MoinMoin: MoinMoinChat/Logs/moin-dev/2008-03-28

2008-03-28T00:02:15  <gizmach> new day
2008-03-28T00:02:16  <gizmach> :D
2008-03-28T00:21:29  <zenhase> well, depends on the timezone ;)
2008-03-28T00:36:43  <TheSheep> xorAxAx: sorry, was afk
2008-03-28T00:38:13  <TheSheep> the idea for splitting goes like this: ([new dom formatter]<--[old formatter api])<--([old parser api]<--[new dom parser])
2008-03-28T00:38:48  <TheSheep> then you can remove the compatibility layers like this [new dom formatter]<--[new moin parser]
2008-03-28T00:40:00  <TheSheep> the first, third and fourth arrows mark passing of DOM trees, the second arrow marks the calls of functions from old formatter api
2008-03-28T00:52:42  * zenhase feels like reborn
2008-03-28T00:53:05  <zenhase> climbing, dinner, shower ... now i feel like i could take on tasks, my mind ist quite at peace right now
2008-03-28T01:18:37  <xorAxAx> TheSheep: that "graph" doesnt make sense
2008-03-28T01:18:51  <xorAxAx> TheSheep: please visualise an api differently than a component
2008-03-28T01:24:30  <TheSheep> xorAxAx: imagine a formatter that outputs a dom tree instead of html
2008-03-28T01:25:40  <xorAxAx> TheSheep: yes, as i said
2008-03-28T01:27:10  <TheSheep> I'm probably too dumb to see the problem in that, I guess I'll need some mockups
2008-03-28T01:28:00  <xorAxAx> which problem?
2008-03-28T01:32:31  <TheSheep> xorAxAx: any of them
2008-03-28T01:32:41  <xorAxAx> ?
2008-03-28T01:32:52  <xorAxAx> not very helpful reply, good night!
2008-03-28T01:35:53  <TheSheep> good night
2008-03-28T08:28:34  <ThomasWaldmann> moin
2008-03-28T09:06:44  <dreimark> moin
2008-03-28T10:34:28  <zenhase> moin
2008-03-28T10:37:48  <dreimark> hi zenhase
2008-03-28T10:39:33  <zenhase> hi dreimark
2008-03-28T10:54:12  <dreimark> zenhase: do you got some appetize for a new soc project ?
2008-03-28T10:56:13  <zenhase> actually yes, after climbing yesterday i thought about it
2008-03-28T10:58:08  <zenhase> i will have to think about what i could do now
2008-03-28T10:58:14  <dreimark> :)
2008-03-28T11:00:36  <zenhase> the DOM tree idea looks quite interesting, but it seems there are already students applying for it
2008-03-28T11:01:38  <zenhase> loosening up formatter/parser coupling could ease up a send_page-refactoring too, i think
2008-03-28T11:14:44  <zenhase> i think i will have to evaluate some options, perhaps take some ideas from my last years analysis and then i will cook up an application today or tomorrow
2008-03-28T13:45:26  <kikka> Moin
2008-03-28T13:55:25  <ThomasWaldmann> hi kikka
2008-03-28T13:55:41  <TheSheep> hi kikka
2008-03-28T14:16:21  <kikka> Hey ThomasWaldmann and TheSheep :)
2008-03-28T14:19:06  <TheSheep> kikka: you're startingi in gsoc?
2008-03-28T14:21:39  <kikka> I think, I'm too young.
2008-03-28T14:23:46  <kikka> And not capable for gsoc (-:
2008-03-28T14:24:36  <TheSheep> ah
2008-03-28T14:41:55  <zenhase> re
2008-03-28T14:44:11  <kikka> hey zenhase. :)
2008-03-28T14:45:20  <zenhase> hiho kikka
2008-03-28T14:58:35  <dreimark> hi kikka
2008-03-28T14:58:59  <kikka> huhu dreimark :)
2008-03-28T15:51:57  <johill> xorAxAx: of course, it's on git.sipsolutions.net
2008-03-28T16:01:23  <johill> damn. I found another design bug in the storage code, it's not possible to implement ACL checking for the 'delete' ACL without messing up storage before-hand
2008-03-28T16:02:28  <xorAxAx> hmm
2008-03-28T16:02:40  <johill> since you need to create_revision and *then* set the DELETED metadata
2008-03-28T16:02:50  <johill> and then the revision already exists. bugger
2008-03-28T16:03:06  <johill> I have ideas how to fix all that, but it requires mostly redesigning it
2008-03-28T16:03:16  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3764:6fa56b0f0de4 1.7-storage-hwendel/ (.hgignore wikiconfig.py): re-add wikiconfig_local a bit differently
2008-03-28T16:03:20  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3765:0917c9d3dc25 1.7-storage-hwendel/ (.hgignore wikiconfig.py): merge in wikiconfig_local change from main branch
2008-03-28T16:03:21  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3766:0f74f7e08e51 1.7-storage-hwendel/docs/CHANGES.storage: more notes on metadata
2008-03-28T16:03:21  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3767:1c48f6a3e2ab 1.7-storage-hwendel/BUGS: add note to BUGS file about edit-locking
2008-03-28T16:03:25  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3768:bb9a23e6a6e3 1.7-storage-hwendel/ (3 files in 2 dirs): add sql backend, notes on implementing storage LOCKFREE
2008-03-28T16:03:25  <xorAxAx> maybe deleted should get special handling?
2008-03-28T16:03:34  <johill> even then
2008-03-28T16:03:43  <johill> you need to check before create_revision
2008-03-28T16:03:51  <johill> since a revision is marked as deleted in any way
2008-03-28T16:03:55  <johill> unless you want to create_deleted_revision!
2008-03-28T16:04:25  <johill> I think the split between create/save is the big problem
2008-03-28T16:04:47  <johill> my design idea removes that, allows operating lock-free, and thus solves a bunch of problems like this
2008-03-28T16:04:57  <xorAxAx> can you write it down somewhere?
2008-03-28T16:05:16  <johill> http://hg.moinmo.in/moin/1.7-storage-hwendel/file/tip/LOCKFREE
2008-03-28T16:06:56  <johill> for the fun of it, I implemented an sql backend
2008-03-28T16:07:04  <johill> to see where the real problems are
2008-03-28T16:07:26  <xorAxAx> yeah
2008-03-28T16:07:33  <johill> it's heavily commented, in http://hg.moinmo.in/moin/1.7-storage-hwendel/file/tip/MoinMoin/storage/backends/sql1.py
2008-03-28T16:08:07  <ThomasWaldmann> johill: but you leave something todo for our soc student? :)
2008-03-28T16:08:38  <johill> oh. do I have to? ;) sure, lots actually since all the problems I see mean that things need to be reworked a lot
2008-03-28T16:09:18  <ThomasWaldmann> bbl
2008-03-28T16:12:47  <CIA-39> Reimar Bauer <rb.proj AT googlemail DOT com> default * 3424:aa3890b17212 1.7/MoinMoin/macro/Hits.py: macro.Hits: incredible speed up using MoinMoin.stats.hitcounts
2008-03-28T16:12:48  <CIA-39> Reimar Bauer <rb.proj AT googlemail DOT com> default * 3425:1729bf752c83 1.7/ (.hgignore wikiconfig.py): merged main
2008-03-28T16:13:04  * dennda rants
2008-03-28T16:27:16  <johill> dennda: about what?
2008-03-28T16:35:17  <dennda> johill: /me needs to do quite some work for university until monday
2008-03-28T16:35:30  <dennda> and my team is not working efficiently
2008-03-28T16:36:39  <dennda> team == toll ein anderer machts :)
2008-03-28T16:37:08  <dreimark> dennda: I think it is clear that you have do/to write tests for your code too
2008-03-28T16:37:33  <dennda> dreimark: yes, I know. It's part of the plan
2008-03-28T16:37:56  <johill> dennda: oh, I know the feeling, I used to get out by implementing a good chunk up front myself and then pointing to that ;)
2008-03-28T16:45:50  <dennda> dreimark: what made you think of that?
2008-03-28T16:47:25  <dreimark> dennda: I did read your deliverables
2008-03-28T16:48:12  <dreimark> testing of old code ends at 7th ;)
2008-03-28T16:49:50  <dreimark> johill dennda : server restart with storage branch should not become much more difficultier as now
2008-03-28T16:50:22  <dreimark> currently sometimes pages are not accessible because of left over edit-lock files
2008-03-28T16:53:13  <johill> dreimark: interesting, they should be timed out though
2008-03-28T16:53:30  <CIA-39> Johannes Berg <johannes AT sipsolutions DOT net> default * 3769:c42396da59f0 1.7-storage-hwendel/BUGS: document acl-checking design problem
2008-03-28T16:56:01  <dreimark> strange "malformed header from script. Bad header=20216: moin.cgi" and a reload solves it
2008-03-28T16:56:23  <johill> sounds like it threw an exception and cgitb threw one as well
2008-03-28T16:56:36  <johill> then you get a plain text exception when running on standalone
2008-03-28T16:57:09  <dennda> dreimark: that was meant to include the code I will be writing
2008-03-28T16:57:25  <dennda> s/the code/the tests/
2008-03-28T16:57:29  <dreimark> dennda: :)
2008-03-28T16:57:53  <dennda> dreimark: or do you think that time span is too short?
2008-03-28T16:57:57  <dreimark> johill: it was on cgi
2008-03-28T16:58:19  <johill> dreimark: yeah, not sure how headers work there, might be a bug with the exception catcher in cgitb
2008-03-28T16:58:20  <dennda> I found out that I can start working on it even before coding starts officially and I intend to do that if I am chosen
2008-03-28T17:02:11  <dreimark> dennda: great
2008-03-28T17:05:10  <dennda> why do these routers die that often?
2008-03-28T17:31:18  <johill> routers die? never happened to me
2008-03-28T17:31:32  <johill> dennda: or is that a wrt* router? they have a firmware bug that kills the nvram
2008-03-28T17:38:10  <dennda> johill: The router here in my flat crashes all the time, the one my parents have at home crashes all the time, the one my girlfriend has crashes all the time... Believe me, they *do* crash
2008-03-28T17:38:30  <dennda> johill: have a look at this: http://youtube.com/watch?v=_y36fG2Oba0
2008-03-28T17:38:31  <kikka> No! :P
2008-03-28T17:43:05  <dreimark> hehe
2008-03-28T18:15:08  <johill> dennda: I know that :)
2008-03-28T18:28:25  <dreimark> bbl
2008-03-28T18:55:17  <johill> dennda: they just have crappy software. install openwrt :P
2008-03-28T18:58:40  <dennda> don't think it's compatible
2008-03-28T18:58:45  <dennda> must check
2008-03-28T19:00:36  <dennda> boy, it's listed! thanks for making me check
2008-03-28T19:00:40  <dennda> I'll tackle that
2008-03-28T19:29:43  <xorAxAx> dennda: crash != die )
2008-03-28T19:31:04  <dennda> xorAxAx: it's the same. it just needs to be resurrected :)
2008-03-28T19:36:49  <gizmach> hi all
2008-03-28T19:36:58  <dreimark> hi gizmach
2008-03-28T19:52:27  <johill> dennda: apparently, many routers have a firmware bug that means on any checksum error the whole nvram is deleted and the whole thing just won't work
2008-03-28T19:52:36  <johill> dennda: hence, the new openwrt no longer uses nvram
2008-03-28T20:04:28  * dennda will follow your advice
2008-03-28T20:17:25  <dreimark> http://www.dynamicdrive.com/style/csslibrary/item/css-tableless-form/
2008-03-28T20:19:32  <dreimark> TheSheep: that on is nice too http://www.dynamicdrive.com/style/csslibrary/item/animated-link-underlines/
2008-03-28T20:22:00  <johill> eww
2008-03-28T20:34:29  <dreimark> johill: http://www.qxm.de/webdesign/20070814-182313/supereinfache-css-schatten-fuer-bilder
2008-03-28T20:34:40  * dreimark looks for nice css for arnica
2008-03-28T20:42:25  <dreimark> bbl
2008-03-28T21:13:47  <ThomasWaldmann> re
2008-03-28T21:23:44  <xorAxAx> 21:23:17 <@kblin> Sledge__: yup, same here. Samba has a grand total of 1 app
2008-03-28T21:37:44  <xorAxAx> 21:37:34 < boote> LH: I like the extension to the deadline as well. We have only had 3 applicants this year. In previous years, we have had more like 15-20 by this time.
2008-03-28T21:49:04  <xorAxAx> it looks like there will be a timeline extension of one week
2008-03-28T21:57:29  <TheSheep> xorAxAx: we have more mentors than applications?
2008-03-28T21:57:48  <xorAxAx> TheSheep: hmm?
2008-03-28T21:59:19  <TheSheep> or is it normal?
2008-03-28T21:59:46  <xorAxAx> well, we dont know the final application count yet
2008-03-28T22:02:30  <gizmach> xorAxAx: did u look at the DOM application
2008-03-28T22:02:47  <xorAxAx> gizmach: no
2008-03-28T22:05:10  <TheSheep> gizmach: what does it mean that 'the tree uses proper namespaces'?
2008-03-28T22:05:29  <TheSheep> gizmach: XML namespaces?
2008-03-28T22:09:00  <johill> and why do you think DOM == XML?
2008-03-28T22:09:38  <gizmach> johill: I don't think DOM == XML? what are we talking about
2008-03-28T22:10:13  <johill> TheSheep
2008-03-28T22:10:32  <johill> but I think I'm confused
2008-03-28T22:10:42  <johill> your application does say the dom tree uses elementtree stuff?
2008-03-28T22:10:49  <TheSheep> johill: XML is the closest thing mentioned in there that has namespaces
2008-03-28T22:10:54  <gizmach> johill:  I didn't posted an application
2008-03-28T22:11:05  <TheSheep> gizmach: so it's not yours :/
2008-03-28T22:11:19  <TheSheep> he fogot to provide his nick
2008-03-28T22:11:23  <TheSheep> forgot
2008-03-28T22:11:30  <gizmach> TheSheep: no its not mine, and I don't have that idea
2008-03-28T22:11:31  <gizmach> :D
2008-03-28T22:11:32  <johill> aha, I was just following TheSheep's assumption it was :)
2008-03-28T22:11:34  <gizmach> TheSheep: ok
2008-03-28T22:11:49  <TheSheep> sorry for the confusion
2008-03-28T22:11:54  <dennda> hu? how can you file an application without a google account?
2008-03-28T22:12:18  <xorAxAx> dennda: not at all
2008-03-28T22:12:22  <TheSheep> dennda: I meant nick on irc
2008-03-28T22:12:38  <johill> bbl
2008-03-28T22:12:41  <dennda> I am wondering if the three of you confused me as well
2008-03-28T22:16:57  * TheSheep spreads chaos on the channel
2008-03-28T22:21:22  <gizmach> TheSheep: I started making my appliction for DOM but I heard that waldi sent an application for that project allready so I was considering of giving up from writing and posting it.
2008-03-28T22:22:11  <TheSheep> gizmach: this might be the worst thing to do, I guess. You can always post several applications.
2008-03-28T22:23:34  <xorAxAx> maybe we like you much better than him ... he seems a bit uncommunicative :)
2008-03-28T22:24:07  <xorAxAx> now i just finished a wesnoth session against my brother and will have a look at the applications
2008-03-28T22:25:35  <xorAxAx> gizmach: what was the url again?
2008-03-28T22:25:37  <dennda> wesnoth rocks
2008-03-28T22:25:45  <gizmach> TheSheep: that was the only project I was really interested in :D Like a little child when it wants green ball with yellow spots and only that one :D ok, that was horrable example
2008-03-28T22:25:46  <xorAxAx> couldnt find it in the logs
2008-03-28T22:25:55  <gizmach> xorAxAx: for what?
2008-03-28T22:25:59  <gizmach> xorAxAx: DOM?
2008-03-28T22:26:01  <xorAxAx> the application text
2008-03-28T22:26:21  <gizmach> xorAxAx: I din't sent u an url for the application
2008-03-28T22:26:25  <xorAxAx> ok
2008-03-28T22:26:40  <xorAxAx> so now i check waldi
2008-03-28T22:26:49  <gizmach> xorAxAx: :)
2008-03-28T22:27:01  <xorAxAx> ah, the first mentor spam
2008-03-28T22:27:13  <dennda> mentor spam?
2008-03-28T22:29:12  <xorAxAx> dennda: random people applying as a mentor
2008-03-28T22:30:38  <TheSheep> enlarge your mentor!
2008-03-28T22:31:34  <gizmach> TheSheep: ?
2008-03-28T22:31:46  <TheSheep> ignore me
2008-03-28T22:37:55  <johill> xorAxAx: tbh, I think DOM stuff is a bit larger than a SoC project
2008-03-28T22:38:07  <xorAxAx> johill: i dont :)
2008-03-28T22:38:20  <johill> you could add some sort of DOM stuff for parser/formatter but all the issues with TOC etc. are quite problematic to solve cleanly
2008-03-28T22:38:26  <johill> xorAxAx: that's because you haven't tried doing it ;)
2008-03-28T22:38:47  <xorAxAx> johill: i agree that you need a clean design
2008-03-28T22:40:33  <johill> I tried that
2008-03-28T22:40:36  <CIA-39> Thomas Waldmann <tw AT waldmann-edv DOT de> default * 2613:5289ff0ea669 1.6/ (3 files in 2 dirs): Fix TableOfContents macro: use correct, new Include macro markup regex, fix some comments
2008-03-28T22:40:39  <johill> it actually worked too :)
2008-03-28T22:42:12  <gizmach> xorAxAx: were you maybe thinking of designing and implementing tree formatter appi as a separate project?
2008-03-28T22:42:39  <xorAxAx> gizmach: as opposed to run together with which other task?
2008-03-28T22:43:11  <gizmach> xorAxAx: with revriting the moin parser to use it
2008-03-28T22:43:21  <johill> well that's the thing
2008-03-28T22:43:27  <xorAxAx> TheSheep: please post your namespaces questions in the google webapp :)
2008-03-28T22:43:27  <johill> you need a parser that creates a tree
2008-03-28T22:43:37  <johill> then you need a formatter that reads it
2008-03-28T22:43:42  <johill> and all macros need to be changed too
2008-03-28T22:43:56  <johill> I haven't found a way to cleanly incorporate old things into a DOM-based wiki
2008-03-28T22:44:20  <xorAxAx> gizmach: a parser without an api is a bit useless. but in case of waldi writing the formatter etc. stuff, you could use that
2008-03-28T22:44:43  <xorAxAx> johill: well, the only issue i see is the dwim trafo
2008-03-28T22:44:56  <xorAxAx> johill: also, i recognised that pickling is completly unnecessary
2008-03-28T22:45:12  <xorAxAx> because the docbook formatter doesnt return anything either
2008-03-28T22:45:13  <gizmach> xorAxAx: I would like to design and write a parser that creates a tree
2008-03-28T22:45:31  <johill> xorAxAx: hold on, dwim?
2008-03-28T22:45:34  <xorAxAx> gizmach: hehe, yeah, might make sense as an application
2008-03-28T22:45:38  <xorAxAx> johill: do what i mean
2008-03-28T22:45:45  <xorAxAx> johill: the part that you deeply disliked
2008-03-28T22:45:50  <johill> xorAxAx: pickling is an easy solution to caching
2008-03-28T22:46:06  <xorAxAx> johill: ah, i meant pickled stuff as a return value from the formatter
2008-03-28T22:46:18  <xorAxAx> how to cache the beast is a different thing
2008-03-28T22:46:28  <johill> xorAxAx: ahh. now you're talking about how I initially tried doing things in the current design
2008-03-28T22:46:30  <gizmach> xorAxAx: yes only for the application, but I dono if there was a case that to people work on implementing two parts of something and then make them work together
2008-03-28T22:46:36  <johill> with returning objects pickled into strings
2008-03-28T22:46:38  <johill> yeah that was useless
2008-03-28T22:46:44  <xorAxAx> yes
2008-03-28T22:46:50  <johill> I figured that out later, you need to change the whole chain to just work on teh objects
2008-03-28T22:46:50  <TheSheep> xorAxAx: done :)
2008-03-28T22:47:03  <johill> caching is actually easy
2008-03-28T22:47:06  <johill> look at what I did in dom.git
2008-03-28T22:47:39  <johill> (git clone http://git.sipsolutions.net/dom.git)
2008-03-28T22:47:41  <xorAxAx> gizmach: well, i think it would run like in real life - both would start and then you would adopt your stuff to his api and in the mean time you should work closely together ... but its not set that waldis proposal makes sense, i am still figuring that out :)
2008-03-28T22:48:29  <johill> xorAxAx: somebody (TheSheep?) has something where just the parser communicates with DOM with some other code, but that's pretty useless too
2008-03-28T22:48:47  <xorAxAx> johill: why is that "useless"?
2008-03-28T22:48:56  <gizmach> xorAxAx: well I don't know if mine will make sense
2008-03-28T22:48:58  <johill> well, it doesn't help
2008-03-28T22:49:06  <xorAxAx> (not sure why i would use that except as some inspiration for the apis and data schemas involved)
2008-03-28T22:49:09  <dennda> How many applications are there now? 3? 4?
2008-03-28T22:49:11  <xorAxAx> johill: for which issue? :)
2008-03-28T22:49:17  <xorAxAx> gizmach: your what? :)
2008-03-28T22:49:19  <xorAxAx> dennda: 3
2008-03-28T22:49:21  <johill> it doesn't gain you anything, the nice thing about DOM is how you can implement a TOC macro trivially for example
2008-03-28T22:49:28  <gizmach> xorAxAx: application
2008-03-28T22:49:47  <xorAxAx> johill: well, not completly true. just compare his code to the moin default parser :)
2008-03-28T22:49:55  <waldi> johill: well, i know at least one output format where TOC is a nop
2008-03-28T22:49:56  <xorAxAx> gizmach: well, we can review before you submit
2008-03-28T22:50:06  <xorAxAx> gizmach: and give you comments
2008-03-28T22:50:23  <johill> xorAxAx: yeah
2008-03-28T22:50:38  <gizmach> xorAxAx:  : I true
2008-03-28T22:50:39  <johill> waldi: hm? explain
2008-03-28T22:50:59  <johill> waldi: something where the viewer in the end builds the toc?
2008-03-28T22:51:03  <waldi> johill: yep
2008-03-28T22:51:09  <xorAxAx> hint hint, docbook
2008-03-28T22:51:20  <johill> right but that's non-trivial to do too without special-casing the TOC macro
2008-03-28T22:51:37  <waldi> i thought about ODF
2008-03-28T22:51:46  <xorAxAx> i guess ODF special cases nearly anything
2008-03-28T22:51:51  <xorAxAx> and OOXML even more :-)
2008-03-28T22:51:59  <johill> anyway, I don't care too much about this, it was a nice thought experiment at the time ;)
2008-03-28T22:51:59  <waldi> xorAxAx: no
2008-03-28T22:52:22  <xorAxAx> waldi: so, how does the toc look like in odf and dont you think that its a bit too heavy? :)
2008-03-28T22:52:40  <xorAxAx> also, you need to think how to come from a macro to a document node that means "toc" :)
2008-03-28T22:52:42  <waldi> xorAxAx: ooxml is nasty. odf is something between docbook and html, its nice
2008-03-28T22:53:05  <johill> xorAxAx: also, afaik, the creole dom thing doesn't have any sort of tree model, it's just a tree
2008-03-28T22:53:12  <xorAxAx> well, ok, i guess this is a just a matter of mapping
2008-03-28T22:53:29  <johill> and the tree-model really is the hard thing
2008-03-28T22:54:31  <xorAxAx> johill: model as in the document semantics or the operational evaluation semantics`
2008-03-28T22:54:47  <waldi> take a look at ODF, they worked several years on the spec
2008-03-28T22:55:03  <xorAxAx> yeah
2008-03-28T22:55:15  <johill> more thinking of the nesting semantics etc.
2008-03-28T22:55:22  <xorAxAx> i guess ooxml did as well according to their marketing stuff, but they would include the work done for word 1.0
2008-03-28T22:56:29  <waldi> ooxml is a xml formated memory dump
2008-03-28T22:56:44  <johill> word always had such file formats
2008-03-28T22:56:56  <xorAxAx> waldi: hehe, yeah
2008-03-28T22:57:10  <johill> I was waiting for something new and surprising ;)
2008-03-28T22:57:20  <johill> I'd much rather talk about storage :P
2008-03-28T22:57:25  <waldi> xorAxAx: bitmasks, binary data
2008-03-28T23:02:04  <dennda> johill: I'm your man
2008-03-28T23:02:42  <johill> have you seen the sql backend? :)
2008-03-28T23:03:27  <dennda> oh did you start it?
2008-03-28T23:03:36  <johill> it works
2008-03-28T23:03:48  <johill> I'm running a local wiki on it
2008-03-28T23:03:51  <dennda> oh
2008-03-28T23:04:00  <dennda> you stole one of my points :)
2008-03-28T23:04:35  <dennda> Unfortunately I can't dive into it right now. I still have a bunch of work to be done before monday
2008-03-28T23:04:37  <xorAxAx> then reword it to "refactor johill's backend after fixing all api quirks" :-)
2008-03-28T23:07:45  <johill> yeah
2008-03-28T23:08:01  <johill> it actually comes with a disclaimer to not use it and rewrite it ;)
2008-03-28T23:08:22  <xorAxAx> you could also try a couchdb backend
2008-03-28T23:08:30  <johill> whatever that is
2008-03-28T23:09:34  <johill> anyway, the api isn't suited to implementing anything properly, imnsho
2008-03-28T23:12:17  <johill> not to diss the work done, it's great to see things abstracted away at all
2008-03-28T23:15:46  <johill> what I'd actually like to see is comments on the lock-free design
2008-03-28T23:17:32  <dennda> you want a lock free design?
2008-03-28T23:17:51  <xorAxAx> johill: where can i see it? :)
2008-03-28T23:18:15  <xorAxAx> in LOCKFREE? :)
2008-03-28T23:18:37  <johill> no, LOCKFREE contains something entirely different ;)
2008-03-28T23:18:43  <xorAxAx> ok, i am having a look
2008-03-28T23:19:02  <johill> dennda: at the backend level it might actually be necessary to lock some things, but from a front-end code POV I think it'd be better
2008-03-28T23:20:04  <johill> xorAxAx: especially the edit-lock stuff I'm a bit wary about. having it as metadata for an item is nice, but lots of code assumes existing page data is immutable (not like Moin's use of the word but rather the proper use)
2008-03-28T23:20:52  <johill> on the other hand, what happens if you edit a page and somebody else decides to rename it at the same time?
2008-03-28T23:21:20  <johill> currently moin totally screws up with that and redirects the editor, when he hits save, to a new editor with no content at all, but rather the "Please describe ... here"
2008-03-28T23:25:54  <johill> I'm not sure though whether the edit-lock is renamed along with it
2008-03-28T23:26:03  <johill> if it is, that would be strange and lead to rather weird behaviour
2008-03-28T23:26:21  <johill> but in the storage code it would be renamed along with it
2008-03-28T23:26:29  <johill> tbh, I'm not sure what should actually happen!
2008-03-28T23:35:29  <xorAxAx> johill: you are confusing levels :)
2008-03-28T23:35:46  <xorAxAx> edit lock is editable metadata
2008-03-28T23:35:48  <johill> I'm thinking holistic
2008-03-28T23:35:53  <johill> :P
2008-03-28T23:36:18  <johill> yeah I have reservations against that, but now I forgot why I didn't like that per-item metadata
2008-03-28T23:36:22  <xorAxAx> well, checking against the holistic picture is nice, introducing layering violations is not :-)
2008-03-28T23:37:07  <johill> just thinking about it doesn't introduce the layering violations :)
2008-03-28T23:37:21  <johill> "layering is just a way to design protocols, not to implement them" ;)
2008-03-28T23:37:47  <johill> seriously though, I think those things need to be defined before you can design a proper storage abstraction that gives you useful guarantees
2008-03-28T23:38:24  <johill> the currently defined backend operations are nicely atomic, but they don't help implement what you want
2008-03-28T23:38:39  <xorAxAx> yes
2008-03-28T23:38:53  <xorAxAx> well, one way to approach that is to describe use cases on a high level with pseudo ops
2008-03-28T23:39:12  <xorAxAx> then you can think how to pull the ops into levels and serialize them
2008-03-28T23:41:00  <johill> not sure
2008-03-28T23:41:27  <xorAxAx> use case as in "register high level edit lock"
2008-03-28T23:41:59  <xorAxAx> or different wording like "opened editor windows (user, timestamp, rev)"
2008-03-28T23:42:27  <johill> I tend to think that you lose the actual technical problems out of sight
2008-03-28T23:43:33  <xorAxAx> the ones that let moin look like linux in the end? :)
2008-03-28T23:43:43  <xorAxAx> no, these would come into play in the serialization stage
2008-03-28T23:43:56  <xorAxAx> then you can reason where to put locks and where to promise acid properties
2008-03-28T23:44:52  <johill> yeah those :P "mom, my moin crashed while booting"
2008-03-28T23:45:08  <johill> (I love how the emacs startup is actually called 'booting')
2008-03-28T23:46:57  <dennda> johill: sorry, I can't free my head before tuesday :(
2008-03-28T23:48:41  <dennda> but comments on my application are appreciated nevertheless :)
2008-03-28T23:50:03  <johill> np
2008-03-28T23:56:44  <zenhase> re