MoinMoin: MoinMoinChat/Logs/moin-dev/2007-05-18

2007-05-18T04:33:46  <flowhase> grzywacz: why are you laughing at my decision? :<
2007-05-18T06:27:48  <grzywacz> flowhase, because bsd didn't quite meet my expectations when I was trying to switch some time ago
2007-05-18T06:27:55  <grzywacz> Maybe it's matured, though.
2007-05-18T08:08:38  <ThomasWaldmann> moin
2007-05-18T08:44:53  <dreimark> moin
2007-05-18T08:52:52  <ThomasWaldmann> dreimark: i guess that "big or" in the pagedir checking code needs some improvement
2007-05-18T08:54:03  <dreimark> yeah, I was wondering if we dont have somewhere the lifetime of the lock but we havent or?
2007-05-18T08:55:30  <ThomasWaldmann> i dont know, look in the code. but I maybe would remove the check for the cache dir from that "or"
2007-05-18T08:56:44  <ThomasWaldmann> btw, that code is similar to the qualifier I used in cleanpages / pagecleaner script
2007-05-18T09:00:10  <ThomasWaldmann> and maybe check if one os.listdir is faster than multiple os.path.exists
2007-05-18T09:03:00  <dreimark> ok
2007-05-18T09:10:17  <dreimark> bbl
2007-05-18T09:59:54  <xorAxAx> lanius: i am just skimming my university slides about transactions, if you want some overview about the topic, i can give you a link :)
2007-05-18T14:16:26  <ThomasWaldmann> johill: what extra methods? IIRC they were removed some time ago from the plan.
2007-05-18T14:23:26  <johill> heh ok
2007-05-18T14:24:12  <johill> I was reading a bit through the diffs and sawa that
2007-05-18T14:24:30  <johill> still I think it should be a separate backend you can configure elsewhere
2007-05-18T15:10:08  <ThomasWaldmann> johill: my idea was to have some data = <some backend> and caching = <some backend> in the config
2007-05-18T15:10:34  <ThomasWaldmann> using same api, but different backends, if you like
2007-05-18T15:10:48  <johill> yeah
2007-05-18T15:12:08  <ThomasWaldmann> but I think that caching stuff can be done rather late and after some benchmarks
2007-05-18T15:13:27  <johill> well, does it replace the regular caching we have too?
2007-05-18T15:13:37  <johill> like for sessions? I'd think it probably should
2007-05-18T15:14:55  <ThomasWaldmann> IF we do it, it replaces caching.py
2007-05-18T15:14:56  <lanius> xorAxAx: sure
2007-05-18T15:15:26  <johill> ok
2007-05-18T15:16:04  <lanius> johill: ThomasWaldmann: Caching of PageLinks is another thing than the usual caching
2007-05-18T15:16:26  <ThomasWaldmann> but I guess it will take some time until the pages/users stuff works and we don't need to kick the working caching stuff we have until we get bored :)
2007-05-18T15:16:38  <lanius> :)
2007-05-18T15:16:42  <johill> heh
2007-05-18T15:17:24  <lanius> i think it will make sense to start implementing soon, but which part to start with?
2007-05-18T15:19:07  <ThomasWaldmann> lanius: there are still some comments that have answers, but need more clarification
2007-05-18T15:19:20  <ThomasWaldmann> or corrections
2007-05-18T15:19:24  <lanius> yes, go on
2007-05-18T15:19:44  <ThomasWaldmann> Where is wiki-wide caching data saved?
2007-05-18T15:19:46  <ThomasWaldmann> * Cache Items are just normal Items with mimetype=cache and no revisioning done.
2007-05-18T15:20:11  <ThomasWaldmann> we had that discussion recently :)
2007-05-18T15:20:16  <lanius> recently ;)
2007-05-18T15:20:34  <lanius> it's not a big thing to go your way or my way, both is done easily
2007-05-18T15:21:01  <ThomasWaldmann> maybe add both ways and also a comment about doing that after the other stuff
2007-05-18T15:21:41  <ThomasWaldmann> How is edit-log stored?
2007-05-18T15:21:42  <ThomasWaldmann> * The edit log is stored in the metadata but also cached
2007-05-18T15:21:51  <ThomasWaldmann> local, global, both?
2007-05-18T15:22:24  <lanius> local
2007-05-18T15:22:31  <lanius> global is stored in the cache only
2007-05-18T15:23:14  <ThomasWaldmann> How are user profiles stored?
2007-05-18T15:23:15  <ThomasWaldmann> * Just as items.
2007-05-18T15:23:36  <ThomasWaldmann> needs more detail
2007-05-18T15:24:38  <lanius> added
2007-05-18T15:24:42  <ThomasWaldmann> Where are item-caches like the pagelinks stored?
2007-05-18T15:24:43  <ThomasWaldmann> * As metadata, handled by page.py
2007-05-18T15:25:05  <ThomasWaldmann> the pagelinks are a rather easy case and they could maybe really go into metadata
2007-05-18T15:25:25  <lanius> yes, true
2007-05-18T15:25:28  <ThomasWaldmann> but there are other caches like text_html that you won't put into metadata, but into cache system, see above
2007-05-18T15:25:33  <lanius> right
2007-05-18T15:26:36  <ThomasWaldmann> when it comes to locking, there should be a more exact terminology
2007-05-18T15:27:30  <lanius> need to talk to xorAxAx if we are clear about locking now
2007-05-18T15:27:42  <ThomasWaldmann> btw, all classes should be derived from object (directly or indirectly)
2007-05-18T15:28:04  <lanius> yes, new style classes
2007-05-18T15:28:37  <ThomasWaldmann> and well, the filters and index stuff is rather vague still
2007-05-18T15:29:00  <lanius> the definition of filters is unclear, but what else
2007-05-18T15:29:22  <ThomasWaldmann> (current implementation uses filter funcs)
2007-05-18T15:29:52  <lanius> but filtering funcs are not an option for a database backend
2007-05-18T15:30:18  <ThomasWaldmann> I know. Maybe check where it is used and how.
2007-05-18T15:30:38  <ThomasWaldmann> and please don't use "filter" as argument name as the current code does
2007-05-18T15:31:26  <ThomasWaldmann> (and your current api draft also)
2007-05-18T15:31:57  <lanius> k
2007-05-18T15:32:24  <ThomasWaldmann> (it overwrites a builtin)
2007-05-18T15:32:36  <lanius> yeah, i realized
2007-05-18T15:33:05  <ThomasWaldmann> the acl layer is still missing
2007-05-18T15:33:38  <lanius> it's no layer, it is integrated in revision/item/data/metadata internal, so it's difficult to add it to an external api
2007-05-18T15:34:47  <ThomasWaldmann> i dont see it on the wiki page
2007-05-18T15:34:54  <lanius> there is a topic ACL
2007-05-18T15:35:21  <ThomasWaldmann> i mean in the listing
2007-05-18T15:35:35  <lanius> yeah, because you can't integrate it there
2007-05-18T15:36:34  <ThomasWaldmann> well, there should be at least some comment in there about it
2007-05-18T15:36:51  <lanius> the comment is with the acl topic
2007-05-18T15:36:53  <ThomasWaldmann> and we need the user being flexible
2007-05-18T15:37:01  <lanius> yes
2007-05-18T15:37:55  <ThomasWaldmann> maybe when specifying the missing __init__ methods, stuff would get clearer
2007-05-18T15:38:23  <lanius> __init__ of itemcollection?
2007-05-18T15:38:38  <ThomasWaldmann> every place where it is missing :)
2007-05-18T15:39:48  <lanius> will do
2007-05-18T15:42:00  <ThomasWaldmann> one has to differentiate according to how long lived the stuff will be. a persistent server wants to keep the storage open all the time. but it will have changing users.
2007-05-18T15:42:45  <lanius> ?
2007-05-18T15:43:48  <ThomasWaldmann> for example, you can't give the user to the object that connects to the sql server.
2007-05-18T15:45:09  <lanius> i don't understand
2007-05-18T15:46:07  <ThomasWaldmann> user changes every few ms, db connection stays open for long
2007-05-18T15:46:54  <ThomasWaldmann> so the user must be either given with every call to some method or given to the __init__ of some short-lived "acl filtered" perspective on the storage
2007-05-18T15:47:42  <lanius> the user is given to the init of the item
2007-05-18T15:47:46  <lanius> which is short lived
2007-05-18T15:48:55  <ThomasWaldmann> ok, that solves it as far as the item is concerned
2007-05-18T15:49:29  <ThomasWaldmann> (btw, you don't want to check ACLs for the caching stuff)
2007-05-18T15:49:54  <lanius> if there is no acl metadata key no acl checking will be done
2007-05-18T15:50:28  <ThomasWaldmann> there is still acl_rights_before, default, and after
2007-05-18T15:50:57  <ThomasWaldmann> (specified in the config)
2007-05-18T15:50:59  <lanius> and they should be respected, no?
2007-05-18T15:51:19  <ThomasWaldmann> for standard items, sure
2007-05-18T15:51:27  <lanius> for the cache as well
2007-05-18T15:51:46  <lanius> otherwise you can't get things from the cache where you don't have access to the coresponding item
2007-05-18T15:52:53  <lanius> sorry, have to go and get my train, will be back in about 1 hour
2007-05-18T15:55:39  <ThomasWaldmann> ok, cu
2007-05-18T16:35:21  <xorAxAx> re
2007-05-18T16:36:45  <xorAxAx> lanius: this doesnt make sense: 15:49:54 < lanius> if there is no acl metadata key no acl checking will be done
2007-05-18T16:37:34  <xorAxAx> lanius: slides (interesting part starting at ~50% about transactions) - http://typo3.cs.uni-paderborn.de/fileadmin/Informatik/AG-Karl/Lehre/SS07/KMS/kmsSS07_04_coordination-1up-mitL.pdf
2007-05-18T17:25:41  <grzywacz> moin
2007-05-18T17:25:57  <xorAxAx> hi karol
2007-05-18T17:26:17  <grzywacz> hi alexander :-)
2007-05-18T18:10:05  <flowhase> hi everyone
2007-05-18T18:17:49  <grzywacz> hello flowhase