MoinMoin: MoinMoinChat/Logs/moin-dev/2007-05-17

2007-05-17T00:03:59  <neagulm> xorAxAx, I will update the wiki with more ideas...
2007-05-17T00:04:14  <neagulm> dreimark, i need to go to sleep
2007-05-17T00:04:33  <dreimark> neagulm: good night!
2007-05-17T00:05:16  <neagulm> dreimark, i will be tomorrow from 8 at work and will join #moin-dev as soon as possible.
2007-05-17T00:05:46  <neagulm> good night
2007-05-17T00:30:03  <dreimark> gn all
2007-05-17T00:48:17  <grzywacz> ThomasWaldmann, does CIA report pushes to SoC branches?
2007-05-17T00:50:42  <ThomasWaldmann> iirc not (yet)
2007-05-17T00:50:50  <grzywacz> ok :)
2007-05-17T00:55:54  <ThomasWaldmann> btw, is there some xmpp based collab editor?
2007-05-17T00:57:21  <grzywacz> You mean something that a couple of people can use to edit a piece of text concurrently?
2007-05-17T00:57:50  <grzywacz> http://xmppcollaborate.wordpress.com/
2007-05-17T00:57:50  <grzywacz> ?
2007-05-17T01:02:13  <ThomasWaldmann> ah, ok
2007-05-17T01:03:43  * ThomasWaldmann was playing with some js-only client side, see my hp
2007-05-17T01:08:54  <grzywacz> xorAxAx, currently email notifications are being sent directly by the PageEditor class, and I think that moving this to a specialised notifications module is a must for at least two reasons: 1) reduce clutter, 2) make it possible to generate notifications from actions
2007-05-17T01:09:00  <grzywacz> Any comments?
2007-05-17T01:09:48  <grzywacz> Basically, the current code could stay as it is for a while, with a difference that _notifySubscribers would no longer be a member of PageEditor, so it would need an additional argument ;-)
2007-05-17T01:10:20  <grzywacz> (the same with _sendNotification)
2007-05-17T01:13:10  * ThomasWaldmann .sleep()
2007-05-17T01:13:17  <grzywacz> Good night, Thomas.
2007-05-17T01:13:41  <ThomasWaldmann> gn * :)
2007-05-17T01:15:38  <starshine> nn TW
2007-05-17T10:45:09  <xorAxAx> grzywacz: yes, you will have to move it :)
2007-05-17T13:33:48  <ThomasWaldmann> TaesuPyo is alife. :) He wrote some email that his stuff is under GPL.
2007-05-17T13:34:25  <xorAxAx> wow
2007-05-17T13:35:15  <ThomasWaldmann> yeah, took a few days X)
2007-05-17T15:38:05  <lanius> xorAxAx: ThomasWaldmann: how is locking of the edit-log currently done?
2007-05-17T15:40:09  <xorAxAx> lanius: its not done at all
2007-05-17T15:40:32  <xorAxAx> lanius: but append only usage reduces the risks of corruption
2007-05-17T15:41:46  <lanius> so the caching interface needs a file like interface as well?
2007-05-17T15:42:02  <xorAxAx> why?
2007-05-17T15:42:11  <xorAxAx> but i think that this is true, yes
2007-05-17T15:42:24  <lanius> because if you do just set/get you don't have appends
2007-05-17T15:42:57  <xorAxAx> editlogs arent used as caches
2007-05-17T15:43:06  <xorAxAx> they are a primary source of information
2007-05-17T15:43:10  <lanius> we decided to that
2007-05-17T15:43:13  <lanius> to do that
2007-05-17T15:43:17  <xorAxAx> yes
2007-05-17T15:43:28  <xorAxAx> well
2007-05-17T15:43:43  <xorAxAx> you dont want to write 100 MB of global editlog when editing a page, yes
2007-05-17T15:45:32  <lanius> so in the end the cache could just be items with mimetype = cache?
2007-05-17T15:46:23  <xorAxAx> yes, but without revisions
2007-05-17T15:46:31  <lanius> or only one revision
2007-05-17T15:47:01  <xorAxAx> yes
2007-05-17T15:47:10  <xorAxAx> i meant to say "without revisioning"
2007-05-17T15:47:13  <xorAxAx> :-)
2007-05-17T15:50:04  <lanius> i don't think we can make data immutable because an admin even wants do edit that, so we need a global locking strategy
2007-05-17T15:51:22  <xorAxAx> well, currently admins edit it as well but it is considered immutable
2007-05-17T16:00:18  <lanius> but if you implement locking you can't just considere it to be immutable
2007-05-17T16:04:22  <xorAxAx> well, it has been working for years :)
2007-05-17T16:10:43  <lanius> what about the following
2007-05-17T16:11:01  <lanius> you can either get an item for reading or for writing from the itemcollectino
2007-05-17T16:11:19  <lanius> if you get it for writing a lock will be set, so no other can get it for writing
2007-05-17T16:11:34  <lanius> when saving the lock will be released
2007-05-17T16:12:43  <xorAxAx> this works as long as you cannot delete or change revisions
2007-05-17T16:12:50  <xorAxAx> otherwise you need read locks as well
2007-05-17T16:13:05  <lanius> why
2007-05-17T16:13:47  <xorAxAx> because thread A might start reading e.g. the metadata, then B removes that revisions or changes the mimetype, then A continues reading the data incorrectly assuming the mimetype
2007-05-17T16:14:47  <lanius> ok
2007-05-17T16:15:07  <lanius> but can't we make the metadata just "immutable" as well then?
2007-05-17T16:15:11  <lanius> if it works for the data
2007-05-17T16:17:09  <lanius> b
2007-05-17T16:29:53  <xorAxAx> lanius: the metadata is often changed
2007-05-17T16:30:25  <xorAxAx> e.g. with wikisync, every page stores some information about remote wikis that synced to this page
2007-05-17T16:34:00  <lanius> expain further
2007-05-17T16:34:26  <lanius> explain
2007-05-17T16:35:29  <xorAxAx> well, in order to synchronise and merge remote wiki pages, the wiki has to know which remote revision corresponds to which local revision
2007-05-17T16:35:55  <xorAxAx> so after every wikisync job run, the list of local revision info is extended/changed
2007-05-17T16:36:22  <xorAxAx> thats just one example, i enumerated others a few days ago :)
2007-05-17T16:38:47  <lanius> list of local revision info?
2007-05-17T16:39:27  <xorAxAx> umm, remote revision info
2007-05-17T16:39:34  <xorAxAx> thats all described on the wikisync docs pages
2007-05-17T16:40:08  <ThomasWaldmann> btw, i dont think we should just put caching stuff into backend storage with just using mimetype=cache.
2007-05-17T16:40:52  <ThomasWaldmann> there should be some separate space for that stuff.
2007-05-17T16:41:07  <lanius> xorAxAx: i still don't understand
2007-05-17T16:41:09  <lanius> ThomasWaldmann: why
2007-05-17T16:41:58  <ThomasWaldmann> how would you store the text_html cache for Page XY?
2007-05-17T16:42:53  <lanius> name something like that: cache/text_html/page xy and the content in revision 1
2007-05-17T16:43:44  <ThomasWaldmann> ok, that means you use some specific part of the namespace
2007-05-17T16:43:57  <lanius> yes, just like the users have a specific part of the namespace
2007-05-17T16:44:49  <ThomasWaldmann> when you create a list of all items you will have to filter out all those items
2007-05-17T16:44:55  <ThomasWaldmann> cache items
2007-05-17T16:45:09  <lanius> list of all pages
2007-05-17T16:45:26  <ThomasWaldmann> so the question is whether we should them put into same namespace at all
2007-05-17T16:45:36  <lanius> sure, that's why list_items() has a filter
2007-05-17T16:45:36  <ThomasWaldmann> or just use some separate namespace
2007-05-17T16:46:20  <lanius> that wouldn't help since the namespaces are transparent to the layer above
2007-05-17T16:48:10  <ThomasWaldmann> the point is if you have 5000 pages, and 5 caches per page, you will have a list of 30.000 items just to throw away 25.000 of them before showing
2007-05-17T16:48:35  <lanius> yes, but that should be fast
2007-05-17T16:49:04  <xorAxAx> lanius: have you read the wikipage about wikisync internals?
2007-05-17T16:49:17  <lanius> http://moinmoin.wikiwikiweb.de/WikiSyncronisation this one?
2007-05-17T16:49:26  <ThomasWaldmann> how would it be fast?
2007-05-17T16:49:30  <lanius> indexes
2007-05-17T16:49:44  <ThomasWaldmann> -v
2007-05-17T16:49:58  <xorAxAx> lanius: yes
2007-05-17T16:50:03  <lanius> xorAxAx: i did
2007-05-17T16:51:01  <lanius> xorAxAx: i don't see the point where any metadata needs to be changed
2007-05-17T16:51:11  <xorAxAx> lanius: well
2007-05-17T16:51:12  <lanius> xorAxAx: what would be different from data changes
2007-05-17T16:51:27  <xorAxAx> lanius: as i said - the wiki needs to know which revisions match each other
2007-05-17T16:51:45  <lanius> yes, compare name and revision no?
2007-05-17T16:52:04  <xorAxAx> now the question is - are tags added without a rev being added
2007-05-17T16:52:19  <lanius> tags?
2007-05-17T16:52:21  <xorAxAx> and the answer is "yes, if the merge resulted in the same page contents"
2007-05-17T16:52:28  <xorAxAx> lanius: i am talking about wiki sync tags
2007-05-17T16:52:32  <xorAxAx> as described on the above page
2007-05-17T16:52:59  <lanius> "They are tuples associated to a page" whatever that means
2007-05-17T16:53:42  <xorAxAx> (IWID, remote_rev)*
2007-05-17T16:53:44  <xorAxAx> basically
2007-05-17T16:53:53  <xorAxAx> umm, and remote_pagename
2007-05-17T16:53:58  <lanius> and where is that stored
2007-05-17T16:54:04  <xorAxAx> in the page directory
2007-05-17T16:55:42  <lanius> and why is that changed on every sync
2007-05-17T16:56:50  <xorAxAx> hmm, given the assumptions that remote pages only change their contents iff the rev number changes, it might make sense to model this as immutable per-rev  metadata
2007-05-17T16:57:04  <xorAxAx> but then there are a few exceptions:
2007-05-17T16:57:23  <xorAxAx> initial syncs from a wiki dont necessarily change the page on one side if it was not available on the other
2007-05-17T16:57:38  <xorAxAx> so you would need to modify the metadata on your side without creating a new revision
2007-05-17T16:58:05  <lanius> you always do a sync from remote to local? or both sides?
2007-05-17T16:59:01  <lanius> ok, both sides
2007-05-17T16:59:12  <lanius> but wouldn't it make sense to lock the whole wiki during synchronisation
2007-05-17T16:59:15  <xorAxAx> yes, both sides, both sides store equivalent information
2007-05-17T16:59:21  <xorAxAx> yes, that would make sense
2007-05-17T16:59:30  <xorAxAx> or at least the single page
2007-05-17T16:59:38  <xorAxAx> but a wiki sync might need a minute
2007-05-17T16:59:38  <lanius> yes
2007-05-17T16:59:46  <xorAxAx> so you dont want to lock too much :)
2007-05-17T17:00:05  <xorAxAx> (because of latency - the code is doing a few requests internally)
2007-05-17T17:00:05  <lanius> so lock the currently syncing page
2007-05-17T17:01:25  <lanius> so locking of whole items which prevents saving until you fetch the item again should be enough
2007-05-17T17:01:41  <xorAxAx> yes
2007-05-17T17:01:48  <lanius> that sounds not too difficult
2007-05-17T17:02:06  <lanius> any other locking situation?
2007-05-17T17:02:22  <xorAxAx> caching files are regularly modified
2007-05-17T17:02:31  <xorAxAx> user files are modified as well
2007-05-17T17:03:53  <lanius> lock the whole item again
2007-05-17T17:04:33  <lanius> i would add three methods to the backend: lock_item, unlock_item, item_is_locked
2007-05-17T17:04:50  <ThomasWaldmann> lanius: for the caching files at least, I want to have the possibility to do a similar backend as the code that is used now
2007-05-17T17:05:17  <ThomasWaldmann> and it doesnt need locking for posix
2007-05-17T17:05:39  <lanius> yes, we are talking about a different locking layer
2007-05-17T17:06:16  <lanius> ThomasWaldmann: since it is a cache the backend doesn't matter and since it must be distributed in some cases it can't be the same
2007-05-17T17:06:32  <xorAxAx> lanius: how would be item_is_locked useful? :)
2007-05-17T17:06:46  <lanius> item_collection must check to know if it returns a mutable or immutable item
2007-05-17T17:07:27  <ThomasWaldmann> lanius: you talked about just put caches into normal items, so why doesnt it matter?
2007-05-17T17:07:59  <lanius> ThomasWaldmann: what do you mean by similar backend?
2007-05-17T17:09:07  <ThomasWaldmann> caching.py uses 2 locking methods, a slow one doing real locking for non-posix (e.g. win32) and a fast one doing nothing for posix
2007-05-17T17:09:21  <ThomasWaldmann> and speed is critical for caching
2007-05-17T17:09:38  <lanius> that locking stuff is on the backend layer
2007-05-17T17:09:49  <lanius> in the backend layer even
2007-05-17T17:09:56  <lanius> data-corruption locking
2007-05-17T17:10:19  <lanius> we are talking about locking so the following is not possible
2007-05-17T17:10:33  <lanius> a reads the page; b reads the page; a changes and writes; b changes and overwrites the changes of a
2007-05-17T17:13:44  <ThomasWaldmann> for some caches, this is not needed
2007-05-17T17:13:48  <xorAxAx> the transactional locking system could use the same primitives internally, though
2007-05-17T17:14:02  * xorAxAx is gone jogging
2007-05-17T17:14:10  <ThomasWaldmann> so will you do it just because the cache is an item?
2007-05-17T17:15:12  <lanius> no, because those locks must be handled by the page level where a cache item must be as well
2007-05-17T17:15:25  <lanius> xorAxAx: have fun
2007-05-17T17:16:27  <lanius> ThomasWaldmann: they are supported by the backend, but if they are used is decided by a layer above
2007-05-17T17:16:38  <ThomasWaldmann> ok, so back to polluting the namespace :) how is it going to be so fast it does not matter?
2007-05-17T17:17:02  <lanius> the backend has to make it fast with indexes, the mime-type is used as an index
2007-05-17T17:17:12  <lanius> and access to indexes is fast per definition
2007-05-17T17:18:38  <ThomasWaldmann> somehow that sounds more like a huge workaround around a problem that we should not create in the first place
2007-05-17T17:19:07  <lanius> how else can we generalize the backend for caches/users/pages/attachements at all
2007-05-17T17:20:11  <ThomasWaldmann> maybe by sharing most of the code, but using different namespaces (at least for caching stuff, maybe for user profiles)
2007-05-17T17:20:24  <lanius> but who handles the namesapces then
2007-05-17T17:21:42  <ThomasWaldmann> you usually know what you want, so you can say "give me item WikiSandBox/text_html from the caching namespace"
2007-05-17T17:22:32  <lanius> what if you have two namespaces for load sharing
2007-05-17T17:22:51  <lanius> s/sharing/balancing/
2007-05-17T17:23:03  <ThomasWaldmann> so the namespace is an additional argument
2007-05-17T17:23:29  <lanius> to what
2007-05-17T17:23:42  <lanius> and how do you know if your item is in the first or in the second one
2007-05-17T17:24:34  <ThomasWaldmann> how do you load balance?
2007-05-17T17:25:43  <lanius> and you can even do a own namesapce for cache/ with the current design
2007-05-17T17:26:41  <lanius> but what you can not do is say that namespace1 only contains cache items
2007-05-17T17:28:25  <ThomasWaldmann> currently you have backend = ... in your config as the only thing doing any data providing
2007-05-17T17:28:46  <ThomasWaldmann> there could be just some caching = ... parallel to that for the caching
2007-05-17T17:29:04  <lanius> and then some users= and some attachements= and whatevery will come?
2007-05-17T17:29:49  <ThomasWaldmann> no, because there are no attachments
2007-05-17T17:30:02  <lanius> and there are no users and there is no cache
2007-05-17T17:30:04  <ThomasWaldmann> (just items :)
2007-05-17T17:30:38  <ThomasWaldmann> the point is mixing what should be mixed and not mixing what should be kept seperate
2007-05-17T17:30:55  <lanius> it's storing of data
2007-05-17T17:31:38  <ThomasWaldmann> and i see no point in mixing tons of internal caching things into the wiki items
2007-05-17T17:32:30  <lanius> the storage layer does not care
2007-05-17T17:33:20  <ThomasWaldmann> but the user will, if the speed sucks :)
2007-05-17T17:34:07  <lanius> but speed won't suck, i don't can promise all for a file backend, but a file backend is not for performance, it is for ease of use, a database backend won't suck in performance
2007-05-17T17:34:10  <ThomasWaldmann> or it eats too much memory because all item lists are 10 times bigger than needed
2007-05-17T17:34:21  <lanius> why 10 times bigger`?
2007-05-17T17:34:55  <ThomasWaldmann> because you might have 9 caching things per real item in them
2007-05-17T17:35:07  <lanius> but then they are not stored in memory
2007-05-17T17:35:16  <lanius> only the item what is used is stored in memory
2007-05-17T17:36:10  <ThomasWaldmann> if i request a list of all items, it will be 10x bigger
2007-05-17T17:36:17  <lanius> then don't do that
2007-05-17T17:37:00  <ThomasWaldmann> ok, so how do i get all INTERESTING items? :)
2007-05-17T17:37:17  <ThomasWaldmann> (from the wiki user perspective)
2007-05-17T17:37:30  <lanius> a filter with mimetype=text/wiki
2007-05-17T17:37:34  <ThomasWaldmann> no
2007-05-17T17:37:40  <lanius> why not
2007-05-17T17:37:57  <ThomasWaldmann> because that are only pages of some specific markup
2007-05-17T17:38:10  <lanius> then a filter with mimetype!=cache
2007-05-17T17:38:15  <ThomasWaldmann> but the user is interested in other markups, too
2007-05-17T17:38:47  <ThomasWaldmann> then i still will have user profiles in it
2007-05-17T17:39:10  <lanius> then exclue user profiles as well
2007-05-17T17:40:02  <lanius> or we have to add another metadata key, mimetype and itemtype
2007-05-17T17:40:23  <lanius> where itemtype is wiki or user or cache and mimetype is then text/wiki or text/python or..
2007-05-17T17:41:04  <ThomasWaldmann> ok, assuming that i do that filters, how will it fetch the stuff that i want then?
2007-05-17T17:41:17  <lanius> who is you
2007-05-17T17:42:31  * ThomasWaldmann is /me
2007-05-17T17:42:46  <lanius> who is it
2007-05-17T17:43:24  <ThomasWaldmann> it = the system we are talking about X(
2007-05-17T17:43:30  <lanius> but what layer
2007-05-17T17:43:57  <lanius> what part of the system
2007-05-17T17:44:08  <ThomasWaldmann> ItemCollection.keys I guess
2007-05-17T17:44:54  <ThomasWaldmann> (following YOUR idea :)
2007-05-17T17:45:48  <lanius> ItemCollection.keys has the filters= argument which is passed to all backends
2007-05-17T17:45:58  <lanius> and the backends only return the according names
2007-05-17T17:46:05  <lanius> s/according/matching/
2007-05-17T17:46:49  <ThomasWaldmann> ok, go on
2007-05-17T17:47:25  <lanius> a db backend just does a fast query "select * from db where mimetype = bla"
2007-05-17T17:47:47  <ThomasWaldmann> it does not know bla
2007-05-17T17:47:47  <lanius> a file backend would have a file for every index where the value for each page and this index is written to
2007-05-17T17:47:58  <lanius> it does since this is specified in the filter
2007-05-17T17:48:11  <ThomasWaldmann> not, the filter only says what not
2007-05-17T17:48:28  <lanius> then it does a query "select * from db where mimetype != bla"
2007-05-17T17:48:44  <lanius> no magic
2007-05-17T17:49:14  <ThomasWaldmann> btw, how do i specify a filter for mimetype!=cache?
2007-05-17T17:49:30  <lanius> not yet specified
2007-05-17T17:50:50  <ThomasWaldmann> ok, assuming we have some spec for that soon, what does the file backend do then?
2007-05-17T17:51:21  <lanius> it opens the index file which is essentially a key,value pair, like Page1=cache\nPage2=wiki/text
2007-05-17T17:51:30  <lanius> and then checks every entry
2007-05-17T17:51:38  <lanius> what the db does internally
2007-05-17T17:52:39  <ThomasWaldmann> ok, that means when it runs into the cache/* part of that all-in-one namespace, it will check 25000 entries of the index and none of them will match.
2007-05-17T17:53:06  <lanius> yes
2007-05-17T17:53:18  <ThomasWaldmann> but we knew that before :)
2007-05-17T17:53:40  <ThomasWaldmann> so why do we do that?
2007-05-17T17:53:59  <lanius> because it's simple to implement
2007-05-17T17:54:05  <lanius> and very general
2007-05-17T17:54:14  <lanius> and since the list_items is not used very often
2007-05-17T17:55:06  <lanius> another storage method would be the way around
2007-05-17T17:55:22  <lanius> cache:Page1;Page2\nwiki/text:Page3;Page4;Page5
2007-05-17T17:55:50  <ThomasWaldmann> $ grep -r getPageList *|less|wc -l
2007-05-17T17:55:53  <ThomasWaldmann> 62
2007-05-17T17:56:06  <lanius> then this is very unoptimized
2007-05-17T17:56:23  <lanius> performance is something databases do, not filesystems
2007-05-17T17:56:49  <ThomasWaldmann> we have to support both in a fast way
2007-05-17T17:56:57  <lanius> databases are alwys faster
2007-05-17T17:57:04  <ThomasWaldmann> no they arent
2007-05-17T17:57:08  <starshine> um no
2007-05-17T17:57:11  <lanius> for a lot of items
2007-05-17T17:57:22  <starshine> believe me db can get themselves tied in a knot
2007-05-17T17:57:30  <ThomasWaldmann> for doing silly things they maybe are faster :)
2007-05-17T17:57:51  <starshine> also the complexity of setup for a small one and difficulty in mgmt outweighs the speed gain of access at a small scale
2007-05-17T17:58:17  <starshine> mysql isam is only faster because it is deliberately gutted of some aspects that would be good for stability
2007-05-17T17:58:30  <starshine> unstable large databases -> dooooom
2007-05-17T17:59:13  <ThomasWaldmann> lanius: please do that grep yourself and look at those places
2007-05-17T17:59:14  <lanius> my last proposal for the indexes reduces the comparison from 25000 to 2
2007-05-17T17:59:15  <starshine> however people like having choices "D
2007-05-17T17:59:25  <starshine> :D
2007-05-17T17:59:32  <ThomasWaldmann> then also please analyse where caching module is used
2007-05-17T17:59:51  <ThomasWaldmann> before continuing on that "the db will rescue us" road
2007-05-17T17:59:56  <lanius> it doesn't make sense to look at the current code and design the backend for that
2007-05-17T18:00:11  <lanius> this way you never get a good design
2007-05-17T18:02:44  <ThomasWaldmann> btw, I currenlty can't imagine the moin code doing some query like "give me all item names, including page-like items, user-like-items and cache-like items"
2007-05-17T18:03:01  <lanius> and i just reduced the comparison from 25000 to 2, which shoudl be fast enough
2007-05-17T18:03:10  <ThomasWaldmann> but ever only one category of them
2007-05-17T18:03:31  <lanius> 4 if you wan't two categories
2007-05-17T18:03:59  <lanius> you don't wan't attachements in the page list, do you?
2007-05-17T18:04:06  <lanius> so you always have this problem
2007-05-17T18:04:41  <ThomasWaldmann> no, what you call attachments is content, like wiki pages
2007-05-17T18:04:58  <lanius> so you want them in the page list?
2007-05-17T18:04:59  <ThomasWaldmann> and that can well be wanted or not wanted
2007-05-17T18:05:07  <lanius> hough
2007-05-17T18:05:10  <ThomasWaldmann> depends on what you do
2007-05-17T18:05:30  <lanius> then what about my idea to split it into mimetypes and itemtypes
2007-05-17T18:06:12  <lanius> because you have the same with users and what else you wan't to store in the storage what will later be added and so a eneral solution is the best
2007-05-17T18:11:47  * lanius is getting a pizza
2007-05-17T18:14:18  <ThomasWaldmann> my main point is not to do unnecessary things. esp. not if they might have big consequences.
2007-05-17T18:14:56  <ThomasWaldmann> so what is the problem of having data = ... and caching = ... in your config?
2007-05-17T18:16:15  <ThomasWaldmann> (this will still be far better than what we have now, we will still have more flexibilty using the storage api)
2007-05-17T18:17:10  <ThomasWaldmann> we just won't query the caching storage for content items, nor the content storage for caching items
2007-05-17T18:19:15  <ThomasWaldmann> (and we can decide later whether we want to do the same thing for the users or not, that is somewhat a border case, depending how you implement user profiles)
2007-05-17T18:22:21  <ThomasWaldmann> btw, even if you would not do it that way, I guess about everybody would "mount" some separate storage backend to the cache/ part of the all-in-one namespace, and if it is only for backup reasons
2007-05-17T18:27:36  <flowhase> moin here
2007-05-17T18:28:08  * flowhase takes a step into a brighter operating system future right now
2007-05-17T18:28:35  * flowhase switches to FreeBSD
2007-05-17T18:30:38  <flowhase> well, at least i hope it will be brighter :>
2007-05-17T18:32:05  <xorAxAx> hehe, the operating system with a closed membership core developers mailing list
2007-05-17T18:33:31  <flowhase> well, i don't really care about that ... i never read LKML
2007-05-17T18:33:58  <flowhase> but i am quite impressed with *BSD documentation
2007-05-17T18:34:38  <xorAxAx> well, freebsd core is much more than LKML
2007-05-17T18:34:41  <xorAxAx> they have a userland
2007-05-17T18:35:02  <flowhase> yep
2007-05-17T18:35:04  <xorAxAx> and they are discussing distribution and release issues
2007-05-17T18:35:30  <xorAxAx> i think freebsd might be nice for niches
2007-05-17T18:35:37  <xorAxAx> but not as a general desktop OS :)
2007-05-17T18:35:47  <flowhase> i object :>
2007-05-17T18:35:59  <xorAxAx> mainly for driver reasons
2007-05-17T18:36:05  <flowhase> i used it for desktop the last few days at the entropia workstation
2007-05-17T18:36:32  <xorAxAx> well, if you dont have much (multimedia) devices, you wont care
2007-05-17T18:36:59  <flowhase> i have mostly standard hardware
2007-05-17T18:37:24  <xorAxAx> i had been playing with freebds for a few months and tried to install it onto my 24/7 utility machine that is running in my living room
2007-05-17T18:37:31  <flowhase> nvidia-graphics, emu10k1 sound ...
2007-05-17T18:37:33  <xorAxAx> but it failed because of pcmcia support issues
2007-05-17T18:37:38  <flowhase> oh
2007-05-17T18:37:46  <xorAxAx> flowhase: i am thinking about stuff like dvb-t support :)
2007-05-17T18:37:54  <flowhase> i see :>
2007-05-17T18:39:36  <ThomasWaldmann> flowhase: which browsers do you have under freesbd, how will you test the IE compat?
2007-05-17T18:40:50  <xorAxAx> IE works nicely on wine
2007-05-17T18:41:10  <xorAxAx> so its not much different from linux :)
2007-05-17T18:44:20  <ThomasWaldmann> flowhase: which python do you have there?
2007-05-17T18:46:36  <ThomasWaldmann> http://linuxwiki.de/LugLudwigsburg < LOL?
2007-05-17T18:46:40  <ThomasWaldmann> http://linuxwiki.de/LugLudwigsburg < LOL!
2007-05-17T18:46:48  <xorAxAx> ThomasWaldmann: anything between 2.3 and 2.5
2007-05-17T18:47:05  <xorAxAx> ThomasWaldmann: will you arrive on monday in ludwigsburg?
2007-05-17T19:06:04  <flowhase> ThomasWaldmann: the IE thingy will be done from vmware i think ;>
2007-05-17T19:08:13  <ThomasWaldmann> vmware works under freebsd?
2007-05-17T19:09:05  <ThomasWaldmann> in any case, make sure that a productive development environment works before coding starts :)
2007-05-17T19:09:39  <xorAxAx> no, it doesnt
2007-05-17T19:09:49  <xorAxAx> yes, e.g. install moin locally
2007-05-17T19:09:52  <xorAxAx> keppy didnt :)
2007-05-17T19:10:36  <ThomasWaldmann> it is easier now with 1.6/1.7 :)
2007-05-17T19:11:40  <flowhase> i already have several local moins thrown all over the boxes i have accounts on :>
2007-05-17T19:11:45  <xorAxAx> hehe
2007-05-17T19:11:55  <flowhase> it's because i often work on different machines
2007-05-17T19:11:58  <xorAxAx> so you are a few leaps further than keppy
2007-05-17T19:12:09  <flowhase> *cough*
2007-05-17T19:12:20  <xorAxAx> :-)
2007-05-17T19:12:54  <flowhase> btw.
2007-05-17T19:13:46  <flowhase> are there any differences (regarding what i intend to work on) between real moin-installs and moin desktop edition?
2007-05-17T19:14:05  <xorAxAx> not at all
2007-05-17T20:42:41  <grzywacz> flowhase, freebsd? *haha* :P
2007-05-17T20:48:16  <dreimark> moin
2007-05-17T20:49:16  <grzywacz> hello dreimark
2007-05-17T20:49:32  <dreimark> hi grzywacz
2007-05-17T20:52:36  * grzywacz waves to his mentor
2007-05-17T20:54:49  <neagulm> helo dreimark
2007-05-17T20:58:33  * xorAxAx waves to his student
2007-05-17T21:01:02  <grzywacz> :)
2007-05-17T21:02:00  <neagulm> bbl
2007-05-17T21:38:51  * ThomasWaldmann submitted some pydev bugs
2007-05-17T22:31:59  <dreimark> neagulm: hi
2007-05-17T22:33:44  <neagulm> hello dreimark
2007-05-17T22:34:36  <dreimark> neagulm: did you thought a bit mor about skipm's plans ?
2007-05-17T22:34:43  <neagulm> dreimark, have you read JeanPhilippeGuérard comment ?
2007-05-17T22:34:50  <neagulm> dreimark, yes
2007-05-17T22:36:28  <neagulm> dreimark, What data do you think we could use to help SB detect spam (IP, username, hostname, presence of attachments, etc) ?
2007-05-17T22:36:37  <dreimark> neagulm: yes, that will make it difficult
2007-05-17T22:37:25  <neagulm> dreimark, Skip's approach is fine with me but care must be taken in order to be compatible with other classifiers.
2007-05-17T22:38:46  <dreimark> neagulm: as they often do attack many sites at once we should incorparete the Badcontent list too, so web adresses are blocked too
2007-05-17T22:38:59  <neagulm> dreimark, as I said before: I think the hole system should be modular, generic and, maybe, as simple as possible
2007-05-17T22:39:34  <dreimark> neagulm: my opinion too :)
2007-05-17T22:42:41  <neagulm> dreimark, brb ~5min
2007-05-17T22:52:03  <dreimark> neagulm: are others classifiers already standardized in their communication ?
2007-05-17T22:59:00  <neagulm> dreimark, back
2007-05-17T22:59:45  <neagulm> dreimark, I don't think that there is a standard related to communication
2007-05-17T23:00:13  <neagulm> dreimark, no informal standard one and no official one :(
2007-05-17T23:01:20  <neagulm> dreimark, but all share the same characteristics: learning, classifying and forget
2007-05-17T23:02:39  <neagulm> dreimark, almost all parse the input into tokens, select the relevant tokens and construct a feature vector for usage in the classifier
2007-05-17T23:02:53  <neagulm> dreimark, the relevant ones for use are: classify/learn/forget
2007-05-17T23:04:29  <dreimark> neagulm: do you have contact to one of the developers of an other categorizer ?
2007-05-17T23:05:10  <neagulm> dreimark, Yes, Bitdefender
2007-05-17T23:07:02  <dreimark> is he interested in getting a simple standard too?
2007-05-17T23:11:42  <neagulm> The bitdef boys are using their classifier only in the mail servers (they have a milter for sendmail and alternate ques for qmail) and in outlook. I don't think they are interested in a specific interface
2007-05-17T23:22:45  <dreimark> neagulm: before skipm gots the idea to use it for a wiki too it was the same with sb
2007-05-17T23:31:13  <neagulm> dreimark, i'm going to sleep... I will check tomorow with the bitdefender friends
2007-05-17T23:31:29  <dreimark> neagulm: fine
2007-05-17T23:31:49  <dreimark> neagulm: good night!
2007-05-17T23:33:37  <neagulm> dreimark, a last question: can you think about a timeline related to my project (in your opinion). What will I do when we finish integrating sb: Start working for text classification or working in other parts of MM
2007-05-17T23:34:41  <neagulm> dreimark, it would be nice to know because I would need something to think about, in the long run... :)
2007-05-17T23:34:55  <dreimark> yeah
2007-05-17T23:34:57  <neagulm> dreimark, i'm leaving know
2007-05-17T23:35:05  <dreimark> ok, cu and good night
2007-05-17T23:35:14  <neagulm> good night