User:Which Linden/Office Hours/2009 June 25

From Second Life Wiki
< User:Which Linden/Office Hours
Revision as of 12:13, 25 June 2009 by Which Linden (talk | contribs) (Created page with '* [11:01] Latif Khalifa: hey which * [11:01] Twisted Laws: hello which * [11:01] Which Linden: Hi there! * [...')
(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.
  • [11:01] Latif Khalifa: hey which
  • [11:01] Twisted Laws: hello which
  • [11:01] Which Linden: Hi there!
  • [11:01] Morgaine Dinova: Hiya Which :-)
  • [11:01] Which Linden: I'm running Snowglobe, it's pretty sexy!
  • [11:02] Latif Khalifa: hehe
  • [11:02] Widget Whiteberry: surveys chairs, wonders which are filled
  • [11:02] Morgaine Dinova: I'm running Snowglove too, but wouldn't call it sexy, hehe. Just greater zoom in map, main visible change, no?
  • [11:02] Morgaine Dinova: globe*
  • [11:02] Widget Whiteberry: Which looks good as bamboo
  • [11:02] Which Linden: the minimap is better too
  • [11:02] Micah Giha: notes that Which moved a safe distance from the herbivore in the area
  • [11:03] Aimee Trescothick: heh, glad to hear you like the minimap :)
  • [11:03] Which Linden: Yeah, great work!
  • [11:03] Which Linden: Weirdly though zooming in all the way doesn't seem to center on my av
  • [11:04] Morgaine Dinova: Woohoo, Aimee's got her mad SpaceNav vehice again, I need another wild ride like yesterday :P
  • [11:04] Aimee Trescothick: ah yeah, that's cos it centres on the camera position
  • [11:04] Widget Whiteberry: focus has been flakey with this last release
  • [11:04] Aimee Trescothick: which is generally a bit behind your av
  • [11:04] Which Linden: Ohhhhh good call aimee
  • [11:04] Aimee Trescothick: wasn't noticeable when you couldn't zoom so close
  • [11:05] Which Linden: yeah... well that's pretty sweet
  • [11:05] Which Linden: Glad I know about that now
  • [11:05] Aimee Trescothick: hoping to make it an option what to focus on in the future
  • [11:05] Morgaine Dinova: Aimee: what's you next big thing for Snowglobe? Give us an exclusive hint ;-)))
  • [11:05] Morgaine Dinova: your*
  • [11:06] Mabalot Tatham: accepted your inventory offer.
  • [11:06] Mabalot Tatham: accepted your inventory offer.
  • [11:06] Aimee Trescothick: well, can't speak for anyone else, but planning to continue tweaking the minimap some more myself
  • [11:06] Morgaine Dinova: Apart from space invaders on minimap of course ....
  • [11:06] Which Linden: now that I know it zooms on camera position I can't imagine it any otherr way
  • [11:07] Aimee Trescothick: yeah, it's most noticeable if you use flycam, as it follows with your camera then
  • [11:07] Morgaine Dinova: I like it, although a tickbox to control the center would be good
  • [11:07] Aimee Trescothick: would be handy to have it stay on the avatar at times though, like working in a sandbox
  • [11:07] Saijanai Kuhn: Signpost just sent me a video of the map with calanderized events marked on it
  • [11:08] Mojito Sorbet: I prefer minimap always follows the flycam
  • [11:08] Saijanai Kuhn: [1]
  • [11:08] Morgaine Dinova: Aimee: just add a control to the minimap right-button menu?
  • [11:08] Which Linden: the minimap shouldn't gett too featureful, because then it's like why not use the regular map
  • [11:08] Aimee Trescothick: where your camera may be away from your avatar, but you'd like to still see who's around your body
  • [11:08] Mojito Sorbet: Regular map way too slow
  • [11:09] Aimee Trescothick: the main map is much fast in snowglobe
  • [11:09] Mojito Sorbet: Turn on "show land for sale" on main map and watch what happens to performance
  • [11:09] Which Linden: I guess that's not really what I meant. It seems a bit weird to have two such different maps.
  • [11:10] Aimee Trescothick: but I see the minimap as more of a real time plan view of what's happening, whereas the Main map is for long distance navigation and planning
  • [11:10] Saijanai Kuhn: Signpost's mod doesn't seem to slow down the main map that much though
  • [11:10] Mojito Sorbet: I view main map as long distance navigation. MiniMap is for near me.
  • [11:10] Morgaine Dinova: They really have different purposes. I guess they could be merged, but that might end up in compromising both.
  • [11:10] Mojito Sorbet: yes
  • [11:10] Which Linden: Yeah they have different purposes, as befits their very different UIs
  • [11:11] Mojito Sorbet: In fact I rarely use LMs for the usual places I go. Bring up map, type sim name, enter, then pick a spot to TP to
  • [11:11] Aimee Trescothick: I would like to be able to use the S3 map tiles on the minimap though
  • [11:11] Aimee Trescothick: that may well be something I look at soon
  • [11:11] Which Linden: I liked the proposal on sldev to make objects above/below the camera pos to fade out
  • [11:11] Latif Khalifa: Aimee, that would be agreat option
  • [11:11] Mojito Sorbet: Fade out, instead of the little arrows?
  • [11:11] Micah Giha: Until I tried to find these office hours sites, I have not been using the main map much at all
  • [11:12] Which Linden: Mojito: objects not avs
  • [11:12] Mojito Sorbet: Oh, ok
  • [11:12] Latif Khalifa: i'd just turn them off, the objects, just use the tile with terrain + objects from s3
  • [11:12] Saijanai Kuhn: for RPG and war games, a "fog of war" option might be good as well
  • [11:12] Aimee Trescothick: I'd like to have the S3 map and recently updated objects overlayed
  • [11:13] Mojito Sorbet: I think disabling map is a feature of RLV, isnt it?
  • [11:13] Aimee Trescothick: so you still see stuff that's moving or changing
  • [11:13] Which Linden: And I really love Merov's notion that we don't need a map at all; you should be able to just fly really high and see very far
  • [11:13] Morgaine Dinova: Aimee: all textures will come from S3 in due course, no? So eventually the sources of both will be unified.
  • [11:13] Latif Khalifa: lol
  • [11:13] Aimee Trescothick: well, the minimap at the moment isn't a ready rendered texture
  • [11:13] Mojito Sorbet: You know how long it takes to get up that hih? lol
  • [11:14] Aimee Trescothick: it's created within the viewer
  • [11:14] Which Linden: Mojito: yeah, but maybe there'd be some supa-fast flying option
  • [11:14] Morgaine Dinova: Which: is Merov's middle name "EagleEye"? ;-)
  • [11:14] Mojito Sorbet: ctrl-M is still faster
  • [11:14] Aimee Trescothick: terraform land and you can watch it changing on the minimap etc.
  • [11:15] Morgaine Dinova: Could be just an overlay on the last recorded terrain texture though
  • [11:15] Mojito Sorbet: Yes, I found that useful when I was building my new island
  • [11:15] Which Linden: I guess it's not really a practical thing but I could imagine that when you hit ctrl-m, instead of bringign up the floater it'd just aot-boost your camera to 1,000,000 m
  • [11:15] Latif Khalifa: i wish merov would fix bugs that make flying high very unpleasant experiene now, with half of horizon flickering in and out of existance
  • [11:15] Morgaine Dinova: Yeah, I hate that flickering in Snowglobe
  • [11:15] Which Linden: *auto-boost (to pick one especially egregious misspelling to correct)
  • [11:15] Which Linden: orly, the water flickers?
  • [11:16] Morgaine Dinova: The horizon flickers when you're up high
  • [11:16] Latif Khalifa: yeah, but when you're high enough most of what you see is water
  • [11:16] Mojito Sorbet: Sometimes goes grey, sometimes blue
  • [11:16] Which Linden: hm maybe it's a shader thing
  • [11:16] Morgaine Dinova: And has a white patch in front. So basically it's got some bug
  • [11:17] Latif Khalifa: well its freshly broken in 1.23
  • [11:17] Which Linden: broken in both 1.23 and snowglobe?
  • [11:17] Latif Khalifa: yup
  • [11:17] Morgaine Dinova: We could add "Rise to 1000m and check horizon" to the Smoke Tests on the wiki
  • [11:18] Latif Khalifa: the purpose of tests is to find bugs, RC process found plenty, the problem is they are not fixed
  • [11:18] Which Linden: could do that
  • [11:18] Latif Khalifa: its not that we don't find bugs
  • [11:18] Which Linden: well there's always a tradeoff when doing a release
  • [11:18] Latif Khalifa: yeah yeah
  • [11:19] Which Linden: because fixing bugs also adds risk of introducing a new bug
  • [11:19] Latif Khalifa: "best viewer ever"... heard that song from Q already lol
  • [11:19] Morgaine Dinova: Snowglobe is very early in its life, should be no surprise. But 1.23.4 has same bugs, and that's quite wrong.
  • [11:19] Which Linden: it took me a long time to understand that sort of release process
  • [11:19] Latif Khalifa: its just that Lindens and residents have a very different view of what is important to fix for a release
  • [11:20] Which Linden: what would be ideal is if we could get away from the "shove bugs around" sort of model
  • [11:20] Morgaine Dinova: 1.23.4 is full of bugs that 1.22.11 doesn't have, so clearly something went very wrong in the SVN update
  • [11:20] Latif Khalifa: what Q classifies as "nice to have" feature, people actually using SL think of as critical bugs
  • [11:20] Which Linden: Yeah, that's awkward
  • [11:20] Mojito Sorbet: And errors in design are just as much a "qaulity" issue as are crashes.
  • [11:21] Which Linden: Hah true but wayyy harder to fix
  • [11:21] Morgaine Dinova: Mojito: very true. Unfortunately, QA isn't tasked to identify design faults
  • [11:21] Latif Khalifa: its the stubborness to admit that there is a problem that is most difficult to understand
  • [11:21] Mojito Sorbet: At some point, you really do need to look at doing the viewer over form scratch, rather than cobbling more stuff onto old code.
  • [11:21] Which Linden: I generally agree that fixing crashes should take priority -- that's basically the worst sort of bug
  • [11:21] Mojito Sorbet: Proper threading, proper modulaization
  • [11:22] Which Linden: Not that I agree with everything Joel Spolsky says, but: [2]
  • [11:22] Which Linden: (that's right, article 69, baby)
  • [11:23] Morgaine Dinova: Nooooooo .... we much continue to grow the monolith!!! Plenty of room for a 100 gig executable on modern drives! ;-)
  • [11:23] Mojito Sorbet: I dont mean throw out every line of course.
  • [11:23] Latif Khalifa: Which, but if a bug makes parts of the scene disappear, that build becomes useless. And 1.23 makes stuff disappear in builds with a lot of sculpties
  • [11:23] Mojito Sorbet: But rendering should be completely divorce from network and user I/O for example
  • [11:23] Morgaine Dinova: Yup
  • [11:23] Which Linden: Our executable is really small relative to pretty much anything out there, Morgaine
  • [11:24] Morgaine Dinova: Which: so there's tons of bad stuff out there ... dooesn't mean it's right :-)
  • [11:24] Which Linden: Latif: ah you're talking about the rendering limits, yes, that's semi-intentional so it's trickier to deal with
  • [11:25] Mojito Sorbet: Is that different from the Mesg Detail slider?
  • [11:25] Which Linden: Morgaine: yeah, but, I'm just sayin' -- hardly our biggest problem
  • [11:25] Mojito Sorbet: *mesh
  • [11:25] Latif Khalifa: Which, there was a problem. The "fix" is a bad one. I wish LL would admit, the fix is bad and look for a correct one
  • [11:25] Latif Khalifa: saying its intentional, its supposed to be like that is not very helpful
  • [11:25] Which Lindenri: Ashton: helllo everyone!
  • [11:26] Morgaine Dinova: 90% of the viewer should be written in a safe scripting language, and only the hot spots done in C++. You'd have far fewer bugs, far greater readability, and roughly the same performance.
  • [11:26] Which Linden: Latif: Agreed. There's risk from fixing bugs, in that you could do so badly! But we'll eventually get it right
  • [11:26] Which Lindenri: Ashton: sorry, had to visit Jack Linden first because of the asset server issues
  • [11:26] Morgaine Dinova: lol
  • [11:26] Mojito Sorbet: Bah
  • [11:26] Aimee Trescothick: but but but C++ is what keeps programmers in work Morgaine
  • [11:27] Aimee Trescothick: we'd have nothing to do but add features without the bugs
  • [11:27] Latif Khalifa: Which, this was discovered early on in RC process... its the subborness to admit that there is a problem on LL part that is... interesting
  • [11:27] Mojito Sorbet: Youy could give them keypunches - that would make them even less productive.
  • [11:27] Morgaine Dinova: There would still be plenty C++ to do --- the bulk of the graphics for starters.
  • [11:27] Morgaine Dinova: 3D I mean.
  • [11:27] Mojito Sorbet: Oh sure
  • [11:27] Morgaine Dinova: No C~~ in the 2D at all.
  • [11:27] Mojito Sorbet: Compute intensive stuff.
  • [11:27] Aimee Trescothick: graphics programming should be done in Logo
  • [11:27] Morgaine Dinova: Hehe
  • [11:28] Mojito Sorbet: Netwrok and user interactions, could all be in C# or something
  • [11:28] Mojito Sorbet: Something with built in support for thrading and locks
  • [11:28] Which Linden: Yeah, or we could do it like EVE does, where everything is in PYthon
  • [11:29] Saijanai Kuhn: obj C with array synntax
  • [11:29] Micah Giha: How many users with low-end hardware would be forced out of SL with less efficient (scripted) code?
  • [11:29] Mojito Sorbet: Especially if client-side scripting hapens, you would have mono in there anyway
  • [11:29] Mojito Sorbet: Mono has a JIT compiler to real machine code
  • [11:29] Morgaine Dinova: Well with Python's greenlets and eventlets, you'd be able to hardness multicores quite nicely without horrendous explicit parallel threading code.
  • [11:30] Saijanai Kuhn: actually, smalltalk with array syntax *trying to learn F-script still)
  • [11:30] Mojito Sorbet: Is that F++? Mono does that too.
  • [11:30] Saijanai Kuhn: Not sure what that is
  • [11:30] Latif Khalifa: hey Dahlia :)
  • [11:30] Which Linden: Micah: the general idea is that in an application, the bulk of the code is spent doing things that aren't time-critical, but there's a small set of code that needs to be really fast. Using this model you could have 80% of the code be in a "slow" language but still be about 95% as performant
  • [11:30] Morgaine Dinova: No, that's F# Mojito
  • [11:31] Mojito Sorbet: ok
  • [11:31] Mojito Sorbet: I bet it would be more than 95%
  • [11:31] Which Linden: On the other hand, since we have our own UI library, pretty much all that library code would want to be in C++
  • [11:31] Mojito Sorbet: Especially if thie higher-level language made parallelism easier
  • [11:32] Micah Giha: ummm... my point... there are still single core machines running SL clients out there
  • [11:32] Which Linden: Mojito: possibly, but then again even now not everyone has a dual-proc machine
  • [11:32] Saijanai Kuhn: P is an array of stuff like [(name, address, phone)...
  • [11:32] Mojito Sorbet: Just because many peple do not have a dual core machine does not mean you should not take advantage of it if it is there!
  • [11:32] Which Linden: The UI library does some pretty heavy data processing (e.g. iterating over a list of 1000 items in response to a mouse click)
  • [11:32] Mojito Sorbet: Design the code to scale with te hardware.
  • [11:32] Latif Khalifa: well SL isn't really cpu bound on a moderm processors anyway
  • [11:33] Mojito Sorbet: Oh? Try shadows
  • [11:33] Morgaine Dinova: Pity F# is MS-only, or I'd have a play. I wonder if anyone's doing an open source one, maybe translated to C# for Mono.
  • [11:33] Saijanai Kuhn: Which, actually the array thingie in F-script optomizes for that
  • [11:33] Which Linden: Mojito: of course but this is Micah's fear: that someone with an entry-level system will be hurt in the quest to boost performance on high-end maachines
  • [11:33] Mojito Sorbet: Sure, so you do not use time-slice scheduling in compute intensive tasks for instance.
  • [11:33] Mojito Sorbet: This is not rocket science.
  • [11:34] Which Linden: Sai: ?
  • [11:34] Morgaine Dinova: That's where green threads would score, they don't need multiple cores to provide great concurrency in a single process.
  • [11:34] Micah Giha: If too much is added client side based on the assumption of computing power of multi core...
  • [11:34] Dahlia Trimble: green threads?
  • [11:35] Saijanai Kuhn: which?
  • [11:35] Morgaine Dinova: Dahlia: like Python's greenlets and eventlets, and Erlang's processes. You can spawn millions, no problem
  • [11:35] Mojito Sorbet: Says on wiki that studies show "Linux native threads have much better performance on I/O and context switching operations."
  • [11:35] Which Linden: Sai: was just curious what you meant about "the array thingie"
  • [11:36] Saijanai Kuhn: ah, F-script extends smalltalk syntax to handle optimized operations on arrays
  • [11:36] Mojito Sorbet: Like APL?
  • [11:36] Saijanai Kuhn: yeah. Explicitly like APL
  • [11:37] Mojito Sorbet: with more readable syntax I hope
  • [11:37] Morgaine Dinova: I'd love to know why LL never put arrays into LSL. It's one of the noddiest things to do (and was 90% done by Don 3 years ago) ... so why didn't it happen?
  • [11:37] Micah Giha: What's not readable about APL?
  • [11:37] Saijanai Kuhn: maybe. Definitely more compact than smalltalk or python
  • [11:37] Mojito Sorbet: Well, it does beat TECO for readability
  • [11:37] Which Linden: Even if you have fast iteration, calling <scripted function foo> on each item in an array is going to be slow
  • [11:38] Morgaine Dinova: APL does improve on line noise for readability ;-)
  • [11:38] Which Linden: Morgaine: risk of introducing new bugs?
  • [11:38] Which Linden: I don't actually know why LSL has no arrays (or more complex data structures)
  • [11:39] Mojito Sorbet: I am ok with the approach of not tacking more stuff onto LSL, and just go to C#
  • [11:39] Which Linden: Yeah probably everyone sees that as more of a win
  • [11:39] Mojito Sorbet: Any new library functions of course would benefit both
  • [11:40] Morgaine Dinova: Which: arrays are so simple though, both in semantics and in implementation, that bugs really don't get much opportunity ... and anyway, all development introduces bugs, yet dev hasn't stopped.
  • [11:40] Mojito Sorbet: And there are engineering process techniques to reduce chances of bugs getting in.
  • [11:41] Which Linden: Morgaine: yeah, I don't claim to defend every prioritization decision we make
  • [11:41] Saijanai Kuhn: E.G.///
  • [11:41] Saijanai Kuhn: > {'oliver', 'henry', 'bertram'} uppercaseString
  • {'OLIVER', 'HENRY', 'BERTRAM'}
  • [11:41] Mojito Sorbet: Just overload the operators.
  • [11:41] Saijanai Kuhn: yar, but this is optimized outside the interpreter
  • [11:42] Micah Giha: Seeing so much lag on sims with too many scripts running now.. adding complexity would only aggravate that situation
  • [11:42] Which Linden: Wow just looked at the APL page. I kinda like the idea of using extended characters as operators. Very concise
  • [11:42] Morgaine Dinova: The amount of sim CPU currently being lost to running up and down lists must be huge, since all none-trivial programs do it. The savings from very simple O(1) arrays would have been vast.
  • [11:42] Mojito Sorbet: Argh, lists. Me hates lists
  • [11:42] Which Linden: Micah: the flip side of that argument is that adding more powerful primitives would allow people to write the same behaviors with more efficient code
  • [11:43] Dahlia Trimble: lists suX0r :(
  • [11:43] Which Linden: Yeah I also hate them
  • [11:43] Which Linden: Struggled with them for that dang escrow demo
  • [11:43] Morgaine Dinova: Lists are great in list-oriented languages. But LSL isn't one.
  • [11:43] Which Linden: Oh that's a good idea --- let's convert LSL to Erlang!
  • [11:44] Morgaine Dinova: That would work :-)
  • [11:44] Mojito Sorbet: LISP!
  • [11:44] Latif Khalifa: lol
  • [11:44] Which Linden: Ha ha, that's actually a hilarious idea, because LSL scripts are quite a bit like Erlang processes
  • [11:44] Morgaine Dinova: At least I'd be earning money from scripting .... nobody else would though :P
  • [11:44] Latif Khalifa: for about two days
  • [11:44] Mojito Sorbet: Keeping track of CAR and CDR diagrams gives me a headache
  • [11:45] Dahlia Trimble: EWWWWW lisp :(
  • [11:45] Morgaine Dinova: Which: they are indeed quite similar to Erlang processes, in their shared-nothing paradigm. In almost nothing else though
  • [11:45] Saijanai Kuhn: Suppose you have an array, KEY, and another array called LOCK. Both arrays are the same
  • length. You need to know whether every element of KEY is equal to the corresponding LOCK
  • element:
  • > KEY := {1.01, 1.763, 1.808, 1.2346, 1.2272, 1.8095, 1.1}
  • > LOCK := {1.01, 1.763, 1.898, 1.2346, 1.2272, 1.8095, 1.1}
  • > KEY = LOCK
  • [11:45] Saijanai Kuhn: FALSE
  • [11:46] Which Linden: Morgaine: and they can receive events, just like Erlang processes
  • [11:46] Morgaine Dinova: LISP rocks! /me ducks at the rock from Dahlia
  • [11:46] Which Linden: And they have a ridiculously limiting set of primitive data structures, just like Erlang
  • [11:46] Saijanai Kuhn: oops Key = Lock \ \ #&
  • [11:46] Which Linden: kids because he loves
  • [11:46] Mojito Sorbet: [3]
  • [11:47] Morgaine Dinova: Hahaha, yeah, that's a classic xkcd
  • [11:47] Dahlia Trimble: hurls a giant rock at Morgaine's caddaddar
  • [11:47] Morgaine Dinova: Well I do both LISP and Perl, so I'm covered :-)
  • [11:47] Morgaine Dinova: catches the rock in a cons.
  • [11:48] Mojito Sorbet: Hey, C# has Lambda functions too
  • [11:48] Which Linden: Guido hates lambdas
  • [11:48] Latif Khalifa: i really really like c#
  • [11:48] Morgaine Dinova: Yeah, and they're quite nice in C#3
  • [11:48] Latif Khalifa: the ease of event handling, threading, built in collections, its just sweet
  • [11:48] Latif Khalifa: too bad Mono is such a POS
  • [11:49] Mojito Sorbet: how so?
  • [11:49] Dahlia Trimble: <3 mono
  • [11:49] Latif Khalifa: well i wrote this app that would end up using 700MB ram in mono
  • [11:49] Morgaine Dinova: The only major thing missing from C# (or actually, from CLR) is green threads. It really needs those, in these days of parallel programming.
  • [11:50] Latif Khalifa: so it was more efficient for me to install Sun Virtual Box... with Windows XP running in VM
  • [11:50] Mojito Sorbet: So it is mono's implemntation of CLR that is the problem, not the concept?
  • [11:50] Latif Khalifa: imagine that, XP virtual machine running on 512mb ram would outperform mono on the same box, using less cpu and ram
  • [11:50] Dahlia Trimble: it probably was a difference in the method you used to measure memory
  • [11:51] Latif Khalifa: Dahlia, my system load went from 10-20 to .3
  • [11:51] Mojito Sorbet: Windows and Linux have opposite policies in how memory and paging get allocated
  • [11:52] Morgaine Dinova: Latif: according to Miguel de Icaza, that used to be the case, but from Mono 2.4 onwards there's proper memory handling. -- [4]
  • [11:52] Which Linden: Hm but mono is much more embeddable, as we observe on a daily basis
  • [11:52] Latif Khalifa: lies, danmed lies and statistics
  • [11:53] Latif Khalifa: mono 2.4 garbage collector and memory usage is still faaaar from optimized
  • [11:53] Mojito Sorbet: All I can get is mono 2.0.1
  • [11:53] Morgaine Dinova: Well he ran the things that the guys on #opensim told him to run.
  • [11:54] Dahlia Trimble: Miguel finally did get back to me and he gave me a few links discussing measuring memory consumption in linux, I'll try to dig them up and share them
  • [11:54] Which Linden: Did he not compare it to the .NET vm? Seems an oversight
  • [11:54] Morgaine Dinova: Cool Dahlia!
  • [11:54] Latif Khalifa: oh well sounds like Linden Lab when they refuse to acknowlege there is a problem
  • [11:54] Morgaine Dinova: By the way, read the Discussion page off that Mono page
  • [11:55] Morgaine Dinova: It says what he measured
  • [11:55] Latif Khalifa: i know on my linux box VM running XP outperforms Mono by factor of 10
  • [11:55] Morgaine Dinova: Speed?
  • [11:56] Mojito Sorbet: Probably different JIT compilers.
  • [11:56] Latif Khalifa: system load and memory usage
  • [11:56] Morgaine Dinova: (Haven't seen Miguel comment on speed, only on memory)
  • [11:56] Latif Khalifa: he can claim there is no problem all he wants :)
  • [11:57] Morgaine Dinova: He didn't claim, he ran what the #opensim guys told him to run. You give him your test scenario, and your results, and it would be interesting to see his tests.
  • [11:57] Which Linden: Doesn't seem to me that he is claiming there is no problem, only that things are improving
  • [11:57] Mojito Sorbet: I think it may depend on exactly what you are doing.
  • [11:57] Dahlia Trimble: dunno... I have code that works better on mono than .net and vice versa
  • [11:58] Morgaine Dinova: Yep
  • [11:58] Mojito Sorbet: People do not worry abouit locality-of-reference any more
  • [11:58] Which Linden: Yeah, Latif: I'd encourage you to show someone on the Mono team your 10x-difference program, maybe they'll be able to fix your issue
  • [11:58] Morgaine Dinova: Which is why one needs to establish the exact test scenario, and then compare .NET and Mono. Can't measure handwaving.
  • [11:58] Latif Khalifa: guess where on the grapsh i switched from running Mono to XP-in-a-VM
  • [11:58] Mojito Sorbet: I want to know how to get 2.4 on my Linux boxes without breaking anyting
  • [11:58] Latif Khalifa: [5]
  • [11:59] Mojito Sorbet: Both ubuntu and gentoo are still at 2.0.4
  • [11:59] Latif Khalifa: memory usage graph shows the same
  • [11:59] Which Linden: It kinda bugs me when someone finds a problem, talks about it, but never tries to fix it
  • [11:59] Which Linden: Or even tell the people responsible for fixing it that there's a bug
  • [11:59] Morgaine Dinova: Gentoo was at 2.0.1 for Mono stable, but it had 2.4 in test branch
  • [12:00] Morgaine Dinova: And it works
  • [12:00] Latif Khalifa: Which, Mono folks were told gazillion times about memory consumption prolbems, but as you have seen in that link they just claim there is no problem
  • [12:00] Mojito Sorbet: Oh, just a masking then? I will check that
  • [12:00] Morgaine Dinova: Mojito: yeah, ~x86
  • [12:01] Latif Khalifa: and I cannot fix evrey problem I find lol
  • [12:01] Which Linden: Latif: I didn't see that in the link
  • [12:02] Which Linden: Just because someone says "In this specific scenario there is no problem" doesn't mean they are also saying "in all situations there is no problem"
  • [12:03] Which Linden: It just seems to me that you have something incredibly useful: a repro. Withholding that repro from the developer seems irresponsible.
  • [12:03] Morgaine Dinova: There might well be a problem, but as things stand, you *claim* there's a problem, while Miguel runs the #opensim test and *measures* that there isn't one. Measurement wins.
  • [12:03] Morgaine Dinova: So provide a suitable test.
  • [12:03] Latif Khalifa: sure
  • [12:03] Which Linden: Whoa open mike
  • [12:04] Which Linden: Cyber
  • [12:04] Cyber Enyo: HI
  • [12:04] Aimee Trescothick: sounds like he's commiting a murder
  • [12:04] Morgaine Dinova: It would be to everyone's benefit to be able to show Miguel that it's crap. Just need an effective test for him to replicate.
  • [12:04] Which Linden: Can you turn off your mike?
  • [12:04] Saijanai Kuhn: plugs ears
  • [12:04] Cyber Enyo: SORRY
  • [12:04] Which Linden: Thx
  • [12:04] Mojito Sorbet: sits in the silence of having voice disabled.
  • [12:05] Aimee Trescothick:  :)
  • [12:05] Which Linden: Ha ha ha good thing this is a text-only OH
  • [12:05] Cyber Enyo:  ?
  • [12:05] Micah Giha: places a tick mark in the column for 'good to not be on voice'
  • [12:05] Latif Khalifa: 1. use libomv, login 10 clients, observe memory usage mono vs. .net
  • [12:05] Dahlia Trimble: at times it's a benefit that sound doesnt work for SL™ on my linux box ;)
  • [12:05] Mojito Sorbet: That too
  • [12:05] Saijanai Kuhn: we chat in text only so we can make a record of hte office hour coversation, Cyber
  • [12:05] Latif Khalifa: and mono folks are aware of it, and were told on numerous occasions
  • [12:06] Saijanai Kuhn: and he gone.
  • [12:06] Micah Giha: He was noewbie... only 3 days old
  • [12:07] Micah Giha: ^newbie ... darn hooves.. so hard to type with them
  • [12:07] Dahlia Trimble: couldnt have been worse than the stuff I did as a 3 day old n00b ;)
  • [12:07] Morgaine Dinova: Maybe #opensim bods would script up a hypergridded setup for Miguel, and when he runs it, they all HG in to him so he can measure.
  • [12:07] Dahlia Trimble: lots of OpenSim users use mono
  • [12:07] Dahlia Trimble: quite a few
  • [12:08] Dahlia Trimble: including me
  • [12:08] Latif Khalifa: btw. anyone looking for a good VM software
  • [12:08] Latif Khalifa: Sun VirtualBox
  • [12:08] Latif Khalifa: highly recommend it
  • [12:09] Which Linden: I've hard good things about that, I'll have to try it
  • [12:09] Which Linden: *heard
  • [12:09] Dahlia Trimble: me too, <3 it
  • [12:09] Latif Khalifa: and its free
  • [12:09] Which Linden: Ha ha, until Oracle makes it for-pay
  • [12:09] Dahlia Trimble: it's Sun
  • [12:09] Latif Khalifa: haha
  • [12:09] Which Linden: I wonder what they're going to do with their N virtualization options
  • [12:09] Which Linden: Oracle bought Sun
  • [12:09] Latif Khalifa: no idea what's gonna happen with mysql either
  • [12:09] Dahlia Trimble: oh didnt know that
  • [12:10] Latif Khalifa: i wish ibm bought them, somehow i dislike oracle more than ibm
  • [12:10] Which Linden: My guess would be that one of the forks (Maria or Drizzle) captures mindshare
  • [12:10] Which Linden: It's cause Larray Ellison is uglier than Sam Palmisano
  • [12:10] Which Linden:  :-P
  • [12:10] Dahlia Trimble: lmao
  • [12:10] Latif Khalifa: well oracle doesn't do any open source for starters
  • [12:10] Latif Khalifa: lol Which
  • [12:10] Which Linden: Actually they do Latif
  • [12:10] Which Linden: Not much though
  • [12:11] Which Linden: So in general that is correct
  • [12:11] Latif Khalifa: oh well I liked PostgreSQL more anyway :P
  • [12:11] Latif Khalifa: and its BSD ;)
  • [12:12] Which Linden: OK well I've overstayed my hour a bit, I should get on with the day
  • [12:12] Which Linden: Thank you so much for such an interesting discussion
  • [12:12] Latif Khalifa: good luck with that Which
  • [12:12] Latif Khalifa:  :)
  • [12:12] Morgaine Dinova: Odd meeting, general tech chat :-)
  • [12:12] Latif Khalifa: take care
  • [12:12] Which Linden: See you soon!
  • Which Linden