User:Zero Linden/Office Hours/2008 Mar 06

[8:30] Wyn Galbraith: Morning Sai & Rex.
[8:30] Rex Cronon: hi wyn
[8:30] Rex Cronon: hi everybody
[8:31] Tao Takashi: Hello everybody!
[8:31] Rex Cronon: hi tao
[8:32] Skotos Koltai: Hey, Tao. Thanks for the TP.
[8:32] Tao Takashi: you're welcome :)
[8:32] Tao Takashi: Skotos is member of the Data Portability Group
[8:33] Tao Takashi: Hey Keystone
[8:33] Keystone Bouchard: Hey Tao! Longtime..
[8:33] Dr Scofield: hi all
[8:33] Tao Takashi: great presentation you have at Metaverse U :)
[8:34] Skotos Koltai: BTW - Anyone wanting to join the DataPortability Project in SL, I've set up a Group for it.
[8:34] Keystone Bouchard: thanks!
[8:34] Jade Lily: hi tao!
[8:34] BigMike Bukowski: That sounds interesting, what exactly are you working on with it?
[8:34] Tao Takashi: Hey Jade :)
[8:34] Keystone Bouchard: we finished that competition on Friday [1]
[8:34] Zha Ewry: Inbound Zero
[8:35] Echo Seigo: hey all
[8:35] Goldie Katsu: Ahh I have excellent timing to day
[8:35] Goldie Katsu: grins
[8:35] tx Oh: moin
[8:35] Tao Takashi: cool, congrats!
[8:35] Wyn Galbraith: Morning Zero.
[8:35] Tao Takashi: Hi Zero!
[8:35] Rex Cronon: hi echo
[8:35] Dr Scofield: hi goldie
[8:35] Keystone Bouchard: thanks! we're pretty excited - looking for our next project now
[8:35] Tao Takashi: btw, what's your email? :)
[8:35] Zero Linden: Good morning all
[8:35] Keystone Bouchard: jbrouchoud@gmail.com
[8:35] Goldie Katsu: Hi Dr. S
[8:35] Rex Cronon: hi zero
[8:35] Goldie Katsu: Hi Zero
[8:35] Tao Takashi: well, I meant Zero :)
[8:35] Dahlia Trimble: nmbMorning :)
[8:35] Dr Scofield: hi zero
[8:35] Keystone Bouchard: lol!
[8:35] Tao Takashi: but thanks :)
[8:35] Zha Ewry: sips RL coffee, and greets Zero
[8:35] JenzZa Misfit: greetings all .... hello Zero :)
[8:36] Zero Linden: checks that his coffee cup is full.... it is
[8:36] Goldie Katsu: looks over at the RL coffee maker and hopes he goes down to make some soon.
[8:36] Rex Cronon: hi
[8:36] Dahlia Trimble: everyone looks grey today :(
[8:36] Wyn Galbraith: bought a 2 cup coffee mug.
[8:36] JenzZa Misfit: sweeet
[8:36] Cel Edman: Hi all, and zero.. get rid of that roof someday. getting clausterfobic being in small spaces in sl
[8:36] Zero Linden: Hmmm... but it makes such a nice, cozy spot
[8:36] Rex Cronon: hi
[8:37] Adam Xinpeng: Hai guys!
[8:37] Zero Linden: kjust be cglad I don't hold these meetings inside the lotus flower!!!
[8:37] Saijanai Kuhn: 34 avatars. Your spamertizing dollars at work
[8:37] Keystone Bouchard: i like the roof! defines the space =)
[8:37] Zha Ewry: Why do people like to sit inside BamBam's chairs?
[8:37] Wyn Galbraith: Cause they're cool.
[8:37] Zha Ewry: On them, tho
[8:37] Zha Ewry: Not inside !
[8:37] Tao Takashi: full house it seems
[8:38] Wyn Galbraith: has spent time inside the lotus. "Oh I don't know about sitting on them."
[8:38] Tao Takashi: either everybody wants grid interop or some protest is planned here ;-)
[8:38] Adam Xinpeng: Cant we want both?
[8:38] Tao Takashi: Hey Adam
[8:38] Zha Ewry: Its good to want things Adam :-)
[8:39] Tao Takashi: WANT WANT WANT WANT!
[8:39] Saijanai Kuhn: wantsw a job
[8:39] Tao Takashi: not so good to be wanted, though
[8:39] Zero Linden: Oh man, and here I thought virtual worlds were all about giving...
[8:39] Zha Ewry: wants to buy Saij a SL spell checker ;_)
[8:39] Zero Linden: wants Zha to buy the second life viewer a spell checker
[8:39] Dr Scofield: i'd like voice info: is parcelvoiceinfo the only piece of information that will cause the client to show an island as voice enabled? or is parcel voice info only part of the game?
[8:39] Tao Takashi: yeah, I want you to give ;-)
[8:39] Zha Ewry: agrees
[8:40] Dahlia Trimble: doesnt want the job, only the money
[8:40] Dr Scofield: LL should integrate aspell
[8:40] tx Oh: but i lifithello grey people
[8:40] Dr Scofield: and be done with it
[8:40] Tao Takashi: and they should integrate the same for voice :)
[8:40] Rex Cronon: hi
[8:40] Adam Xinpeng: We should totally have one of these meetings in OpenSim. ;)
[8:40] Dr Scofield: lol
[8:40] Zero Linden: Well - I'm down to working out the freakin' work flow for the spec document.... I've got a version in Pages
[8:41] Zero Linden: but I can't really work from that -> not very standards friends
[8:41] Zero Linden: friendly
[8:41] Cel Edman: for me I try to give 60% and would be nice to get 40% back.. But encountered some black-holes irl. that only sucks, people like that i try to avoid
[8:41] Zha Ewry: Ah
[8:41] Dr Scofield: Pages?
[8:41] Tao Takashi: Mac OS X Word app
[8:41] Dr Scofield: ah
[8:41] Zero Linden: Pages is the Apple, iWork text processor
[8:41] Saijanai Kuhn: minimal functionaly tres imple
[8:41] Dr Scofield: neooffice
[8:41] Saijanai Kuhn: tres simple
[8:41] Zero Linden: It is a nice environment for writing...
[8:41] Zero Linden: but see, I need to be able to get this thign into various formats:
[8:41] Zha Ewry: will take it in almost any format that can be converted and looked at
[8:42] Zero Linden: wiki, PDF, and probably rfc2629 XML
[8:42] Zha Ewry: will get out the old kiln, and bake clay tablets if needed
[8:42] Zero Linden: heh - true enough
[8:42] Wyn Galbraith: If you use PDF can't that be read by mostly any OS?
[8:42] Dr Scofield: shudders to think of the fedex fees
[8:42] Saijanai Kuhn: Mac OS X has a print to PDF option
[8:42] Zero Linden: But I've got to eb able to actually encorporate peoples input and have a work flow worked out....
[8:42] Wyn Galbraith: Or has Adobe not made a reader for Linux
[8:42] Zero Linden: otherwise it is going ot be lots of manual labor for me
[8:42] Zero Linden: and I'm a wuss
[8:42] Tao Takashi: Pages can also directly export to PDF but I think the problem is how to work on it together
[8:42] Dr Scofield: kpdf
[8:42] Rex Cronon: how about OpenOffice docs?
[8:42] Zha Ewry: No good, that
[8:43] Adam Xinpeng: Push it to a PDF, then someone can cut and paste it onto the wiki
[8:43] Adam Xinpeng: as long as it's out there, I think we're good?
[8:43] Tao Takashi: I am not sure why it's not simply a wiki ;-)
[8:43] Wyn Galbraith: Oh, true, you need the whole adobe package to work on it.
[8:43] Rex Cronon: brb.......
[8:43] Zero Linden: I'm glad you are all willing to do the work - but I think you underestimate how long these docs are going to get and how often we're going to have to do this process
[8:43] Saijanai Kuhn: has some vague idea... ;-)
[8:44] Adam Xinpeng: Just do it on the wiki to begin with?
[8:44] Dr Scofield: i think we never underestimated that
[8:44] Zha Ewry: We need a nice clean process
[8:44] Adam Xinpeng: Or make it a .txt file we can do .diffs on?
[8:44] Zha Ewry: Easy in easy out, easy edit
[8:44] Zero Linden: Anyhow,
[8:44] Zero Linden: I'm going to publish this stuff today (me hopes soooo much)
[8:44] Tao Takashi: well, other groups work with a wiki (like Data Portability, Microformats, ...) so I guess we can do that, too. There just needs to be a process in place who should edit the official version when
[8:44] Saijanai Kuhn: https://wiki.secondlife.com/wiki/Login_Protocol
[8:44] Tao Takashi: then again this wiki is still damn slow
[8:45] Gareth Ellison: hi
[8:45] Dr Scofield: exactly
[8:45] Dr Scofield: text with markdown markup and all in repository
[8:45] Zero Linden: going to publish it as:
[8:45] Gareth Ellison: may i ask for logs in a notecard?
[8:45] Zero Linden: 1) PDF file (fixed, "hard" reference)
[8:45] Saijanai Kuhn: chat log will be available after the meeting
[8:45] Rex Cronon: back
[8:45] Tao Takashi: Gareth: There wasn't too much info yet, just the question on how to publish the document Zero created
[8:45] Zero Linden: 2) Wiki pages, the text will be locked, but the talk pages unlocked
[8:46] Saijanai Kuhn: https://wiki.secondlife.com/wiki/AW_Groupies#Chat_Logs
[8:46] Gareth Ellison: what document? :) i'll sit and observe
[8:46] Magic Bracelet: -- current (ACTIVE): Repeat Sensor Has been Touched.
[8:46] Magic Bracelet: -- current (ACTIVE): Turning sensor on
[8:46] Zero Linden: 3) Co-oment page for markup
[8:46] Tao Takashi: the initial grip interop doc from Zero
[8:46] Zero Linden: The reason for the locked version on the wiki is to make sure that we can discuss the same thing
[8:46] Tao Takashi: grid
[8:46] Zero Linden: between revs
[8:46] Dr Scofield: ;-)
[8:46] Rex Cronon: wiki
[8:46] Zero Linden: I'm find if people want to collaborative edit sections on the side
[8:47] Zero Linden: but we need to keep a reference copy clean in the wiki at all times --- such is the needs of standards
[8:47] Saijanai Kuhn: Maybe put a lnjk to a stub page for each section for discussion?
[8:47] Saijanai Kuhn: a link to a stub page*
[8:47] Zero Linden: Sai - yes, that is a good idea - I'll be sure to make sure they are there
[8:48] Zha Ewry: We have to have the "This is the current consensus" version
[8:48] Zero Linden: well, I'd call it "the current proposal"
[8:48] Zha Ewry: Sure
[8:48] Zha Ewry: But. as time goes on
[8:48] Zero Linden: since we are not going to get complete consensus on everything!
[8:49] Zha Ewry: We need the current 'This is what we've all agreed to is the current version"
[8:49] Saijanai Kuhn: I have a feeling that I've seen a bit of the standard format. Angela, Tess and Donovan edited the straw login-page a day or two ago with a nice format
[8:49] Zero Linden: Ah yes- have you folks been following that?
[8:49] Zha Ewry: including the "We find points X, Y and 67" to be contentious
[8:49] tx Oh: dreams of a set of rfc about sl protocols
[8:49] Saijanai Kuhn: was wroking on a revision myself and then saw what they had done
[8:49] Leffard Lassard: The current proposal is always relative. There should be a way to refer comments to concret versions.
[8:49] Zero Linden: So, Donovan and Angela have been documenting the login protocol as they implemented it
[8:50] Zero Linden: their pages reflect what is running in the code
[8:50] Saijanai Kuhn: https://wiki.secondlife.com/wiki/Second_Life_Login_API_Strawman
[8:50] Gareth Ellison: will there be some kind of draft document in a static form short-term besides the wiki?
[8:50] Zero Linden: that protocol will probably lag the reference doc for some time - but I think we are comfortable with that
[8:50] Zha Ewry: That is wicked good Zero
[8:50] Zha Ewry: Are you making the rest of the shop keep an eye on the process?
[8:50] Zero Linden: Gareth, you mean for Donovans and Angela's work?
[8:50] Dahlia Trimble: protocol lag the document? thats a switch
[8:51] Zero Linden: Yes, Dahlia - it is
[8:51] Zha Ewry: And a good one
[8:51] Gareth Ellison: i mean the arch specs in generral
[8:51] Gareth Ellison: ugh, tyops
[8:51] Zero Linden: And please don't underestimate the amount of effort it is going to take to make this kind of switch
[8:51] Zha Ewry: God no
[8:51] Dahlia Trimble: would seem quite a task
[8:52] Saijanai Kuhn: knows what effort he has done. Sobbed quietly whenn he saw the new format
[8:52] Zha Ewry: spent a year, teaching a team of 15 programmers to document as they coded
[8:52] Tao Takashi: that's why it's quite nice to have doctests :)
[8:52] Tao Takashi: but it means teaching people to test as they code ;-)
[8:52] Zero Linden: Oh - I so love test-driven development
[8:53] Zha Ewry: Spent a week, rejecting check ins because ti included code fragments like p*++ = i[j]+k[j
[8:53] Adam Xinpeng: haha nice
[8:53] Tao Takashi: I will definitely try to do that in the agent domain stuff
[8:53] Zha Ewry: "What, I have a comment!"
[8:53] Zero Linden: On my own projects I'm the kind of guy that writes the test first and watchs the compile first complain about the missing .h file before creating the .h file
[8:53] Zha Ewry: Yep
[8:53] Zero Linden: then waits for the error about the missing function declaration .....
[8:53] Gareth Ellison: crazy hack below, don't try and understand it:
[8:53] Zero Linden: then the link error
[8:54] Zero Linden: etc....
[8:54] Gareth Ellison: that's my fave comment
[8:54] Wyn Galbraith: Careful Tree!
[8:54] Zero Linden: It's like page-faulting in the code
[8:54] Dr Scofield: lol
[8:54] Zha Ewry: Forces you to have all the bits
[8:54] Tree Kyomoon: sorry everything rezzing still
[8:54] Zero Linden: And forces you not to write stuff you don't actually need
[8:54] Dr Scofield: not a bad approach
[8:54] Gareth Ellison: zero - how much of the grid arch currently is anything similar to TDD-based dev?
[8:55] Zero Linden: You mean, how much of Linden Lab's engineering efforts are test-driven?
[8:55] Gareth Ellison: i can gues based on the messy viewer code the sims aren't but the python web services are sexy TDD
[8:55] Saijanai Kuhn: so that format for the login is pretty much how things are going to go, Zero?
[8:55] Gareth Ellison: yeah
[8:55] Zero Linden: Almost none
[8:55] Gareth Ellison: heh
[8:55] Zero Linden: Studio Icehouse and Studio DaBook do a fair bit
[8:55] Zha Ewry: We're hoping to fix that, tho, as a community, right?
[8:55] Zero Linden: but mostly in the python services they provide
[8:56] Gareth Ellison: :) Zha
[8:56] Zero Linden: I believe that Studio Icehouse's current work on the next phase of HetGrid
[8:56] Zero Linden: is entirely test driven
[8:56] Saijanai Kuhn: Zero, just double-checking: the strawman login is in the new docs format for sure?
[8:56] Tao Takashi: do you use doctests there or unit tests?
[8:56] Zero Linden: and has (or at least had) 100% source coverage via tests
[8:57] Zero Linden: Sai - I the current documented work in the wiki for login is, say 90% what's going to be in the doc I produce
[8:57] Tao Takashi: and the viewer in the agent domain branch is working against that strawman login protocol?
[8:57] Zero Linden: There is some breaking apart of the legacy stuff into three legacy methods
[8:57] Saijanai Kuhn: KK, know what to make anything I do look like, thanks.
[8:57] Zero Linden: and moving some odd ball stuff from the basic login capability into a legacy one
[8:57] Gareth Ellison: regarding current protocol documentation - internalyl is it just a matter of reading the code? (please excuse my lag-induced typos)
[8:58] Zero Linden: Oh, you mean style?
[8:58] Saijanai Kuhn: yeah. style
[8:58] Zero Linden: Ah - well, that is part of the work flow stuff I'm working on
[8:59] Saijanai Kuhn: Oh, OK. I was working on a mergin of things when I saw what they had done. Seemed much beter than my attempts
[8:59] Zero Linden: if you want a reference, I'm targeting the capabilities of rfc2629 - which is an XML schema for IETF RFCs
[8:59] Saijanai Kuhn: KK Thanks
[8:59] Zero Linden: Tao - I think the bulk of our tests are using unit test, but some use doc test
[8:59] Saijanai Kuhn: [2]
[8:59] Gareth Ellison: zero - about protocol docs? do LL devs just have to read the code or is there any pre-written specs?
[9:00] Zero Linden: Personally, I find doc test somewhat brittle
[9:00] Zero Linden: since it relies on final text formatting of your results
[9:00] Zero Linden: Internally, generally it is "read the code"
[9:00] Zero Linden: we have some internal wiki pages about high level organziation of various sub-systems
[9:00] Zero Linden: but alas, it is generally out of date
[9:01] Zero Linden: that said, the pattern of coding for the message system
[9:01] Saijanai Kuhn: knows
[9:01] Zero Linden: is regular enough that the structure of messages comes through pretty clear
[9:01] Zero Linden: it is the suble semantics that fall through the crachs
[9:01] Zero Linden: *cracks
[9:01] Saijanai Kuhn: Zero is there any direct translation of that xml format to xml available?
[9:01] Gareth Ellison: any chance of releasing those pages?
[9:01] Saijanai Kuhn: xml to html*
[9:01] tx Oh: thinks rfc is the way to go
[9:02] Whoops Babii: I agree tx
[9:02] Gareth Ellison: i find that the wiki is horrendously out of date or incomplete in parts - some of the messag categories are just empty
[9:03] Zero Linden: Sai - yes, there are several XLST systems for converting that to HTML and PDF
[9:03] Gareth Ellison: hmm, saijanai has me thinking - an automated conversion of mesage template comments to an HTML document would rock
[9:03] Zero Linden: see file:///Users/markl/Downloads/rfc2629xslt/rfc2629xslt.html#grammar
[9:03] Zero Linden: whoops
[9:03] Gareth Ellison: i.e put each message with it's comments
[9:03] Whoops Babii: yes?
[9:04] Echo Seigo: lol
[9:04] Whoops Babii: hehehe
[9:04] Tao Takashi: Zero: Well, I just can say that we use doctests wherever we can in Plone development because it makes writing tests easier and it can serve as some sort of documentation already
[9:04] Tao Takashi: so you basically write a story and try to make it work afterwards
[9:05] Saijanai Kuhn: [3]
[9:05] Zero Linden: [4]
[9:06] Zero Linden: Tao - yes - I think of my unit tests in the same way - just that sometimes, for deeper, internal frameworks,
[9:07] Zero Linden: constanting having to express the state in terms of the text the read-eval-print loop would produce I find awkward
[9:07] Zero Linden: I like the unit tests to tell a story that the user of a system can read
[9:07] Jussi Hudson: good evening form above the arctic circle
[9:08] Zha Ewry: Not only awkward, Zero, but actually, misleading, since it's filtered al ittle by the formatting
[9:08] Jussi Hudson: my brain is frozen but hearth is still warm:)
[9:08] Zero Linden: Right - so if the domain *is* text, then it feels more natural to me
[9:08] Rex Cronon: hi. u eat a popsicle?
[9:08] Wyn Galbraith: Greetings Arctic Circle!
[9:08] Jussi Hudson: pse dont let me interrupt
[9:09] Jussi Hudson: i just rezzed and sa w so many ppl
[9:09] Jussi Hudson: blushes
[9:09] Tao Takashi: you can do the same assert checks in doctests
[9:09] Zero Linden: As for releaseing the internal wiki pages - I don't think they are complete enough or even up-to-date enough that engineers feel comfortable doing so
[9:09] Tao Takashi: so I am not sure I can follow you ;-) But anyway, main thing is probably that it's tested :)
[9:09] Dr Scofield: do we have time for a non-document question perchance?
[9:09] Zero Linden: where they have, Pheonix has been quite good about putting them into the public wiki
[9:09] Zero Linden: I notice that our offical internal reference for LLSD is the public wiki version
[9:10] Zero Linden: Well - I do have a current "state of the grid" question for ya'll
[9:10] Zero Linden: Does anyone have a script that definitively demonstrates the recent grid slow down?
[9:10] Zero Linden: of scripts?
[9:10] Zha Ewry: Oh, the stalling, and really odd timing?
[9:10] Dr Scofield: shakes her head
[9:11] Zha Ewry: is it just scripts?
[9:11] Tao Takashi: here's is some example btw: [5]
[9:11] Dahlia Trimble: a friend has a scripted object that says there is lag, but I didnt see any problem in the stats
[9:11] Zha Ewry: I've been on sims where the whole interaction pattern
[9:11] Zero Linden: Yes - I think we've all felt it - even my personal region is experienceing it
[9:11] Zha Ewry: feels slow
[9:11] Zha Ewry: Very long delays on texture loads
[9:11] Zha Ewry: and timeouts on attaching objects
[9:11] Dahlia Trimble: texture downloads are terrible in some regions
[9:11] Dr Scofield: LL's C64 cluster is aging
[9:11] Zero Linden: Right - so it might be a whole-region effec
[9:11] Zero Linden: effect
[9:12] Tree Kyomoon: I havent noticed, but mabey on satellite you are all finally seeing it as I do :)
[9:12] Zero Linden: but if we can find a single script that ehibits it that would be easier to track down
[9:12] Dahlia Trimble: grey avatars everywhere
[9:12] Zero Linden: on the other hand - if you have a personal region that clearly shows the issue we can use simstates
[9:12] Adam Xinpeng: Speaking of texture downloads, pardon the interruption.
[9:12] Dr Scofield: thinks C64 cluster running out of disk drive memory to use
[9:12] Adam Xinpeng: Can we get HTTP Texture downloading enabled?
[9:13] Tree Kyomoon: 2nds http texture downloading
[9:13] Rex Cronon: what exactly does this script need to do, or not to do?
[9:14] Whoops Babii: I've been able to reproduce the texture problem by just rebaking. My avie looks fine when the viewer bakes the texture but when the grid updates it it blurs.
[9:15] Dahlia Trimble: lsl scripts dont have much data available to them to do a good job of pinpointing or measuring lag
[9:15] Zero Linden: Rex - if we had a script that did something and it was measurably slower in 1.19.1 than before ( prefereably chatting time to execute)
[9:15] Free Radar: HUD v1.1 by Crystal Gadgets
[9:15] Zero Linden: that woudl be a great thing to have
[9:15] Whoops Babii: it takes forever for the texture to become right again.
[9:15] Zero Linden: HTTP texture is in the top list for Q2
[9:15] Adam Xinpeng: That means next week right? :P
[9:15] Adam Xinpeng: ;)
[9:15] Zha Ewry: Sweeet
[9:16] Saijanai Kuhn: the beta grid is still on earlier release right? I thouht the havok four sims weren't being updated... So you can still test scripts for that on the main grid, right?
[9:16] Rex Cronon: 1.91.1 is the server version, not viewer. right?
[9:16] Zero Linden: Uhm Q2 starts in three weeks....
[9:16] Samuel Linden: there is both Rex
[9:16] Adam Xinpeng: Pfft!
[9:16] Adam Xinpeng: Next week! *whips*
[9:16] Adam Xinpeng: Hehe
[9:16] Zero Linden: and there's a whole three months to do it, eh?
[9:16] Samuel Linden: the viewer RC for 1-19-1 will be out in the next couple of days
[9:17] Zero Linden: I believe that Havok is currently in het-grid style test on the main grid
[9:17] Zero Linden: Oh - another resource question for you all
[9:18] Saijanai Kuhn: right, using pre 1.19.1 server code so you can still test scripts for comparison on the main grid
[9:18] Tree Kyomoon: march 30=end of Q1
[9:18] Tree Kyomoon: 31st
[9:19] Zha Ewry: March 40th.
[9:19] Zero Linden: I'd like to spend some time talking with someone who has deployed XMPP in a medium to large scale installation
[9:19] Zero Linden: anyone know anyone like that
[9:19] Zero Linden: Saijani - clever
[9:19] Dr Scofield: folks from google must
[9:19] Tao Takashi: Twitter people?
[9:20] Zha Ewry: FFull xmpp?
[9:20] Zha Ewry: Dr. sco? S/t Isn't really full is it?
[9:20] Dr Scofield: s/t is proprietar :-(
[9:20] Zha Ewry: Thought so, but similar
[9:20] Zha Ewry: Sigh
[9:20] Dr Scofield: and can use SIP
[9:20] Dr Scofield: but no XMPP :-(
[9:20] Morgaine Dinova: Zero: the people who market the commercial version of ejabberd would be a good bet, since they ran the 1 million user XMMP tests.
[9:20] Goldie Katsu: I can send out a twitter request.
[9:21] Dr Scofield: would dearly like to listen in ;-)
[9:21] Adam Xinpeng: Google?
[9:21] Tree Kyomoon: bets you could find some folks at www.xmpp.org
[9:21] Adam Xinpeng: Google-Talk is XMPP
[9:21] Dr Scofield: it is
[9:21] Adam Xinpeng: AOL are switching to XMPP too
[9:21] Dr Scofield: yep
[9:21] Tao Takashi: I just sent out a twitter request ;-)
[9:22] Tao Takashi: but more than one can't be bad ;-)
[9:22] Zero Linden: Well, on the face of it I'm not enthused about XMPP as a transport, but since so many people bring it up
[9:22] Saijanai Kuhn: wouldn't Donovan be a likely resource for contacts?
[9:22] Zero Linden: I feel like I should get an understanding about using it in a large scale environment
[9:22] Zha Ewry: blinks as two nearly identical twitters flow by
[9:22] Zero Linden: I do have to say so far I've only heard moaning when I mention it to people who have deployed it
[9:23] Morgaine Dinova: Zero: [6]
[9:23] Tao Takashi: Goldie used her 140 characters more wisely ;_)
[9:23] Dr Scofield: i does look nice...but i'd like to learn about real life deployments as well
[9:23] Dr Scofield: i guess it depends on how many open connections you want/must support
[9:23] Zero Linden: Right - I talked with them about that test
[9:24] Zero Linden: the traffic patterns they tested were considerably different than SL IM usage
[9:24] Zero Linden: And, using 8 Sun Fire™ T2000 servers is no small amount of HW
[9:24] Whoops Babii: why 8?
[9:25] Goldie Katsu: smiles at Tao
[9:25] Dr Scofield: we had a discussion about using XMPP on opensim-dev mailing list and #opensim-dev
[9:25] Whoops Babii: that's a hell of a lot of horsepower.
[9:25] Zero Linden: Also, sigh, Jabber XCP uses prioretary extentions over XMPP
[9:25] Morgaine Dinova: Oh joy
[9:25] Zha Ewry: pitty that
[9:26] Zero Linden: Exactly, and all that - and it was just handling IM traffic --- whereas if we wanted the open grid protocol to use it as a transport
[9:26] Dr Scofield: the consensus so far is that keeping 50K connections open to one server is not going to be the solution
[9:26] Zero Linden: it would be handling much more
[9:26] Tao Takashi: somebody asks what "big" might mean (on twitter)
[9:26] Tao Takashi: or in your words what "large" means :)
[9:27] Saijanai Kuhn: eventually, 40 million to 10 million concurrent logins
[9:27] Saijanai Kuhn: 50 to 100% million concurrent logins
[9:27] Zero Linden: Tao - I'd say supporting 100k active endpoints (where active means engauged in sending data)
[9:27] Morgaine Dinova: Well I think the suggestion to use XMMP was really just to get you out of the firefighting arena with a COTS solution. For the collosal AWG scaling numbers, we'd have to be more creative than that.
[9:27] Zero Linden: and supporting 1M or more inactive (just hanging on presence)
[9:28] Zero Linden: that's "very large" - I'd be interested in folks within an order of magnitude of that
[9:29] Saijanai Kuhn: so this is a short-term solution, eh Zero?
[9:29] Zero Linden: It's just that various people (here and inside LL) keep saying "well, why not XMPP?" and we should all know the answer to that
[9:29] Zha Ewry: its not the IM traffic, per se, that bothers me
[9:29] Echo Seigo: I think Mickal Rmond blogged about this some a few weeks ago
[9:29] Zero Linden: and hey, I'd be happy to learn the answer is "yes!"
[9:29] Dr Scofield: yep, that would be good to find out
[9:29] Zha Ewry: its the pattern, the steady updating, and change and such
[9:29] Zero Linden: Zha - exactly
[9:29] Morgaine Dinova: Sounds like we need another VAG ... "Megascale IM", starting base 10m because that's where we are, and looking only upwards.
[9:30] Dr Scofield: nods
[9:30] Zero Linden: We have to keep a bidirectional flow with each logged in viewer - and on the surface, XMPP looks right (symetric, extensible message based,etc...)
[9:30] Zero Linden: but I don't think in practice it can handle what we'd want to throw at it
[9:30] Zero Linden: Okay all -
[9:30] Zero Linden: it's 9:30
[9:31] Zero Linden: must go to studio directors meeting
[9:31] Dr Scofield: cu
[9:31] Zero Linden: thanks for all comeing
[9:31] Zero Linden: \later
[9:31] Morgaine Dinova: Cya Zero, thanks
[9:31] Tree Kyomoon: anyone have the transcript from 830-9?
[9:31] Saijanai Kuhn: thanks for having us, Zero
[9:31] Dahlia Trimble: thanks Zero :)
[9:31] Adam Xinpeng: Thanks
[9:31] Goldie Katsu: Thanks Zero
[9:31] tx Oh: ahoi
[9:31] Adam Xinpeng: let us know when you post the pdf
[9:31] Adam Xinpeng: I'm very keen to see it
[9:31] Tao Takashi: we all are :)
[9:31] Whoops Babii: Thanks everyone.
[9:31] Rex Cronon: bye zero
[9:31] Tree Kyomoon: XMPP works with flex too by the way

User:Zero Linden/Office Hours/2008 Mar 06

Navigation menu

Search