User:Which Linden/Office Hours/2008 Oct 2

From Second Life Wiki
< User:Which Linden/Office Hours
Revision as of 14:12, 2 October 2008 by Which Linden (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
  • [11:05] Which Linden: hiya party ppls
  • [11:05] Cenji Neutra: hey
  • [11:05] FWord Utorid: hola which
  • [11:05] Saijanai Kuhn: hey dude
  • [11:06] Saijanai Kuhn: thanks for the pointers last week, Which. Thinking I'm starting to get things close to working
  • [11:06] Which Linden: Awesome
  • [11:06] Saijanai Kuhn: now if I can just convince Zero that implementing the current IM in OGP as a shortterm solution, is the easiest way to go...
  • [11:06] FWord Utorid: the opensim people care about that too, sai
  • [11:06] Which Linden: heh ... zero's not quite the one for "the easiest way to go"
  • [11:07] FWord Utorid: this morning in opensim-dev they were talking about doing an irc maneuver for chat
  • [11:07] Saijanai Kuhn: sure, but the upside is that you can test lots of different issues with the existing system without having to change the client GUI
  • [11:07] FWord Utorid: of course they are determined to not let zero tell them what to do, but they will fall in line if they want to stay compatible
  • [11:07] FWord Utorid: tess was talking xmpp but that's an unknown too
  • [11:08] Saijanai Kuhn: inventory, group IM, etc., all goes through that. Trying to rearrange them to go through some other protocol is going to be hard word due to the fragile GUI thing
  • [11:08] FWord Utorid: well, that's kind of things that you and I are doing
  • [11:08] FWord Utorid: hopefully producing gui alternatives
  • [11:08] Saijanai Kuhn: and ifyou want to test the trust issues and so on you need a working system, no matter how clunky
  • [11:08] FWord Utorid: although inventory isn't so important to me
  • [11:08] Cenji Neutra: I read that someone said that XMPP won't scale - though my uninformed view is that it should
  • [11:08] Which Linden: but you gotta admit that pipe-separated bodies with an integer enum describing the class of communication is not forward-thinking
  • [11:09] Which Linden: if we don't kick the current IM system for OGP, we're gonna be saddled with it forever
  • [11:09] Saijanai Kuhn: oh, well, yeah. Not talking about anything but getting SOMETING working now, instead of trying to mod the GUI for each new part
  • [11:09] FWord Utorid: which, i don't think irc will cut it, it's not extensible enough
  • [11:09] FWord Utorid: it might scale, but it's very limited in scope
  • [11:09] Saijanai Kuhn: this isn't meant to be the official OGP IM, just something to get us started with the current viewr GUI, which is a pain to mod at that level
  • [11:10] FWord Utorid: xmpp sounds like it could pull it off, but you're doing ground up work there.
  • [11:10] Which Linden: I guess I don't understand, then, Sai: why can't you just start using the current IM system then?
  • [11:11] Saijanai Kuhn: well, the current IM system isn't hooked up in the OGP viewer
  • [11:11] Which Linden: FWord: yeah..... I'm happy that I'm out of the debate of "what would be the best chat protocol" because it sounds hairy
  • [11:11] Which Linden: Sai: oh
  • [11:11] FWord Utorid: which: yeah, the dust has to settle on it I think
  • [11:11] FWord Utorid: sai: if improvedinstantmessage packets and chat packets are shipped, you can get at what you need
  • [11:12] Saijanai Kuhn: my suggestion was to have teh Agent Domain pose as a sim to the IM server and let teh AD sort out how/when to send things to each agent. And, the only real change after that is to get the IIM packet working with the CAP instead of UDP
  • [11:12] FWord Utorid: if you have an event model, you could maneuver the event to abstract away whatever protocol is going to be used.
  • [11:12] Which Linden: You could just mess around with packets to the sim itself, the "true legacy" route
  • [11:12] Cenji Neutra: What is the 'OGP viewer'? Isn't the python stuff just meant as a test-bed implementation of the OGP for testing?
  • [11:12] Saijanai Kuhn: FWord, right. And that would be the next step after getting *something* working
  • [11:12] Which Linden: I think the OGP viewer is like the regular viewer, just OGP-enabled
  • [11:12] Saijanai Kuhn: OGP viewer is what gridnauts use to TP around
  • [11:12] Cenji Neutra: ok
  • [11:13] Which Linden: PyOGP is different
  • [11:13] Cenji Neutra: ah
  • [11:13] Which Linden: (though I guess Sai is working on making a viewer out of PyOGP)
  • [11:13] Cenji Neutra:  :)
  • [11:13] Saijanai Kuhn: we can do anything we want with pyogp, but no 3D view possible for a LONG time (if ever)
  • [11:13] FWord Utorid: what keeps 3d view out of the possibilities?
  • [11:14] Saijanai Kuhn: not sure that Python is up to handling the graphics stuff external to OpenGL
  • [11:14] Saijanai Kuhn: night end up with a 1 fps viewer
  • [11:14] Which Linden: There are frameworks for that
  • [11:15] Which Linden: PyOGRE is actually not bad
  • [11:15] Saijanai Kuhn: don't thnk its the rendering speed that would be the issue. More the higher-level stuff
  • [11:16] Saijanai Kuhn: Gareth put a python wrapper around llvolume.cpp but not sure if that is sufficent to get things speedy enough
  • [11:16] FWord Utorid: idk. if they could reduce the memory usage in the viewer so it doesn't crash so frequently, i would be pleased. but, of course, i want to make c# 3d viewers of different kinds
  • [11:17] Saijanai Kuhn: if all else failed, you could put a wrapper about all the graphics subsystem libs, that might work, but not familiar with how 3D game engines work at that level
  • [11:17] Which Linden: Yeah, me neither
  • [11:17] Saijanai Kuhn: anyway, I'm just promoting getting IIM ported to the OGP viewer as a test case for group IM, not as the "real" group IM
  • [11:18] Saijanai Kuhn: we knowit works. So what can we do to improve the protocol. Having it all stream in via the AD EQG cap is nice, but are their gotchas to even doing that?
  • [11:19] Saijanai Kuhn: anyway, the gist of my conversation with Zero this morning
  • [11:20] Saijanai Kuhn: soI'm all talked out.
  • [11:20] FWord Utorid: i don't have much to add, i just came for the grass
  • [11:21] Which Linden: Heh
  • [11:21] FWord Utorid: well, i do want to know more about that big happy diagram behind us for L$
  • [11:21] FWord Utorid: but something tells me that is a slow mover.
  • [11:21] Which Linden: Oh... do you want the tour of it?
  • [11:21] Saijanai Kuhn: its called bamboo f0wrd. Not sure if its safe to smoke
  • [11:21] FWord Utorid: sai, are you suggesting I smoke questionable substances? :O ;)
  • [11:21] Which Linden: I mean, we hashed that thing over almost a year ago
  • [11:21] Which Linden: Heh I said "hash"
  • [11:21] FWord Utorid: which, I missed it, maybe there was a log of the talk
  • [11:21] FWord Utorid: I noticed.
  • [11:21] Saijanai Kuhn: he he he he he said hash
  • [11:22] Saijanai Kuhn: he heh heh*
  • [11:22] FWord Utorid: I am just interested in my own little microenterprise of sorts
  • [11:22] FWord Utorid: and if LL is gearing up to offer that sort of thing in conjunction with opensim, it would be cool
  • [11:23] Which Linden: It's a slow-moving project because of all the backend work we have to do in order to start supporting it ourselves
  • [11:23] Which Linden: BTW, previous discussion: [1]
  • [11:24] FWord Utorid: thanks. that will keep you from having to ... re-hash
  • [11:24] FWord Utorid: oh, ok
  • [11:24] Which Linden: Oh that's not a good one
  • [11:24] FWord Utorid: don't bogart on the log, bro
  • [11:25] Which Linden: Crap yo can't find a good transcript
  • [11:25] Which Linden: ah: [2]
  • [11:26] Which Linden: also: [3]
  • [11:26] Saijanai Kuhn: well, if Tao can get his AD working withthe new protocols, I'm wondering if you canstart testing things using that. COmpletely unsecure bogus money server for testing in OGP viewers
  • [11:26] Which Linden: Yeah -- could definitely do stuff in parallel
  • [11:26] FWord Utorid: thanks, which. that will go into the read later column, after the hours of contemplating Dahlia.
  • [11:27] Which Linden: The thing is, Sai, the viewer doesn't have much to do with this, it's mostly an AD thing
  • [11:28] Which Linden: Well I take that back -- the viewer initiates transactions
  • [11:28] FWord Utorid: hmm... one thing I think about these new school IM transports... one could make a proxy of sorts and pretend the sims / servers are all old school for now
  • [11:28] Saijanai Kuhn: right. Though, I would THINK it would be easier to modify Tao's AD scripts than the "real thing"that LL is using for Aditi AD
  • [11:29] Saijanai Kuhn: though, I guess those are mostly python scripts too anyway
  • [11:29] Which Linden: Yeah, they're just python
  • [11:29] Which Linden: FWord yeah
  • [11:30] Saijanai Kuhn: Fword that was my idea. Just have teh AD fake being a sim that all the avies happen to be sitting on
  • [11:30] Saijanai Kuhn: and the IM server just talks to the AD like any other sim
  • [11:31] Saijanai Kuhn: just the avies neer go anywhere, as far as the IM server is concerned
  • [11:31] FWord Utorid: hmm... sai, i think the idea is similar though not necessarily the same. I am proposing something that would keep the existing viewer compatible but allowing a bypass for OGP-Powered Testosterone Machines.
  • [11:32] Saijanai Kuhn: oh, you mean ignore OGP and just implement thecurrent IM as is. Gives you IM and such, but doesn't help the OGP planning/design which is the otehr half of doing it
  • [11:32] Saijanai Kuhn: zero's original idea was to send IM through the AD in some way.
  • [11:32] FWord Utorid: sai, what i mean is, a proxy or 'dual delivery system' which allows separation of church and state
  • [11:32] FWord Utorid: church being the old school IM system, state being the OGP way of cool
  • [11:32] Dahlia Trimble: lol
  • [11:33] FWord Utorid: when the viewer logs in, it sends it's old school charm to the sims, and it gets old school messages. OGP sends a special 'I am better than him' byte and gets the new.
  • [11:33] Saijanai Kuhn: well, not sure how muchdifference that is, except that we're trying to get away from the ims handling normal group IM at all
  • [11:34] Which Linden: How, in general, does OGP handle "legacy"?
  • [11:34] FWord Utorid: I am just trying to address the issue which seems to be a primary obstacle, which is trying to have corvettes on the train tracks.
  • [11:34] Saijanai Kuhn: well,, then the sims are all dealing with the AD as different avies at the same time
  • [11:34] Which Linden: ha ha corvettes on the train tracks -- nice
  • [11:34] FWord Utorid:  ;)
  • [11:35] Saijanai Kuhn: Which, right now, the only parts of OPG are login and TP. to establish original presence on a sim/grid. For walking between sims, legacy isused. For local chat, legacy is used
  • [11:35] Saijanai Kuhn: Ithink for TP between sims on the same grid, new TP is also used
  • [11:35] Saijanai Kuhn: but not sure
  • [11:36] Saijanai Kuhn: it might be old school for intra-grid
  • [11:36] Saijanai Kuhn: the group IM/inventory/etc stuff is simply not hooked up right now
  • [11:37] Which Linden: So ... it seems that for OGP one challenge will be fallback ability. E.g. if you have your OGP-enabled viewer but are trying to walk around in Old Town
  • [11:37] Dahlia Trimble: are there any forms of documentation publically available on the design/implementation of the current group IM system? it seems to be a black box to me
  • [11:38] Which Linden: It's kind of a black box to us too
  • [11:38] Dahlia Trimble: lol
  • [11:38] Saijanai Kuhn: So I'm think we let teh AD pretend its a sim to the IM server, and let the AD squirt the IM info to the viewer viia the Event Queue cap
  • [11:38] Saijanai Kuhn: those would be the first two changes to group IM. THEN you can start debating which protocol(s) are better and how to get them into the AD
  • [11:38] Saijanai Kuhn: and then to the viewer
  • [11:39] Which Linden: Sai: that's actually a lot of work to implement
  • [11:39] Dahlia Trimble: I think you could let the client talk directly to the IM server and remove 2 extra points of failure
  • [11:39] Which Linden: The sim code implementing current IM is nontrivial, and snipping that out would be similarly nontrivial
  • [11:39] Saijanai Kuhn: the biggest holdup with the SL viewer is that the IM panels are tied directly to the IM protocols
  • [11:39] Saijanai Kuhn: well, you juts have the AD pretend its a sim...
  • [11:40] Which Linden: Isn't the notifications project addressing that to some degree?
  • [11:40] Saijanai Kuhn: possibly
  • [11:40] Dahlia Trimble: wouldnt have to take te IM code out of the sim, just ignore it
  • [11:40] Which Linden: The AD would have to proxy the packets to the sim
  • [11:40] Saijanai Kuhn: but doesn't that require us to wait until the noficiation project is complete before testing IM, inventory, etc etc
  • [11:40] FWord Utorid: I love unsolvable problems.
  • [11:41] Saijanai Kuhn: don't think its unsolveable.
  • [11:41] Which Linden: I dunno if we have to wait for the project, but it might improve things
  • [11:42] FWord Utorid: this notification project, that's about the interesting flash animation for the corner dialog that was out a while ago?
  • [11:43] Saijanai Kuhn: sure, but I don't think the transition from not having to having would be very difficult. The IIM packet would be coming over teh CAP instead of UDP. That's the only real change (I think)
  • [11:43] Which Linden: Hey, random question for those who are in groups that sell objects: what the frack is supposed to go in the 'Sales' tab of the group info floater?
  • [11:43] Saijanai Kuhn: if notification needs to send something it can get its info from the CAP and send it to the IM panels just as it would for UDP packets
  • [11:43] Dahlia Trimble: land sale accounting?
  • [11:44] Saijanai Kuhn: pretty sure that was part of hte Dia LIberaccion design intent anyway
  • [11:44] Which Linden: Dahlia: does it have doubled-up entries for each transaction?
  • [11:44] Dahlia Trimble: cant remember, been a while since I sold any land
  • [11:44] Saijanai Kuhn: thught notification project was to get the GUI factored into haveing an event queue instead of hardwired handlers for each event/message
  • [11:44] Which Linden: Sai: that's true, converting to event queue should be straightforwrad
  • [11:45] Which Linden: Dahlia: ok, thanks, was just curious
  • [11:45] Saijanai Kuhn: so transform it from teh tk model to something more OOP
  • [11:46] FWord Utorid: oic
  • [11:46] Dahlia Trimble: Can you deed an object to group and have it sell? I've never tried
  • [11:46] Saijanai Kuhn: but maybe I'm confusing projects here
  • [11:46] Which Linden: Dahlia: I think so, but also, haven't tried
  • [11:46] Which Linden: Sai: you could be right
  • [11:46] FWord Utorid: it's a mystery
  • [11:47] Which Linden: Looking at the project page, no, it doesn't go that far
  • [11:47] Which Linden: It basically is just a viewer rearchitecture
  • [11:48] FWord Utorid: I didn't think it would need to be a protocol change
  • [11:48] FWord Utorid: the visuals given off were enough that the way the packets come in could handle it
  • [11:48] Which Linden: Well, not to imply that a viewer rearchitecture is a trivial thing to do, just that it doesn't really touch the protocol level at all
  • [11:48] FWord Utorid: the viewer is the thousand pound gorilla
  • [11:49] Saijanai Kuhn: the viewer GUI isthe 900 pounds of fat
  • [11:49] Which Linden: You let it sit wherever it wants in the movie theater?
  • [11:49] FWord Utorid: well, pretty much
  • [11:49] Saijanai Kuhn: do you have a link to that page, Which. COuldnt' find it via google
  • [11:49] FWord Utorid: i am sure if there was a viewer that was easily modifiable then the server would be more flexiprim
  • [11:50] FWord Utorid: but it's quantum entanglement between both ends of the phone and the phone company
  • [11:50] Saijanai Kuhn: the viewer GUI is basically tk done in C++ and OpenGL with some stuff bolted on at odd places
  • [11:50] Saijanai Kuhn: like folders and IM panels
  • [11:50] Which Linden: [4]
  • [11:50] FWord Utorid: yeah, they did an interesting thing coming up with the XMLish UI
  • [11:50] Which Linden: I think. We have a different page in the internal wiki
  • [11:51] FWord Utorid: just open the internal wiki for us to look at and we'll take it from there.
  • [11:51] Saijanai Kuhn: ah OK. Ithink this is differnt than what I was talking about. There's some mention of event factoring in teh viewer roadmap (I think)
  • [11:52] Which Linden: heh
  • [11:52] FWord Utorid: bummer. i was hoping I could ... bamboozle Which into giving us sekrit access.
  • [11:53] Which Linden: Classic
  • [11:53] FWord Utorid:  ;)
  • [11:54] Which Linden: So yeah, the impression I get is that it's mostly a local-event refactoring, treating the network as a mysterious source of messages
  • [11:54] Saijanai Kuhn: ah, the UI callback slink is red, which means it doesn't even have a stub page :-(
  • [11:54] FWord Utorid: it looks like there is a rotating server over there with an escrow demo
  • [11:54] Which Linden: Yeah, there is
  • [11:55] FWord Utorid: for sale for L$0
  • [11:55] FWord Utorid: i bought one
  • [11:55] Which Linden: I kinda forget how to kick it off though
  • [11:55] Saijanai Kuhn: thing is, I was learnign tk for python, and it seemed like the entire basic viewer GUI is a tk paort.
  • [11:55] Which Linden: Have to take a look at the script
  • [11:55] Saijanai Kuhn: so, whatever strategies that have been used to make tk more OOPish (event handlers) could be used to make the SL GUI more oopsish
  • [11:56] Saijanai Kuhn: and I know there's at least one lib that does this (I think I know)
  • [11:56] Which Linden: there we go
  • [11:56] FWord Utorid: yeah, interesting
  • [11:58] FWord Utorid: i need more coffee
  • [11:58] Which Linden: Yeah, me too
  • [11:58] FWord Utorid: after years of web development i am still trying to find something that makes php form handling happy
  • [11:59] FWord Utorid: i think it does not exist
  • [11:59] Saijanai Kuhn: is still learnign PHP
  • [11:59] Which Linden: Man, don't leanr php
  • [11:59] FWord Utorid: i want to make some sort of a class to contend with it and do validation and so on, but NooOoOoOo 10,000 years have passed and the 2nd ice age has come
  • [12:00] Saijanai Kuhn: my current job (such as it is)requires php. I know *some* but not nearly enough
  • [12:00] FWord Utorid: php is good for web apps afaic. i am sure there are other ways to do it, but none currently that I like
  • [12:00] Which Linden: [5]
  • [12:00] FWord Utorid: apache2 and mysql are happy enough together to get the job done
  • [12:01] Which Linden: IMO, mod_wsgi and python is the platform to be gunning for
  • [12:01] Dahlia Trimble: lol I've always used python for cgi, and I thought I was an outcast
  • [12:02] Which Linden: hah, well, wsgi is nicer than cgi
  • [12:02] Which Linden: ok, hey, time for lunch and stuff
  • [12:02] Saijanai Kuhn: I dont' get it ... LOL
  • [12:02] FWord Utorid: which, a db platform to go with that is?
  • [12:03] FWord Utorid: enjoy your lunch.
  • [12:03] Which Linden: FWord: well we're a mysql shop here
  • [12:03] Saijanai Kuhn: Oh, truth test vs string equality?
  • [12:03] Which Linden: [6]
  • [12:03] Which Linden: OK, thanks all for showing up
  • [12:03] FWord Utorid: ok. learning python is on my list anyway, since i plan to eventually try to get a job at your place
  • [12:03] FWord Utorid: take care
  • [12:03] Dahlia Trimble: thanks :)
  • [12:04] Which Linden: Appreciated this conversation
  • [12:04] Saijanai Kuhn: python is fun once you get the hang of it with a pretty formatter
  • [12:04] Saijanai Kuhn: thanks which
  • [12:04] Dahlia Trimble: poofs... bye all :)