Web   ·   Wiki   ·   Activities   ·   Blog   ·   Lists   ·   Chat   ·   Meeting   ·   Bugs   ·   Git   ·   Translate   ·   Archive   ·   People   ·   Donate

#sugar-meeting meeting, 2009-06-09 10:02:00

Minutes | Index | Today     Channels | Search | Join

All times shown according to UTC.

Time Nick Message
10:03 homunq quick roll call...
10:03 bemasc is here
10:03 homunq silbe bemasc lucian iwikiwi subzero BryanWB
10:03 iwikiwi yep here
10:03 lucian homunq: here
10:03 silbe is listening
10:03 BryanWB homunq: here
10:03 homunq walterbender aa
10:04 aa hola!
10:05 homunq OK, who wants to run the meeting today?
10:05 subzero homunq: here
10:05 mchua is here
10:05 iwikiwi aa: should we start?
10:06 homunq #topic iwikiwi aa karma
10:06 aa s/karma/print/
10:06 iwikiwi: go ahead :)
10:07 iwikiwi okay, we had been having lots of problems with the moodle side, but thankfully martin had took his time, and we have a clear procedure to follow
10:07 on the coding side, a coupla bug fixes to print, and making xml-rpc work
10:07 blogged it
10:08 http://materializingsweetness.wordpress.com/
10:08 I estimate in 2 and half weeks moodle side should be done with good
10:09 walterbender is here
10:09 lucian walterbender: hey
10:09 aa I think it might take a bit longer
10:09 but that's ok
10:10 iwikiwi hmm, i guess it could, we need to be in touch with martin regularly
10:10 homunq my only recommendation is not to put all the bells and whistles on moodle until you've finished the rest
10:11 I mean, "finished" the first draft.
10:11 get something working then move on for now.
10:11 that will also let you block on Martin less - of course, always make time to talk to him when he has time.
10:12 aa homunq: that was my recommendation as well
10:12 iwikiwi aa, homunq, so for now I get back to print?
10:12 aa but we also want to deliver something that works asap
10:12 homunq iwikiwi: get the basics done in moodle
10:13 aa and take some feedback
10:13 homunq but try to draw the line.
10:13 iwikiwi oh okay, i already have a inferior model up
10:13 that is its under assignments as opposed to directly under modules
10:14 so its one more layer down the heirarchy
10:14 I could always go back to it as fallback
10:14 s/a/an
10:14 homunq ok.
10:15 aa writing a module is not that harder than writing an assignment, but thats ok if you already have it
10:15 homunq It sounds as if you're doing OK, keep it up
10:15 #topic lucian walterbender
10:15 lucian homunq: i've been playing around with Browse and how I could integrate SSBs into sugar
10:16 i have rather primitive 'create ssb' in Browse right now
10:16 SSBs are just copies of Browse with an extra toolbar
10:16 but i have run into various trouble with trying to replicate an activity
10:16 walterbender (Rainbow)
10:17 lucian walterbender: yes, mostly that
10:17 tomeu lucian: we can spend some time later in #sugar if you want
10:17 lucian after a chat with bemasc, i've decided to allow the SSBs to create copies of themselves
10:17 basically like curry-ing functions
10:17 so users can make an SSB and customise it with sugar-specific functionality
10:18 and after customisation, they can create an .xo with all the customisation bundled in
10:18 so they can put it on activities.sugarlabs.org or share it
10:18 i've blogged about this a bit, i'll blog some more
10:18 homunq how do they customize it? Develop???
10:18 link
10:18 lucian honeyweb.wordpress.com
10:19 homunq: no, they use the extra toolbar in SSBs
10:19 they never get to see python code
10:19 they can (or will be able to) add userscripts (greasemonkey), userstyles (custom css) and bookmarklets
10:19 there will be 'add script' buttons and such
10:19 homunq OK
10:20 bemasc (javascripts, css, and javascript, respectively)
10:20 lucian yes
10:20 it will for example allow a user to make a GMail activity without seeing python code or fumbling about with creating .xo files
10:20 and functionally, the activity will rival or surpass the existing GMail activity let's say
10:21 so for all this to happen, there are still some pieces that need to be solved
10:21 greasemonkey, javascript-dbus and global bookmarks (as opposed to the current session bookmarks)
10:22 javascript-dbus and global bookmarks i've looked into, it shouldn't be too hard
10:22 bemasc lucian: global bookmarks?
10:22 oh, you mean stored-in-an-SSB bookmarks.
10:22 lucian bemasc: not really, for Browse
10:22 bemasc: it's on the Browse roadmap for 0.86
10:22 bemasc disapproves.
10:23 lucian bemasc: i mean bookmarks not stored in the Journal session, but in the activity settings/preferences
10:23 homunq i have to agree with bemasc, though not authoritatively.
10:23 bemasc Yeah.  I think that's a bad idea, but it's not my decision.
10:23 Anyway, don't let me derail the meeting train.
10:23 lucian http://erikos.sweettimez.de/?p=678
10:24 bemasc: alright
10:24 for my project, i can just store bookmarks in the SSB itself
10:24 but i thought that since the functionality is due in 0.86, it would better to not reinvent it
10:25 walterbender lucian: I remain concerned about how a SSB will store other data in the Journal for access by non-browse activities...
10:25 homunq don't depend on future work by anyone.
10:25 walterbender lucian: a good set of examples would be helpful
10:25 lucian walterbender: like files you mean?
10:26 walterbender files or any data format... JSON
10:26 lucian walterbender: that's rather easy, SSBs will simply offer download links
10:26 walterbender lucian: a working example would be appreciated
10:26 bemasc lucian: and sometimes those "downloads" will actually be entirely on the local machine?
10:26 lucian walterbender: sure, i'll work on that
10:26 bemasc: yep
10:26 bemasc cool
10:26 homunq OK, sorry, we have to move on.
10:27 lucian bemasc: it's how google docs offline works for example
10:27 homunq #topic homunq silbe versions
10:27 link and report?
10:27 silbe #link http://wiki.sugarlabs.org/go/V[…]atastore/Progress
10:28 homunq I saw your benchmarks.
10:28 silbe i've been working on a benchmark to help choose a VCS backend
10:28 homunq Do you have conclusions yet?
10:29 silbe there has been further progress (the second run finished just a few minutes ago), still need some memory usage estimation
10:29 there are a few loosers but no clear winner
10:29 it's all about compromises
10:29 bemasc silbe: are you sure you want to use an existing VCS?
10:29 homunq what are your considerations besides speed? Do you care about python-ness?
10:29 silbe darcs is the fastest but also takes a huge amount of disk space for example
10:30 homunq sorry, I meant besides the benchmarkable...
10:30 silbe bemasc: not sure yet, but might be the fastest way to go forward
10:30 tomeu silbe: I was thinking we would start for something so simple that a vcs would be really overkill
10:30 bemasc silbe: I find this approach very surprising, mostly because these VCSs typically have their differential compression algorithms hardwired.
10:30 homunq I would ++ at least considering existing options. If you can have a format with existing tools outside sugar, that is worth a fair amount.
10:30 silbe a python module would be great, but for the first prototype invoking CLI tools would suffice
10:31 bemasc And they use line-based diff, which is useless for us.
10:31 homunq bemasc, that is a very good point.
10:31 (not useless, but it covers small fraction of our use cases)
10:32 tomeu what I thought we would be doing inside the scope of this gsoc is "just" tracking the place of an entry inside its version tree and, perhaps, doing delta compression
10:32 silbe actually the current benchmark results point toward inventing our own stuff, but i'd prefer encapsulating that in a module that can be easily replaced
10:32 homunq absolutely.
10:32 bemasc silbe: I'm all for a nice hard abstraction barrier, if you can make one. That would be really cool.
10:33 homunq Inventing own overall (package-ish) format != inventing own diff algorithm and format.
10:33 lucian about python-ness, mercurial can be used as a module afaik
10:33 silbe homunq: exactly
10:34 tomeu lucian: we have a prototype that did exactly that
10:34 lucian tomeu: oh
10:34 tomeu but had to implement its own scheme for delta compression anyway
10:34 because hg didn't had that
10:35 homunq OK. silbe, we have to continue this a bit after in #sugar.
10:36 silbe To summarize: I think doing our own backend will be best long term, but using an existing VCS in the meantime will allow me to come up with working code faster
10:36 abstraction => can switch backends later
10:36 homunq (side-note: I'm running the meeting because I don't know if others can #topic....)
10:37 silbe: OK. In that case, go with whatever is easiest, drop the benchmarks now.
10:37 They've told you what you need to know - homegrown is the long-term answer.
10:37 silbe ok
10:38 homunq #topic subzero BryanWB
10:38 subzero I'm still working with can & can't activity joining versions :S
10:39 homunq link?
10:39 subzero I have done the animated version and the everything else version, so now I'm having troubles joining both versions into one
10:39 about the problems: http://karmaproject.wordpress.[…]hat-little-beast/
10:39 the demo (need ff 3.5beta4+): http://karma.sugarlabs.org/can_cant/
10:40 the difficult part is the animation, I need to run js in parallel
10:40 Bryan and kbrosnan (#firefox) has directed me to web workers
10:41 BryanWB_ sorry, the Internet just died for last 10 mins in my apt
10:41 subzero BryanWB_: np
10:41 homunq BryanWB_: you just started. I'll pm you the log of this topic
10:41 subzero BryanWB_: I'm talkin homunq about the current problems with animation stuff
10:41 BryanWB_ tks
10:42 homunq web workers?
10:42 BryanWB_ homunq: threads in js ;)
10:43 subzero yep, for firefox
10:43 lucian talked me about WorkerPool (for chrome)
10:43 homunq so, this issue is inevitably browser-dependent? too bad.
10:44 BryanWB_ subzero: lucian: i thought web workers were a standard
10:44 lucian homunq: no, Gears has WorkerPool
10:44 BryanWB_: the WebWorkers are standard, but only lastest firefox&webkit have it
10:44 subzero lucian: my mistake, sorry
10:44 homunq that's what I meant, each browser has it's own thing, no abstraction layer on top which hides the differences.
10:44 BryanWB_ lucian: which is fine, since only the latest browsers support html5 anywars
10:44 lucian the WorkerPool in Gears is for older browsers
10:44 subzero yes, that's right
10:45 homunq oh
10:45 heh, oops
10:45 lucian so basically WebWorkers in HTML5, WorkerPool in Gears for legacy
10:46 subzero afaik
10:46 homunq ok. is it doing the right tricks for you?
10:46 BryanWB_ homunq: either chrome and mozilla will settle on a standard web worker api or js libraries like jQuery will build an abstraction that works across them
10:46 homunq "will" :(
10:46 subzero homunq: an approach, since it runs js in parallel but don't modify dom elements
10:46 homunq I guess you are on the bleeding edge.
10:47 subzero a little..
10:47 homunq only one thread can modify, the others can just look? OK, so sound is in the background, right?
10:48 BryanWB_: still here, right?
10:48 BryanWB_ is still here
10:48 subzero about sound, I think each thead can run its proper sound element
10:49 lucian subzero: from what usage of WebWorkers i've seen, people use them for processing data and just return things to the main thread
10:49 subzero: like John Resig
10:50 subzero lucian: yes, I've seen the same
10:50 what I think is to process the anited stuff in threads
10:50 homunq anited?
10:50 BryanWB_ subzero: and them start/stop them according to user input?
10:51 subzero s/anited/animated/
10:51 and the master will do the changes in the dom elements
10:52 that is what has happened to me...
10:53 homunq sounds a bit clumsy... you'll need some clever hacks, not clean solutions.
10:54 which makes doing your roadmap/planning especially hard
10:54 subzero I'm thinking how to get something better, the other solution is setTimeOut (js function)
10:54 homunq how can you address that?
10:55 can the main thread sleep on some flag that the other ones set?
10:55 BryanWB_ subzero: how does flash handle this kind of stuff? and can web workers do something similar?
10:55 homunq because if it's polling, that would kill you IMO
10:56 subzero homunq: about sleep the main thread, I think it's not possible...
10:56 mchua notes that marketing mtg is due to start soonish
10:56 subzero BryanWB_: it runs animation in parallel
10:56 BryanWB_ subzero: flash does?
10:57 homunq mchua: yes
10:57 #topic bemasc
10:57 assimd is not here
10:57 bemasc Hi all. Lots to report.
10:57 subzero BryanWB_: yes, that's why you can do a lot of actions whit that
10:57 homunq (I hope that we'll continue to chat after in #sugar)
10:57 link
10:57 subzero BryanWB_: I mean, multitouch
10:58 bemasc no link. You can see the git log I guess...
10:58 homunq You don't have to write much, but try to have some report to link here.
10:58 bemasc Last night, after midnight, I got a restructured version of my operational transformer working.
10:58 homunq oooh
10:59 what does that mean?
10:59 bemasc It passes my unit tests and shows a 60-fold speedup in my benchmark.
10:59 homunq wow
10:59 well, carry on then
10:59 bemasc SharedTextDemo-1 started to get slow after you'd typed a few lines.  Hopefully, this version should be fast enough for a whole essay.
10:59 homunq another 60 for next week :)
10:59 garycmartin bemasc: so we can look forward to seeing a faster 'sharedtextdemo' then?
11:00 bemasc garycmartin: hopefully by tomorrow.
11:00 Also, I found a bug in libdbus, and got them to patch it.  Once fixed, it will allow automatic persistence for all groupthink-using apps.
11:00 garycmartin bemasc: fab. (I'm very interested in using your backend to add colab to other activities)
11:00 bemasc So activity authors never have to think about the datastore; groupthink will handle serialization/deserialization for them (if they wish).
11:01 homunq yay! groupthink goes beyond groupthink, to rethink!
11:01 that is great news.
11:01 OK, I think we have to yield the room.
11:01 let's continue after-chat in #sugar.
11:01 bemasc yep
11:02 homunq any last-minute words?
11:02 5
11:02 4
11:02 3
11:02 2
11:02 1
11:02 #endmeeting

Minutes | Index | Today     Channels | Search | Join

Powered by ilbot/Modified.
Webmaster