User:Zero Linden/Office Hours/2008 August 28
Jump to navigation
Jump to search
- [8:36] Tara5 Oh: hi zero
- [8:36] Zero Linden: hello
- [8:36] Zero Linden: ya know
- [8:37] Tao Takashi: well, I am not a fan of such traditional conferences though
- [8:37] Tara5 Oh: we were just discussing LA - will you be coming?
- [8:37] Zero Linden: if you TP here and quick turn to look at that bean stalk across the way
- [8:37] Inara Marquette: Hi Zero :)
- [8:37] Inara Marquette: Yeah isnt that Kates place?
- [8:37] Zero Linden: then the brid objects look way cool as the invisible prims they are attached to are still visible for a while
- [8:37] Zero Linden: I always think --- wow what is that great piece of kinetic art....
- [8:38] Zero Linden: ...then it becomes just the birds
- [8:38] Cadis Blackadder: the day is full of birds
- [8:38] Inara Marquette: its the little things in SL :)
- [8:38] Cadis Blackadder: :)
- [8:39] Cadis Blackadder: (sorry I know... obscure reference)
- [8:39] Zero Linden: Sorry about Tuesday - unavoidable all day strategy planning session at Linden
- [8:40] Inara Marquette: Cadis I thought it was profound :)
- [8:40] Cadis Blackadder: :) (Kate Bush)
- [8:40] Zero Linden: So...........
- [8:40] Zero Linden: What's on people's minds?
- [8:41] Tao Takashi: vacation ;-)
- [8:41] Cadis Blackadder: haha
- [8:41] Tara5 Oh: just recovering from vacation
- [8:41] Zero Linden: I've spent the last week with my head in the lofty smog that is corporate strategy
- [8:41] Zero Linden: so I've forgotten all my bits....
- [8:41] Tara5 Oh: two weeks in the Santa cruz mountains no wifi
- [8:41] Cadis Blackadder: ykes
- [8:41] Zero Linden: though I did have a short fun time talking with someone about how to implement an LLSD class in Java....
- [8:42] Xugu Madison: sleep....
- [8:43] JayR Cela: hmmm a NetBeans plug in would be nice for that mabey
- [8:43] Xugu Madison: I've got an LLSD class in Java....
- [8:43] Zero Linden: publish it!
- [8:44] Tao Takashi: maybe a collective svn is a good idea
- [8:44] Xugu Madison: I'll wrap it up and see if I can pry it from the framework it's in, over the weekend
- [8:44] Zero Linden: I'm not touching that....
- [8:45] Tao Takashi: like we have for Plone, just as a place where people can put their projects
- [8:45] Tao Takashi: finished or not
- [8:45] Tao Takashi: most will probably never finished ;-)
- [8:46] Zero Linden: One wonders what the stats on levels of completeness are on SourceForge
- [8:46] Tao Takashi: it also might depend on the definition of completeness
- [8:46] Tao Takashi: but you might start with "it runs"
- [8:46] JayR Cela: lol
- [8:47] Tao Takashi: I actually should continue on my blog post on trust and the stuff Bartholomaeus and I discussed last friday
- [8:47] JayR Cela: so whats up with the new server rollout
- [8:47] Tao Takashi: unfortunately some on-site client appointment came up this week
- [8:47] JayR Cela: seems to bee taking a long time
- [8:48] Rex Cronon: hello everybody
- [8:48] JayR Cela: hi there Rex :_)
- [8:48] Rex Cronon: hi
- [8:49] Zero Linden: well, there was a crash mode that only appeared on the main grid.
- [8:49] Rex Cronon: .
- [8:49] Zero Linden: The source was found and fixed and yesterday they redeployed
- [8:49] JayR Cela: ahhh / kool :_)
- [8:49] Zero Linden: I think we'll know this morning if the crash rates are back to normal
- [8:49] Cadis Blackadder: oh good
- [8:49] Xugu Madison: You happy enough if I just publish the LLSD code as a "Here's a start, but it isn't ready to really do much" example for now? GPLv2 okay?
- [8:49] Zero Linden: for the curious: and STL error involving improper handling of the end() iterator
- [8:50] Zero Linden: Xugu - sure - the most important thing is to make your license intention clear: public domain, BSDish, GPLish, etc...
- [8:50] Zero Linden: The end() interator in STL seems to trip up a lot of programmers
- [8:51] Zero Linden: indeed, the first time I read the spec, I was a little lost....
- [8:51] Zero Linden: but, defining all operations on half open intervales: [ a, b )
- [8:51] Zero Linden: from a inclusive, up to but not including b
- [8:51] Zero Linden: has many advantages
- [8:52] Zero Linden: like: dividing a range [ a, c) at any arbitrary point b, is just the two ranges [ a, b ), [ b, c )
- [8:53] Zero Linden: and a test for contiguous ranges is just a comparison for equality of end points [ a, b ) directly preceeds [ c, d ) iff b == c
- [8:53] Zero Linden: but that means that operating on the whole list [ begin(), end() )
- [8:53] Zero Linden: implies that end() has to point beyond the list
- [8:54] Zero Linden: and that trips people up to have a valid iterator on a list that isn't dereferencable
- [8:54] Zero Linden: ah well
- [8:54] Zero Linden: (realizes that no one asked for that little essay..... but he enjoyed it!)
- [8:54] Tao Takashi: :-)
- [8:54] JayR Cela: :_)
- [8:54] Cadis Blackadder: :)
- [8:55] Zero Linden: [ :, - )
- [8:55] Zero Linden: wow - now I want to go off and design a language where : means begin() and - means end()....
- [8:56] Zero Linden: and you can write: sort: somearray on: [ :, - )
- [8:57] Tao Takashi: the new LSL is born! :)
- [8:57] Zero Linden: actually, the real sad moral of this story is that in most implemetnations of STL, dereferencing end() doesn't give an error (just junk data)
- [8:57] Tao Takashi: you can call this language smiley
- [8:57] Zero Linden: and if you compile/rum then in debugging mode, they are too slow to use in the sim
- [8:58] Cadis Blackadder: the operators in the language would make u giggle
- [8:58] Tao Takashi: btw, I think some OGP blog would be nice for announcements like new specs and such
- [8:59] Zero Linden: LL is in the middle of revamping it's whole blog strategy internally -- they are trying to make the communications more rational
- [8:59] Zero Linden: I'll ask about how they'd like to incorporate OGP communications in there.
- [9:00] Zero Linden: e-mail question sent
- [9:00] Zero Linden: (see - just in time!)
- [9:01] Damian Delacroix: politely raises a paw.
- [9:01] Tao Takashi: maybe it would also be nice to have not only LL news on it
- [9:01] Tao Takashi: new versions of components etc. might also be interesting
- [9:01] Tao Takashi: thanks for asking :)
- [9:02] Saijanai Kuhn: Tess was asking about next steps for OGP. I suggested that we impement the group IM with the OGP protocols to help establish community for the gridnauts and to get a leg up on what the issues are for getting a real group IM i OGP working
- [9:03] Saijanai Kuhn: implement the *current* group IM in the Agent Domain that is
- [9:04] Zero Linden: I have to admit - that I've been a bit out of the loop with them and am not sure what else is being considered
- [9:04] Zero Linden: I do know that right now Tess and Infinity and Leyla are all trying to get a revision of the docs together that
- [9:05] Saijanai Kuhn: well appearance was one suggestion, but thatgets into the inventory issue
- [9:05] Zero Linden: both reflect reality and what we already know is "wrong" there
- [9:05] Zero Linden: limited appearance gets into only a small part of the issues
- [9:05] Zero Linden: which is why it is tackleable
- [9:06] Saijanai Kuhn: its also dear to my heart cause I never got group IM working in my old python scripts. Would be nice to finally see it working (with documentation) in the pyogp
- [9:06] Tao Takashi: profile?
- [9:06] Zero Linden: (the agent domain would need to server ONLY the baked textures - and we could serve them directly to the other viewers just as we do on the main grid today)
- [9:07] Saijanai Kuhn: how do you mean directly to the ohter viewers? Other viwers hooked to the same AD?
- [9:07] Tao Takashi: doesn't this complicate communication channels?
- [9:07] Zero Linden: nope
- [9:07] Zero Linden: consider this rough flow
- [9:08] Tao Takashi: I guess you can see this as one part of the inventory
- [9:08] Zero Linden: your AD -> RD: "I'm wearing a baked texture available at URL x"
- [9:08] Tao Takashi: but it would mean that viewers also need to contact various ADs
- [9:08] Zero Linden: RD -> other avatars: "Joe Avatar is wearing a baked texture available at x"
- [9:08] Zero Linden: other viewers -> your RD: "gimme x"
- [9:09] Zero Linden: to the viewers is no more than a texture fetched from a URL
- [9:09] Tao Takashi: you mean "your AD" in the last sentence then?
- [9:10] Zero Linden: uhm, yes
- [9:10] Zero Linden: other viewers -> your *AD*: "gimme x"
- [9:10] Xugu Madison: [1]
- [9:10] Tao Takashi: and you have to get a cap first?
- [9:10] Tao Takashi: or is that the cap you get from the RD?
- [9:10] Zero Linden: well - the viewer can't tell the difference between a cap and a plain URL, no?
- [9:11] Zero Linden: so if the AD wants to give a cap to the RD that it will expire (or replace) is fine
- [9:11] Zero Linden: the RD can redistribute that cap (go caps!)
- [9:11] Zero Linden: if the AD wants to give out just a plain ole' URL, that's really the same
- [9:11] Tao Takashi: ok, ic
- [9:11] Zero Linden: On the other hand....
- [9:12] Zero Linden: Your AD could give a seed cap for other viewers to use to the RD:
- [9:12] Zero Linden: your AD -> RD: "Here's a seed cap for all sorts of info to give to other viewers: cap-x"
- [9:12] Tao Takashi: that would increase network traffic though
- [9:12] Zero Linden: RD-> other viewers: "Here's a seed cap for Joe Avatar: cap-x"
- [9:13] Zero Linden: other viewr -> your RD (via cap-x): "I'd like a cap to your baked texture, and to your profile, and to private-IM to you"
- [9:13] Saijanai Kuhn: This starts to feel more like the caps issues raised in that masters thesis
- [9:14] Saijanai Kuhn: [2]
- [9:14] Zero Linden: Yes, this later form is more like a more general purpose cap system
- [9:14] Gareth Ellison: wow, great timing
- [9:14] Gareth Ellison: i'm TPing out of here but my ghost will stay behind
- [9:14] Morgaine Dinova: I've been thinking recently that the concept of inventory as it currently stands isn't logical, and is too limiting anyway in allowing only one per av. Inventories should be simple container items, wear them, move them, give them to others etc.
- [9:15] Zero Linden: to tie those to together
- [9:16] Zero Linden: Chip Morningstar has talked about Habitat having such a more pure system: All information was only available via capabilities - including for example every object in the scene and those object's properties
- [9:16] Zero Linden: and that inventory was done with just objects in the system
- [9:16] Xugu Madison: I'd also love too have a "region" on my own computer, where I can keep inventory safely in wardrobes or shelves or other things... maybe... make sense?
- [9:16] Tao Takashi: I should think more about how this stuff (e.g. profile etc.) can be used in a web scenario as well as this is a common problem
- [9:16] Morgaine Dinova: Sounds interesting. Is there s link to his idea?
- [9:17] Zero Linden: Morgaine - his idea was a running system and product: Habitat
- [9:17] Zero Linden: [3]
- [9:17] Morgaine Dinova: Even better. :-) Habitat is too generic a name to Google I think, but I'll see
- [9:17] Saijanai Kuhn: a similar thing is descrbed in the pdf I just linked
- [9:17] Morgaine Dinova: Aha,. thanks
- [9:18] Zero Linden: this is an excellent paper he wrote about the experience: [4]
- [9:19] Tao Takashi: *bookmarked* thanks :)
- [9:20] Morgaine Dinova: Heh, it ran on the Commodore 64. And I bet that subjectively it ran faster than anything runs on Vista ;-)
- [9:20] Zero Linden: However, I don't think the Habitat team had to contend with the realities of a real 3D scene graph, with this much detail, on the kind of bandwidth that is commonly available
- [9:20] Tao Takashi: the demos i wrote for the c-64 definitely run faster than vista ;-)
- [9:21] Morgaine Dinova: Hehe Tao
- [9:21] Zero Linden: and as such, doing EVERYTHING, down to the last texture and menu item as a capability I don't think scales within today's resource constraints
- [9:21] Tao Takashi: I even recently found some in the depths of the internet.. it's very strange to suddenly see things popup you thought nobody has ever seen
- [9:22] Zero Linden: at present, at about 2k textures and a few k objects and 20 or so AVs....
- [9:23] Zero Linden: if it were all capabilities
- [9:23] Zero Linden: I'd estimate it at about 50k capaibilities
- [9:24] Zero Linden: and managing that is a little unrealistic by today's standards
- [9:26] Zero Linden: wow - this is our last office hours of Summer
- [9:26] Morgaine Dinova: Well you only get into scaling issues when inventory is monolithic, and when searches are full scans. But the low numbers we're talking about are noddy, not a research area. If we had billions of possessions, then yes, it would be research, but not with current numbers.
- [9:26] Zero Linden: next week will be just after Labor Day here in the US
- [9:26] Zero Linden: and I think back to work for most of Eurpoe
- [9:26] Zero Linden: Europe
- [9:27] Morgaine Dinova: Eurpoe sounds nice. Typos often produce interesting words :-)
- [9:27] Zero Linden: well, consider if each inventory item were a capability and needed to be fetched individually with an HTTP (or some other) transaction
- [9:28] Zero Linden: my 3k inventory (small by many AV standards) would require 3k fetches to get... if it were capability pure (since even the name of the item would be in the cap)
- [9:28] Saijanai Kuhn: that was the issue when someone wanted to batch delete inventory items
- [9:28] Zero Linden: (would be via the cap)
- [9:28] Morgaine Dinova: But "full fetch" is tantamount to "full scan" .... that's a bad design for starters :-)
- [9:28] Zero Linden: "Eurpoe" was one of the muses... right? :-)
- [9:28] Morgaine Dinova: Hehe
- [9:28] Zero Linden: "muse of scary stories"
- [9:29] Saijanai Kuhn: Let me consult my Sandman collection
- [9:30] Zero Linden: [5]
- [9:30] Zero Linden: I want a Wikipedia HUD in SL!
- [9:31] Cadis Blackadder: lol
- [9:31] Morgaine Dinova: Inventories (the metadata) should be cached locally anyway, so the only time a full list it brought in from the servers should be when a given level of the hierarchy is first opened.
- [9:32] Zero Linden: right - so one can being to make "compromises" in the pure capability model: Like having the container maintain the name of the contained items - so that you can fetch the contents of a folder and get a name->cap map, and display the contents without de-ref'ing all the caps
- [9:32] Zero Linden: s/being/begin/
- [9:34] Zero Linden: well all - enjoy the rest of your vacation
- [9:34] Tao Takashi: I want ubiquity for SL ;-)
- [9:35] Saijanai Kuhn: [6]
- [9:35] JayR Cela: yep / byee everyone / enjoy the weekend :_)
- [9:35] Zero Linden: In two weeks all the smoggy^h^h^h^h^h^h^hlofty LL strategy work will be done for me and I'll be back into guts of protocols
- [9:35] Rex Cronon: u too
- [9:35] Saijanai Kuhn: whee
- [9:35] Zero Linden: and I expect that we'll all be back into the bits again!
- [9:35] Tao Takashi: sounds like new TM policies ;-)
- [9:35] Cadis Blackadder: yey
- [9:35] Tao Takashi: bits sounds good though :)
- [9:35] Zero Linden: (loves that like kids in RL - when we stand up in SL, we stand on the couch!)
- [9:36] Cadis Blackadder: buhbye :) good weekend to u 2
- [9:36] Zero Linden: later
- [9:36] Tao Takashi: cya Zero