User:Which Linden/Office Hours/2008 Aug 28

• [11:02] Which Linden: hey sai, how's it goin?
• [11:02] Saijanai Kuhn: pretty good but must leave in about 30 minutes
• [11:02] Which Linden: S'cool
• [11:03] Which Linden: Thanks for putting up a transcript for last week
• [11:04] Which Linden: Also, that was very cool to have the erights dudes here
• [11:04] Which Linden: I asked Zero if he know about them, and he totally had
• [11:04] Which Linden: I guess "dude", singular, it was just Graph
• [11:04] Which Linden: Hey Tara5
• [11:06] Saijanai Kuhn: Yewah, there's another guy in SL who is a big capabiltiies fan. Just reminded him of your OH
• [11:06] Which Linden: Oh... cool, ok
• [11:06] Armon Aeon: Greetings!
• [11:07] Which Linden: Hey Armon
• [11:07] Which Linden: Sheesh
• [11:08] Which Linden: What's on everyone's mind these days?
• [11:08] Which Linden: Sad about the lack of Mono :-(
• [11:08] Saijanai Kuhn: Zero had a bit of discussion about using a more generalized capabilities thing for sharing baked textures via the Agent Domain
• [11:08] Saijanai Kuhn: still on hold? Haven't checked the blog lately
• [11:08] Which Linden: Ah... *pokes at transcript*
• [11:09] Which Linden: Yeah -- they might roll out some more mono regions if the crash rate is down
• [11:09] Cel Edman: Hi all :)
• [11:09] Saijanai Kuhn: me, I'd like to see the current SL group IM brought into the AD to help establish community and to see what happens to scaling when you distribute parts of the load via the AD isntead of the sim/central chat server
• [11:10] Saijanai Kuhn: hey Cel
• [11:10] Which Linden: Oh I see
• [11:11] Which Linden: Yes, definitely the group IM would see vast improvements if refactored in any way, especially if refactored into AD
• [11:11] Saijanai Kuhn: was talking to Zha about that. Kinda use the email model. The server forwards messages to the AD, which descdes if the avie is online and whatnot
• [11:11] Which Linden: Honestly we should be reconsidering the entire IM infrastructure
• [11:12] Saijanai Kuhn: so the server just needs to track the AD associated with the members
• [11:12] Which Linden: What's "the server"? The IM server?
• [11:12] Saijanai Kuhn: right
• [11:12] Which Linden: Oh... I see...yeah, that is a lot like email
• [11:12] Cel Edman: yep the message capped is kinda strange somehow
• [11:12] Saijanai Kuhn: and the AD tracks the avie and makes decisions about delivery
• [11:13] Which Linden: Right -- it could decide to not deliver it
• [11:13] Which Linden: Well -- here's the thing -- right now there's a major difference between P2P IM and group IM, in that the latter is never delivered when you're offline
• [11:13] Which Linden: So, yes, refactoring that so that they're both delivered in the same fashion gives more flexibility to the architecture
• [11:13] Saijanai Kuhn: right. WHich may or may not be a good thing, but the AD knows when an avie is online or not anyway
• [11:14] Cel Edman: same time when i log on, and instant crash, lucky i log all my conversations. but gotta check it in my logfolder.
• [11:15] Which Linden: The question remains whether we'll use an off-the-shelf IM server or continue to roll our own
• [11:15] Saijanai Kuhn: and an AD might be many different servers behind the scenes but the connection points for the client and chat server need not know this
• [11:15] Which Linden: Yeah, they'd have a uniform interface
• [11:15] Which Linden:  :-)
• [11:15] Which Linden: (love that phrase)
• [11:16] Saijanai Kuhn: thing is wouldn't require much of a redesign to the current system, as far as I can tell. And need not be the perfect solution, just something reasonably quick and dirty that might alleviate many problems even so
• [11:16] Cel Edman: Well I imagine when offline, it could be like the same with e-mail?
• [11:16] Cel Edman: kind of inbox
• [11:16] Which Linden: An inbox, exactly
• [11:16] Saijanai Kuhn: might be, or it could be AD specific. You pay more for better features...
• [11:17] Which Linden: Ha ha, like the ability to have more than 25 queued messages...
• [11:17] Cel Edman: with like a limit of x-kilobyte/megabyte ^^ and spam filters incase
• [11:18] Which Linden: As far as I know, Second Life is one of very few services that offers hybrid realtime/email-time chat
• [11:18] Which Linden: Can't think of any other actualy
• [11:18] Which Linden: Everyone else -- you're either online and receiving messages or not
• [11:18] Which Linden: Or like Facebook the messages are just sorta in an inbox all the time
• [11:19] Cel Edman: when its working Which, when now i`m offline I only get group messages by email, not normal questions of users
• [11:19] Which Linden: Wait, what Cel?
• [11:20] Which Linden: You can get group chat by email?
• [11:20] Which Linden: Or do you mean group *notifications*
• [11:20] Which Linden: which are totally different and strange
• [11:20] Cel Edman: group notifications
• [11:20] Saijanai Kuhn: well they're still part of ImprovedInstantMessage
• [11:21] Saijanai Kuhn: the bane of my SL existence
• [11:21] Cel Edman: Well I got send im to email checked. And before when people asked me sometimes when i was offline..
• [11:21] Cel Edman: i got a notice by email
• [11:21] Cel Edman: now not anymore
• [11:21] Saijanai Kuhn: ah, that's personal IM
• [11:22] Which Linden: Sai: at this point, ImprovedInstantMessage is just "send generic packet to viewer"
• [11:22] Which Linden: Cel: that's ... interesting that they stopped
• [11:22] Saijanai Kuhn: I know, but the ins and outs of populating it with the right data at the right time still are beyond me. That's why my chatbot in python never happened
• [11:22] Cel Edman: maybe I gotta uncheck all then check all again
• [11:23] Which Linden: Might work
• [11:23] Saijanai Kuhn: love that universal checkbox bug
• [11:23] Which Linden: Sai: yeah, I guess my point is though that things go via that message that have nothing to do with IM at all
• [11:23] Which Linden: There are at least three different backends for sending messages, probably more, and they all comunicate with the viewer through the narrow straw of ImprovedIM
• [11:23] Saijanai Kuhn: sure. But in fact, I think all of them are still AD related
• [11:24] Which Linden: Yes, that is true
• [11:24] Saijanai Kuhn: so break them up or not, they're still going to be using the same distribution model, I would think
• [11:24] Which Linden: We need bracelets: "What Would the Agent Domain Do?"
• [11:24] Armon Aeon: LOL
• [11:25] Which Linden: In an ideal world, the AD would host a "send a message to this user, with the following delivery rules"
• [11:25] Which Linden: And the rules could include "store in inbox/email if you support that"
• [11:26] Saijanai Kuhn: Thats another of those thorny issues. ZEro likes the 1:1 correspdonance of the AD and Asset Server. Zha likes the many to many concept that OpenSIm is heading for
• [11:26] Which Linden: Hm..
• [11:27] Which Linden: I guess I don't have a strong opinion on the asset server
• [11:27] Which Linden: I kinda think that the asset server is a red herring, actually
• [11:27] Saijanai Kuhn: well, its more flexible to do it the mant to manyway, but I'm sure its simpler with 1 to 1
• [11:27] Saijanai Kuhn: how so?
• [11:27] Which Linden: Well it's just a file store
• [11:28] Saijanai Kuhn: with certain access keys (?) built into the design
• [11:28] Which Linden: You only know it's there because the urls you're given start with "asset.agni"
• [11:28] Saijanai Kuhn: maybe inventory server would be a better term
• [11:28] Which Linden: Ahhhhh
• [11:29] Which Linden: Right -- so the asset server is just big dumb file hosting -- the interesting questions are -- who gets to see the urls of your inventory
• [11:29] Saijanai Kuhn: but even so, you still are left with the concept of the Central Database that holds all things for the avies in one place
• [11:30] Saijanai Kuhn: But is it more realistic to assume that different assets might be stored with different companies?
• [11:30] Saijanai Kuhn: Bob's Bargain PRims Assets vs the assets created on the LInden servers
• [11:30] Cel Edman:  :)
• [11:30] Which Linden: So -- are you talkign about asset data or inventory here?
• [11:31] Which Linden: "inventory" being a url referencing the asset data
• [11:31] Saijanai Kuhn: good question.
• [11:31] Which Linden: I think it's clear that inventory resides in the AD
• [11:31] Saijanai Kuhn: I would think the AD tracks the urls while the various data can be stored anywhere
• [11:32] Which Linden: Yeah...ok, so what if they're stored on a server you have no access to?
• [11:32] Which Linden: Via firewall rules and whatnot
• [11:32] Which Linden: That would basically just be a abroken asset
• [11:32] Saijanai Kuhn: that scenario might happen for any design, though...
• [11:32] Which Linden: Usable in whichever RD has access to the firewalled asset server but nowhere else
• [11:33] Saijanai Kuhn: sounds reasonable to me.
• [11:33] Which Linden: Such a thing is certainly possible (in fact, that's how things are with our own asset server)
• [11:34] Saijanai Kuhn: so things that aren't accesible aren't rezzed and if we're lucky get greyed out int he interface
• [11:34] Which Linden:  :-) Exactly
• [11:35] Cel Edman: Sorry people gotta go, some work todo
• [11:35] Armon Aeon: see you Cel..
• [11:35] Saijanai Kuhn: alters cel. BTW, we're getting closer to getting pyogp useable
• [11:35] JayR Cela: byee Cel :_)
• [11:37] Saijanai Kuhn: what I'm hoping with pyogp is to get things to the point where we can be the official testers for the OGP before they get released for general beta testing. WOuld save a lot of time in the long run
• [11:38] Which Linden: Yes, would love that too
• [11:38] Saijanai Kuhn: Tao's got a Plone-based Agent DOmain on a webpage right now. IF we can get that to be part of hte official OGP test suite, would be much fun
• [11:38] Which Linden: Cool!
• [11:38] Armon Aeon: yay!
• [11:39] Which Linden: I've seen some of his commits flying by
• [11:39] Saijanai Kuhn: [1]
• [11:39] Which Linden: Definitely it's key to have a serious test suite and simple, readable example code
• [11:40] Saijanai Kuhn: Enus has been quite strickt about having unit tests for almost everything. Unforntuately, the scheduling the last week has made thigns a bit out of joint
• [11:40] Which Linden: Ha ha, love the "proof of concept" sticker
• [11:41] Which Linden: I also am quite a unit test nazi when writing code
• [11:41] Saijanai Kuhn: never done it in my life :-/
• [11:42] Which Linden: It's really helpful to be able to verify code you just wrote, piece by piece
• [11:42] Saijanai Kuhn: got my start with early 80s methodology at a 2 years college. And we're int eh backwoods anyway
• [11:42] Which Linden: Saves debugging time later.
• [11:42] Saijanai Kuhn: sure, I understand the concept. Just never got into the practice
• [11:42] Which Linden: Yeah, continuous unit testing is a very "agile" concept.
• [11:42] Armon Aeon: yes, that would be great
• [11:42] Which Linden: It's not hard to get into it, at least in python
• [11:42] Which Linden: More or less difficult in other language
• [11:43] Saijanai Kuhn: I've been trying to get how its working. With Lock gone, I need to start coding cause he was the other "go getter" for the pyogp
• [11:43] Which Linden: Hah!
• [11:43] Saijanai Kuhn: and, its a good discipine to have, obviously
• [11:44] Which Linden: Well... one easy way to get started is to add a single test case in an existing test suite
• [11:44] Which Linden: And just copy some other test case as a starting point
• [11:45] Saijanai Kuhn: yeah. I sorta see how the unit tests thing works in Python. Have to read up on the details though
• [11:45] Saijanai Kuhn: or is it doc test... darn
• [11:45] Which Linden: Just be like "ok, this one tests for when pi is 3.14, I'll try setting pi to 3.5 in this new test"
• [11:46] Saijanai Kuhn: sure. its the syntax of describing the tests that I don't get mroe than anything else
• [11:46] Disconnected from: in-world Voice Chat
• [11:46] Saijanai Kuhn: I mean THAT kidn of testing I ometimes do anyway in my hacked code
• [11:46] Which Linden: Did you have to run, btw? I recall you saying something like that
• [11:46] Saijanai Kuhn: its just the uniform syntax I don't get
• [11:47] Saijanai Kuhn: yeah just checked schedule. was 1:45 no 11:45
• [11:47] Saijanai Kuhn: so I'm good, thanks
• [11:47] Which Linden: k
• [11:47] Which Linden: What do you mean by "uniform syntax"?
• [11:47] Saijanai Kuhn: BTW, do you ever use GUIs in python?
• [11:47] Which Linden: No, never have
• [11:48] Which Linden: Never even used an IDE
• [11:48] Which Linden: though I hear good things about Komodo
• [11:50] Saijanai Kuhn: well, you have this thing where you describe a test in plain text and then give a >>>> command below it. Not sure how that actually is set up (for example)
• [11:52] Which Linden: Oh, doctests
• [11:52] Saijanai Kuhn: yeah, get them confused (obviously need much more reading)
• [11:52] Which Linden: Yeah you can see one here: [2]
• [11:53] Which Linden: The way it's run is a little snippet of boilerplate at the bottom of the file:
• [11:53] Which Linden: import doctest
• [11:53] Which Linden: doctest.testmod()
• [11:53] Which Linden: then you just run the file
• [11:54] JayR Cela: well / thanks all
• [11:54] JayR Cela: see ya :_)
• [11:54] Which Linden: Thanks for stoppin' by JayR!
• [11:54] Armon Aeon: laterz..
• [11:54] JayR Cela: :_)
• [11:55] Saijanai Kuhn: later JayR
• [11:55] Which Linden: [3]
• [11:56] Saijanai Kuhn: ah ok thanks
• [11:57] Which Linden: The other way of doing tests in python is unittest
• [11:57] Saijanai Kuhn: so anything with >>> is executed as if it were run from the console?
• [11:57] Which Linden: Which has a different syntax
• [11:58] Which Linden: Some people vastly prefer one over the other -- me, I don't care so muuch
• [11:58] Saijanai Kuhn: well I can certainly see this would be good for simple testing
• [11:58] Which Linden: Sai: yeah, that's the idea, and then it checks the following line to see if the results match
• [11:58] Saijanai Kuhn: yeah.
• [11:59] Which Linden: Probably... again... my home internet
• [11:59] Saijanai Kuhn: but for more complext tests, I assume unittest is used?
• [11:59] Which Linden: Crap man I have mega chat lag here
• [11:59] Which Linden: I should maybe mess with my QoS settings
• [11:59] Armon Aeon: yes.. it's been laggy lately..
• [11:59] Saijanai Kuhn: yeah I've had chat loag for so long I thought client crashed
• [12:00] Armon Aeon: har!
• [12:00] Saijanai Kuhn: then I get 5-10 chat liens i a row
• [12:00] Which Linden: For more complex tests, yeah, unittest is used. The most major benefit from my perspective (and I know this is possible with doctest but haven't figured out how to do it) is setup/teardown methods that run before and after each test, setting up state that's common to all tests
• [12:00] Saijanai Kuhn: was bad because I was doing transcript for two meetings at the same time. Lost both
• [12:00] Armon Aeon: my personal favorite is when my chat is delayed for like over 90 seconds...
• [12:01] Which Linden: I'd say we might be having network problems but I'm getting presence notifications all this time
• [12:01] Which Linden:  :-(
• [12:01] Saijanai Kuhn: well its something I gotta learn fo rth pyogp project, and for my own use as well, so its all good
• [12:02] Which Linden: Maybe the Illuminati are intercepting my chat and subtly inserting fnords
• [12:04] Saijanai Kuhn: wouldn't be surprised, Thisis SL aftterall..
• [12:04] Which Linden: I gotta meet a friend for lunch -- so I'll see you all later? *waits for another 5 minutes for all chat to flow*
• [12:04] Saijanai Kuhn: enjoy and thanks for stopping by
• [12:04] Armon Aeon: aaahhhh..
• [12:04] Saijanai Kuhn: Want me to post transcript or will you
• [12:04] Which Linden: I got it
• [12:04] Saijanai Kuhn: KK
• [12:04] Which Linden: Thanks for the offer though. :-)