder if this could be transformed to something like djangobook uses so that it makes adding comments easier
  • [9:08] Zero Linden: and we have CSS style sheets that can style the source directly for viewing (!)
  • [9:08] Zero Linden: I'm sure
  • [9:08] Tao Takashi: I guess theoretically that should be possible if you have that app ;-)
  • [9:08] Zero Linden: it is a pretty easy schema to transform, which is why I choose it
  • [9:08] Kopilo Hallard: I like the use of standardised measures
  • [9:08] Zero Linden: since at some point there is a high likelyhood we'll have to transform it into rfc2xml format
  • [9:08] G2 Proto: wow cool news Zero on the CSS
  • [9:08] Zero Linden: er, xml2rfc format
  • [9:08] Oliver Szondi: wowha!
  • [9:09] Zero Linden: also - we've put this all in the source tree, so I think at the next drop you'll all have the full suite of supporting files we use to build this
  • [9:09] Zha Ewry: Nice
  • [9:09] Kopilo Hallard: sweet
  • [9:10] Zero Linden: Of course, you can use some other XML editor and tool chain if your prefer -- the Komodo editor is free and works well, for one
  • [9:10] Saijanai Kuhn: so back to categories... Have at least one that can be used to index all OGP doc related stuff
  • [9:10] ribanna Magic: helllo
  • [9:10] Tao Takashi: very good :)
  • [9:10] Zero Linden: Good idea, Saij.. .please remind me to do so if I forget when we put the next set out
  • [9:10] Kopilo Hallard: xml is pretty straight foward could even use a cli app :p
  • [9:10] Zero Linden: okay
  • [9:11] Tao Takashi: once had some oxygen license
  • [9:11] Zero Linden: I need to move on....
  • [9:11] Zero Linden: #2) rHTTP
  • [9:11] Tao Takashi: ok, I might maybe post some process proposal to gridnauts after looking at opensocial
  • [9:11] Zero Linden: Sigh....
  • [9:11] Tao Takashi: and will post other things with [spec
  • [9:12] Zero Linden: so - we are, after many months, still facing this archtectural road block: How to invoke resources on the viewer at the instigation of the agent domain or region
  • [9:12] Zero Linden: Original SL solution: UDP messages
  • [9:12] Zero Linden: 2nd SL solution: The Event Queue
  • [9:12] Zha Ewry: perks up
  • [9:12] Zero Linden: Various technologies looked at by this group: AMQP, XMPP, BEEP, SCTP
  • [9:13] Zero Linden: none of these really made us want to jump on them --- for quite a variety of reasons:
  • [9:13] Zero Linden: poor library support, heavy-weight standard with much we didn't need,
  • [9:14] Zero Linden: difficult to work with model, etc....
  • [9:14] Tao Takashi: reminds me that I still wanted to read up on XMPP :)
  • [9:14] Tao Takashi: as I think it still will play a role in the future
  • [9:15] Zero Linden: Then we got very mired in a not-really-relevant debate over issues of the common understanding of the term "comet" and "ajax" --- and we often wrestled with how to talk about keep-alive and it's role
  • [9:16] Zero Linden: Somewhere along the way, Donovan and I cooked up running HTTP in reverse, and then Donovan enshrined it in the rHTTP proposal with a proof-of-concept implementation in Python
  • [9:16] Zero Linden: It supposed an event queue like fall back
  • [9:16] Tao Takashi: I need to post the splitted version of this
  • [9:16] Tao Takashi: so people can more easily see what's client and server code
  • [9:16] Zero Linden: though the form of *that* event queue differs from the SL-in-production Event Queue
  • [9:16] Zha Ewry: Have we tested that proof of concept yet?
  • [9:17] Zero Linden: And - that's where we are today
  • [9:17] Tao Takashi: and maybe that people can test it :)
  • [9:17] Tao Takashi: I will try to do this latest this weekend and inform people on how to use it
  • [9:17] Zero Linden: "test" as in - does it work and play well with various HTTP infrastructure doodads? Somewhat
  • [9:17] Zha Ewry: Yes, tha sort of test
  • [9:18] Lucias Carnell: ok jusr define "reverse"http for me O.o its not something ive ever stumbled across
  • [9:18] Zha Ewry: It would be really nice to be able tosay to skeptics "Hey, we got 10,000 people to try this in lots of settings, and we ran into the following X managable issues"
  • [9:18] Tao Takashi: maybe one could build some example chat app on top of it
  • [9:18] Zero Linden: It uses the same technique as a number of other extensions to HTTP 1.1 --- (including the "upgrade to SSL" extensions that the http framers so want, but nobody uses)
  • [9:18] Tao Takashi: should be simple enough and would show how this can be put into some lib and where the problems are
  • [9:19] Tao Takashi: and I also wanted to find out how to use this with the standard lib instead of eventlet
  • [9:19] Tao Takashi: maybe somebody can also implement some example in C# etc.
  • [9:19] Saijanai Kuhn: I'm silly I thought rHTTP was already part of the HTTP just that no-one used it
  • [9:19] Zha Ewry: I may do some simple C# example, while I'm off on the road without my usual bevy of test machines
  • [9:20] Lucias Carnell: c# or python would be nice to see <3
  • [9:20] Tao Takashi: cool, Zha
  • [9:21] Tao Takashi: and you don't take a server park with you on vacation? ;-)
  • [9:21] Zero Linden: So, at this stage, I'm wondering if we can narrow our options to one of:
  • [9:22] Zero Linden: 1) require rHTTP, the end
  • [9:22] Zero Linden: 2) rHTTP with the better EventQueue-Like fallback
  • [9:22] Zero Linden: 3) just do the EventQueue-Like thing, and not bother with rHTTP
  • [9:23] Zha Ewry: well, the rHTTP model really lets us think of the whole architecture symettrically, which is awfully nice
  • [9:23] Zero Linden: *IF* Keep-Alive is implemented in the whole stack, then #3 is just a little chattier than #1
  • [9:23] Zha Ewry: especially, when we try to allow things like proxy clients, and such
  • [9:23] Zero Linden: They both let us think of it Symetrically
  • [9:23] Zha Ewry: they do, but for #3, you have to do it with slighty of hand
  • [9:23] Zero Linden: and they both require that one side queue requests and wait to be contacted
  • [9:23] Zero Linden: so that those parts don't change
  • [9:24] Zha Ewry: Right, but you have to keep sayign to people "The queue, should be thought of as a hack"
  • [9:24] Zero Linden: How about "The queue should be thought of as transport"
  • [9:25] Zha Ewry: LOL
  • [9:25] Zero Linden: Well, HTTP proxy's will have to treate rHTTP as they do SSL - establishing an end-to-end connection (the CONNECT method)
  • [9:25] Zha Ewry: I'd love.. deeply love to be abel to model the client as offering a set of presentation services, which the rest of the cloud icalls, when the client asks for that
  • [9:25] Zero Linden: and then simply passing data
  • [9:26] Zha Ewry: Its about 70% a question of semantics
  • [9:26] Zero Linden: Zha - That is exactly how the simulator->viewer HTTP messages, passed through the event queue, are modeled to day
  • [9:26] Zha Ewry: and 30% more real, as if we don't do rHTTP, then software acting *liket* the client, but being abel to be
  • [9:27] Zha Ewry: in a pure HTTP servinng abiliyt, won't be written that way
  • [9:27] Zero Linden: in fact, the viewer HAS a full HTTP server in it - it just isn't connected to a TCP socket, instead it is feed by the EventQueue
  • [9:27] Zha Ewry: right, but we don't code the rest of the cloud as if its there...
  • [9:27] Zha Ewry: I mean, we do at the design level
  • [9:27] Zha Ewry: but not at the code level
  • [9:28] Zha Ewry: I keep suspecting that the more we enshrine that in our next iteratoin, the more it will come back and bite us
  • [9:29] Tao Takashi: so for 3) would speak that new devs don't have to learn a new concept but use something which is known to them to some extent in the form of comet
  • [9:29] Zero Linden: Well... actually, in the LL simulator we do: We simply get an end-point object and fire a message/request at it...
  • [9:29] Tao Takashi: and the question might be if 1) alone works all the time
  • [9:29] Zero Linden: the underlying end-point logic says (oh, gee, this thing is at the wrong end of a NAT firewalll, okay, sure, I'll just stick it in this queue...)
  • [9:29] Zha Ewry: nods
  • [9:29] Tao Takashi: and it not we need to ask people to implement 2 things
  • [9:30] Zha Ewry: I guess the trick if we end up doing 3 is to spec it properly, so that you *can* do it with a server ont he recieveing end, not a queue
  • [9:30] Zero Linden: Tao - exactly - there is the rub (same question, really about llsd+xml and/or llsd+json)
  • [9:30] Tao Takashi: and I personally of course would always go with stuff people know and maybe have implemented already
  • [9:30] Zha Ewry: I badly want, if I'm actually buildng a "client" which looks like the SL client, but is a server which streams out video, for example
  • [9:30] Zero Linden: If none can agree, and almost everyone has to do both anyway, why not just pick one
  • [9:30] Zha Ewry: to build it properly, with a web service
  • [9:31] Zha Ewry: not a queue
  • [9:31] Tao Takashi: and I must also admit that rHTTP is sort confusing to me, esp. on how to implement it
  • [9:31] Cane Janick: I have a coin
  • [9:32] Tao Takashi: ok, so pick the comet style event queue and json then ;-)
  • [9:32] Zero Linden: So, rHTTP allows one to use the exact same frameworks as you'd expect (ex.: HTTP Server framework in the viewer and HTTP Client lib in the region)
  • [9:32] Zero Linden: IF those frameworks and libs have the right hooks to be used by the rHTTP extension --- Donovan showed that in Python they do (though just barely)
  • [9:32] Tao Takashi: yes, but it took a long time for me to actually understand this code
  • [9:33] Zero Linden: I don't know if things like libcurl does or if Apache can be made to
  • [9:33] Tao Takashi: and I still need to think somewhat more to know how to implement it myself
  • [9:33] Tao Takashi: add to it that I don't have any other endpoint right now where I can test it aganist
  • [9:33] Zero Linden: On the other hand the EventQueue-Like fallback definitely can be built with all the standard compentry that is out there....
  • [9:33] Tao Takashi: so I don't know if I am doing the right thing
  • [9:33] Zha Ewry: mmm.
  • [9:34] Zero Linden: ....but requires you to build an abstraction layer if you want to hide this odd transport from your application
  • [9:34] Tao Takashi: and event queue stuff shouldn't be too hard, just wait with sending until you have somethng to send
  • [9:34] Zha Ewry: Iguess, one thing to ask, in the near term, possibly beyond, is how badly do we want to make it behave well for things which aren'tt hecurrent client?
  • [9:35] Tao Takashi: what do you mean, Zha?
  • [9:35] Bartholomew Kleiber: to have an abstraction layer gives the chance to change the decision we make now at a later time ...
  • [9:35] Zero Linden: The last though I have on this is: One *could* build a rHTTP adapter as either a separate process or at least an independent component - and have it ACTUALLY SPEAK HTTP over a localhost TCP socket
  • [9:35] Zero Linden: to the unmodified frameworks and libraries
  • [9:35] Zha Ewry: The whole queyue and such, is 90% about the fact that the client lives in so many places behind so much NAT and Firewall gear
  • [9:35] Zha Ewry: (as is expect in a very client centric bit of code)
  • [9:36] Zha Ewry: but.. there are lots of things we might put at the edge of a sim, which aren't like that at all, bnut will look like a "presentation endpoint" at the protocol level
  • [9:37] Zero Linden: well all
  • [9:37] Zero Linden: alas, I want to go
  • [9:37] Zero Linden: but I think this last few minutes of discussion has lead to a very good framing of the issue
  • [9:37] Zero Linden: and perhaps after a few days of mulling we'll have a concensus viewpoint...
  • [9:37] Tao Takashi: btw, do we do voting on decisions? ;-)
  • [9:38] Tao Takashi: might be a process question
  • [9:38] Zero Linden: i *need to go --- not want.... I'd actually much rather stay here than go off to look at business projection slides.... :-)
  • [9:38] Tao Takashi: I know the opensocial people do
  • [9:38] Kopilo Hallard: given the amount of firewalls both physical and software... hmmm
  • [9:38] Cadis Blackadder: LOL
  • [9:38] Rex Cronon: bye zero
  • [9:38] Tao Takashi: thanks for the meeting, Zero! :)
  • [9:38] Zero Linden: Not sure, Tao, honestly!
  • [9:38] Tao Takashi: quite productive
  • [9:38] Cane Janick: thx
  • [9:38] Kopilo Hallard: take care Zero
  • [9:38] Zha Ewry: Thanks Zero, see you in three 1/2 weeks
  • [9:38] Zero Linden: later
  • [9:38] Bartholomew Kleiber: bye zero