Experimental IRC log straw-2008-01-04

Available formats: content-negotiated html turtle (see SIOC for the vocabulary)

Back to channel and daily index: content-negotiated html turtle

These logs are provided as an experiment in indexing discussions using IRCHub.py, Irc2RDF.hs, and SIOC.

07:52:31<DrFrasierCrane>hey
08:26:08<DrFrasierCrane>ha
08:26:19<DrFrasierCrane>I think I've found one bottleneck when loading thousands of items
08:27:10<DrFrasierCrane>if we don't select every item's content and load it on-demand, then speed-up is about 3x
08:28:29<DrFrasierCrane>and query takes about the same amount of time as creating Item objects for queried items hmm
08:28:41<DrFrasierCrane>maybe we could have a pool of Items and don't create them every time we query
08:28:47<DrFrasierCrane>just pack data in them
09:02:38<DrFrasierCrane>janm, hi
09:05:39<janm>hello DrFrasierCrane
09:12:31<DrFrasierCrane>have you followed the discussion recently?
09:13:02<DrFrasierCrane>specifically, a suggestion to use bugzilla more intensively for project management
09:13:11<DrFrasierCrane>tracking tasks, roadmaps etc
09:13:59<DrFrasierCrane>first, of course, clean up is needed there, I may have some extra time over the weekend so I could do it if you give me some privileges in bugzilla for the straw product
09:17:45<janm>DrFrasierCrane: i'll read up the discussion on the chatlogs thanks. as for the bugzilla privilege, i'll check it out a bit later. i'll give you one once i figured out how =)
09:18:17<DrFrasierCrane>cool, I'm leaving for work in a minute anyway
09:18:21<DrFrasierCrane>btw..
09:18:27<DrFrasierCrane>http://bugzilla.gnome.org/show_bug.cgi?id=501029
09:18:32<DrFrasierCrane>is this still an issue?
09:18:41<DrFrasierCrane>looks serious
09:19:41<janm>maybe i haven't looked.
09:19:48<DrFrasierCrane>ok
09:23:26<DrFrasierCrane>I'm off, see you later
09:23:30<janm>ok, cya!
11:54:46<DrFrasierCrane>tuukkah, hehe they correted CDATA syntax at jdocs.com :)
11:56:23<tuukkah>web gets better one page at a time
11:56:33<DrFrasierCrane>lol
11:58:06<DrFrasierCrane>have you read my braindump about db performance?
11:58:14<DrFrasierCrane>it doesn't look that bad..
11:58:29<DrFrasierCrane>although for now showing 5000 takes 1 seconds from start to finish
11:58:43<DrFrasierCrane>50% is a query, another 50% creating Item objects and filling gtk store
11:59:01<DrFrasierCrane>(at my 1.3ghz laptop)
11:59:42<DrFrasierCrane>on the other hand, I think we should avoid premature optimization after all
12:01:09<tuukkah>do you create a new item object for each query result?
12:01:42<DrFrasierCrane>new item object for each row of each query result, yes
12:02:59<DrFrasierCrane>I guess we must also keep memory usage constraints in mind
12:03:18<DrFrasierCrane>memory is cheap nowadays, but look at how people whine about Firefox using 100 MB etc.
12:06:38<kallepersson>Yeah
12:07:00<kallepersson>Just because memory is cheap doesn't mean that we should get 4 gig of ram just to run some basic apps
12:07:07<kallepersson>look at Vista for instance
12:07:13<kallepersson>it takes 512 mb RAM to just RUN.
12:07:16<kallepersson>That's crazy.
12:07:35<kallepersson>Requires a 2 gig memory to run without lag
12:08:47<DrFrasierCrane>I agree that not every application should have the right to take as much memory as it wants, it should be ordered by application importance
12:09:06<DrFrasierCrane>web browser nowadays is part of the core in every "home" system
12:09:18<DrFrasierCrane>so in case of Firefox I don't understand people really
12:11:23<DrFrasierCrane>btw kallepersson did you read my comment on Straw menus?
12:13:26<tuukkah>the problem with firefox has long been that it leaks, and it leaks a lot. but i hear they've fixed a lot of this for firefox 3
12:14:25<tuukkah>DrFrasierCrane, re creating an item object for each row of each result, why don't you cache the item objects by id?
12:14:40<DrFrasierCrane>tuukkah, I've been using their trunk builds for 3 years now and I haven't noticed *any* extensive memory usage, although I've seen screenshots where people showed firefox.exe using > 1GB
12:15:35<tuukkah>DrFrasierCrane, do you run the browser for weeks or restart it every 4 hours or so?
12:16:52<DrFrasierCrane>weeks - no, days - maybe, but most often is about 10 hours a day, I know that leaks exist, but they really got around to fixing this stuff, although given their wicked architecture, it's not easy.. but people are helping with tools like leak monitor etc.
12:18:25<tuukkah>yeah like i said, it's getting better
12:19:59<DrFrasierCrane>yes I'm thinking about implementing some kind of cache
12:20:09<DrFrasierCrane>but details on how to invalidate are still blurry
12:20:29<tuukkah>i don't see how you need to invalidate
12:22:18<DrFrasierCrane>hmm I found some case but can't recall it now :)
12:27:34<DrFrasierCrane>I guess I can try that then
12:31:05<tuukkah>well, we can think about this more, i'm just mapping the situation
12:31:56<DrFrasierCrane>actually I haven't even thought about such simple by-id caching :) I tend to overcomplicate things ;)
12:32:16<DrFrasierCrane>I was thinking more like a per-query or per-feed
12:32:32<DrFrasierCrane>but then invalidation gets really tricky
12:42:36<DrFrasierCrane>also I'm thinking about some more generic way of making sure that every change made to a persistent property is guaranteed to be persisted
12:43:28<tuukkah>acid?
12:44:08<DrFrasierCrane>I don't mean on the database level, but on application level
12:52:35<tuukkah>but are you talking about the same kind of guarantees in case of loss of power etc?
12:55:08<DrFrasierCrane>yeah that too, but in general contract ORM mapper includes synchronization of the object model and the db
12:55:31<tuukkah>right
12:55:44<DrFrasierCrane>our little ORM engine doesn't even manage the objects so this burden is on the application currently
12:56:06<tuukkah>our little ORM engine seems to cause some burden ;-)
12:56:47<DrFrasierCrane>I guess it shields us from more burden than it creates :)
12:56:59<DrFrasierCrane>(hopefully)
12:58:09<tuukkah>i know :-)
12:59:04<tuukkah>our objects seldom change, right?
13:00:32<DrFrasierCrane>yes
13:02:05<tuukkah>so it doesn't really matter how we do it, as long as we do it
13:02:59<DrFrasierCrane>yeah that is the current philosophy
13:03:26<tuukkah>except perhaps during some mass operations
13:06:32<tuukkah>updating a feed, or let's say, marking all items of all categories as read at once
13:06:50<tuukkah>all items of all feeds of all categories :-)
13:07:17<DrFrasierCrane>that's technically one sql update :)
13:08:06<DrFrasierCrane>but no ORM will generate one update for such operation
13:09:05<tuukkah>why not
13:10:01<DrFrasierCrane>sorry I mean when you do it in bulk on all objects and then commit transaction, flush session or whatever
13:10:12<DrFrasierCrane>same goes for foreign key cascade delete operations
13:10:19<DrFrasierCrane>they don't use it
14:39:49<DrFrasierCrane>tuukkah, just got back from the meeting, mentioned Straw
14:41:06<DrFrasierCrane>I am to prepare a general document describing discussion tracker's goals, where we are and how we plan to reach these goals
14:45:46<tuukkah>that sounds good, doesn't it?
14:46:19<tuukkah>i just got back from the customs, got my olpc xos :-)
14:46:40<DrFrasierCrane>yeah sounds good indeed :)
14:46:42<DrFrasierCrane>xos?
14:47:10<DrFrasierCrane>aah
14:47:29<DrFrasierCrane>I misread that as sioc xos :)
14:55:19<DrFrasierCrane>how is this olpc laptop?
14:55:44<DrFrasierCrane>can you install any apps on it? I read that it has something called "activities" instead of apps
15:18:37<DrFrasierCrane>janm, ping
15:40:33<tuukkah>i found how to start xterm via the python activity =)
15:40:50<DrFrasierCrane>:D
16:54:41<DrFrasierCrane>see you folks later
19:39:03<DrFrasierCrane>janm, when you have the time, please add me to devs in bugzilla, I would really want to spend some time on it this weekend.. thanks
19:39:12<DrFrasierCrane>in the meantime, I'm off, bye all
19:40:25<janm>DrFrasierCrane: what's your userid/email address?
19:41:26<janm>DrFrasierCrane: just email it to me or privmsg it
19:44:29<janm>tuukkah: you there?
19:45:14<tuukkah>yeah
19:45:24<tuukkah>hi janm
19:45:29<tuukkah>what's up?
19:46:37<janm>hey, would you like your bugzilla account to be added as well?
19:47:48<tuukkah>sure, thanks
19:48:15<janm>hmm, can you msg me your bugzilla email? thanks
19:52:32<tuukkah>how do we want to use bugzilla?
19:53:59<janm>i think we should it as our todo, plans, defects and tests manager
19:54:34<tuukkah>but anything that should be editable text stays in the wiki right?
19:56:06<janm>tuukkah: yeah. hmm.. should plans stay in the wiki as well?
20:04:35<tuukkah>i don't know
20:07:37<DrFrasierCrane>we'll see how it goes I think..
20:07:42<DrFrasierCrane>good nigth
20:09:55<janm>tuukkah: with release stuff, are we looking at improving the current future and do a release based on that?
20:10:19<tuukkah>that's the big question. what do you think?
20:11:48<janm>the reason i'm askin is if that's the case (i.e. improving future), then i'm inclined to move future into HEAD
20:12:05<janm>what do you think?
20:12:35<tuukkah>we're taking a big risk with the future
20:14:31<janm>what do you think are the risks?
20:14:42<tuukkah>i can't even approximate how much work remains to be done since there's so much that has changed and i don't know much about a lot of it
20:15:15<janm>ok, i'll leave it for now
20:15:23<tuukkah>specifically, at least the orm stuff
20:16:24<janm>yeah same here.
20:19:52<tuukkah>i noticed you've done good work on pygtkwebkit
20:20:01<tuukkah>are there apps using it already?
20:23:20<janm>there were few people asking me questions about if it does this and that
20:23:34<janm>i don't know any specific app that uses it though
20:24:46<tuukkah>perhaps it's a bit too new for that
20:25:09<tuukkah>conduit might now be in the same stage as straw
20:25:39<janm>yeah. conduit? are they using it?
20:25:54<tuukkah>they wanted to use documentviews for it
20:26:45<janm>cool
20:27:34<tuukkah>seems they've built on it: http://www.conduit-project.org/browser/trunk/conduit/Web.py
20:27:42<DrFrasierCrane>changed my mind about going to sleep :D I see that you folks have doubts about "future"
20:28:29<DrFrasierCrane>what can I do to make you more comfortable? :)
20:28:51<tuukkah>actually, gtkhtml2 and webkit parts were removed in revision 926
20:28:53<DrFrasierCrane>should I start documenting stuff? I know there's not much docs for now, I'm just trying to do as much as I can
20:29:35<janm>tuukkah: eh, they just like the abstraction? =)
20:30:01<DrFrasierCrane>I understand your concerns, but I think it can really be a big step forward, escpecially with tuukkah's discussion tracker on the horizon and all..
20:30:27<tuukkah>janm, the comment now says "Basic webbrowser abstraction to provide an upgrade path to webkit from gtkmozembed"
20:31:14<janm>tuukkah: ahh, makes sense =)
20:32:19<janm>DrFrasierCrane: i think we just need to try and poke inside the hood and ask you specific questions. we can always ask you when you're awake, i guess =)
20:32:20<tuukkah>DrFrasierCrane, i know it's a big step forward
20:33:13<DrFrasierCrane>janm, sure
20:33:20<janm>tuukkah: i'm thinking of starting another project to provide python bindings to JavaScriptCore, but i don't know the details of that yet or if I can do it
20:33:37<janm>*time to do it
20:34:33<tuukkah>janm, what are those bindings needed for?
20:36:34<janm>tuukkah: it will allow you to manipulate the DOM in a pythonic way
20:37:28<janm>so instead of using javascript, you use python
20:38:23<tuukkah>hmm, so a PythonCore in that sense :-)
20:40:40<tuukkah>there are so many uses for "a html widget" the requirements can vary a lot
20:41:44<tuukkah>in that sense, more example apps would be good in showing what you can do already
20:43:11<tuukkah>and incomplete example apps would be good in showing what you can't do yet :-)
20:44:07<janm>indeed. we should probably ship a widget engine using pywebkitgtk
20:45:02<tuukkah>would that fall into what you can do or can't do ?-)
20:46:52<janm>can do
20:47:36<tuukkah>hmm, what kind of widget engine do you have in mind?
20:47:44<janm>i think i managed to get jackfield to render widgets using webkit
20:48:00<tuukkah>cool
20:49:34<janm>.. or probably port some widget engine to use pywebkitgtk and add those in "Applications that use PyWebKitGtk" =)
20:50:04<tuukkah>heh =)
20:50:59<tuukkah>or make widgetengine.py abstraction
20:51:15<tuukkah>no jackfield news since september :-(
20:51:30<janm>yeah, cool and bundle that :)
20:51:58<janm>tuukkah: yeah i think it's just a toy. i sent him a webkit patch but never got a reply :(
20:53:13<tuukkah>but if you hacked it to work with pywebkitgtk, it's a valid example still
20:58:03<janm>let's fork it!
20:58:10<janm>:)
21:00:01<tuukkah>yeah =)
21:14:33<DrFrasierCrane>bye
22:04:57<kallepersson>speaking of WebKit, I just got Safari working through Wine
22:05:06<kallepersson>Oh happy day!
22:07:27<tuukkah>can i somehow provocate you to work on a web browser that's better than safari and written using pywebkitgtk ?-)
22:21:09<kallepersson>Epiphany
22:21:37<kallepersson>not written in pywebkitgtk though, but webkitgtk nonetheless
22:22:49<tuukkah>do you prefer epiphany?
22:27:02<kallepersson>Over?
22:27:21<tuukkah>firefox?
22:27:40<kallepersson>For regular browsing, yes. For developing, no.
22:28:05<kallepersson>All the extensions I use in Fx are web development tools.
22:28:14<tuukkah>i see
22:28:36<kallepersson>Since I don't need them when I'm just browsing stuff I tend to use the lighter Epiphany instead
22:28:49<tuukkah>what abot safari then?
22:29:07<kallepersson>That's on my 265mb RAM home computer then. At work I use Fx for everything.
22:29:31<kallepersson>I just need Safari to test websites with, I'd never use it for regular surfing, since it's slow and quite buggy
22:29:54<kallepersson>Also, I don't like to run stuff trough Wine that I "depend" on, no matter how weird it sounds :-)
22:30:10<tuukkah>:-)
22:30:23<kallepersson>What is your opinion on the matter?
22:31:22<tuukkah>somehow it's stupid that firefox is like an operating system of its own, but then again web is so important that i'm willing to use firefox for the best experience
22:32:22<kallepersson>Epiphany is not different to Firefox at all web experience-wise, IMHO
22:33:37<tuukkah>yeah, the using iceweasel instead of the actual firefox removes some of the advantage too
22:34:22<tuukkah>but when i've tried to use epiphany it has been too bare-bones and too fringe
22:35:37<kallepersson>I see
22:35:45<kallepersson>Well, it's quite different
22:35:49<kallepersson>But not very
22:36:02<tuukkah>although i suppose most of my extensions are for developers too
22:36:30<tuukkah>and epiphany may have some stuff built-in like session manager
22:38:20<kallepersson>Speaking of pywegkitgtk: http://blogs.gnome.org/johan/2007/11/30/future-of-gnome-language-bindings/
22:38:21<kallepersson>:-)
22:41:39<tuukkah>cool although it's only a prototype at this point
22:42:28<kallepersson>Yeah
22:43:38<janm>ACTION hopes to have a GObject bindings to the DOM so we can instrospect the DOM!
22:45:45<tuukkah>"Encourage upstream projects to include the metadata" - haven't there been binding systems that would've worked with this before?
22:46:07<tuukkah>guile?
22:47:05<tuukkah>swig?
22:48:28<tuukkah>but perhaps here the point is to not support more than gobject so it can be supported better

Back to channel and daily index: content-negotiated html turtle