MoinMoin: MoinMoinChat/Logs/moin-dev/2006-07-25

2006-07-25T00:23:35  * xorAxAx implemented load_IWID
2006-07-25T00:39:32  <xorAxAx> ThomasWaldmann: the multiconfig move will annoy several users
2006-07-25T00:39:46  <xorAxAx> a) they wont notice it directly because of the pyc file b) they have to modify their config file
2006-07-25T00:52:56  <ThomasWaldmann> it's documented
2006-07-25T00:54:54  <ThomasWaldmann> and if they just install it over old stuff, there is not only the pyc, but also the old py file
2006-07-25T00:55:59  * xorAxAx pushed some changes into main because of a bug in my code that stopped moin from working :)
2006-07-25T01:08:34  <starshine> whoopsy
2006-07-25T13:13:18  * xorAxAx implemented security for cgitb
2006-07-25T17:07:06  <ThomasWaldmann> a, nice
2006-07-25T22:24:15  <Kepplar> mmm
2006-07-25T22:24:18  * Kepplar strokes his chin
2006-07-25T22:26:11  * Kepplar fires up Visio
2006-07-25T22:52:47  <Kepplar> ThomasWaldmann: Architecture
2006-07-25T22:52:48  <Kepplar> problem
2006-07-25T22:53:12  <Kepplar> Um, its about specialised wikiitem classes, still not entirely sure how to handle
2006-07-25T22:54:03  <Kepplar> effectively its like a mimetype with a "tree" structure which I probably going to impliment like the \ notation in mimetypes
2006-07-25T22:54:27  <Kepplar> (ill prod you when ive finished what im saying, i need to think)
2006-07-25T22:54:59  <Kepplar> (im ranting)
2006-07-25T22:55:18  <Kepplar> mimetypes talk from moinmoin to storage
2006-07-25T22:55:40  <Kepplar> storage to si have itemtypes
2006-07-25T22:55:56  <Kepplar> mimetypes are mapped to storage types really
2006-07-25T22:56:04  <Kepplar> although its incomparable
2006-07-25T22:56:33  <Kepplar> maybe its parsers im thinking of better
2006-07-25T22:57:15  <Kepplar> nah references wont cut it, it has to be serialisable
2006-07-25T22:57:30  <Kepplar> to allow storage implimentation to traverse it and then handle it its own special way
2006-07-25T22:57:37  <Kepplar> it cannot be tight-coupled with the item
2006-07-25T22:57:39  <Kepplar> revisiomn
2006-07-25T22:58:28  <Kepplar> should a particular itemtype not be explicatly handled by the si then the nearest match should be implimented
2006-07-25T22:58:36  <Kepplar> identical to how mimetypes are handled with its default
2006-07-25T22:59:46  <Kepplar> so pointless implimenting the datastructure twice
2006-07-25T23:01:05  <Kepplar> at the momment mimetype is used to determine itemrevsiontype
2006-07-25T23:01:27  <Kepplar> and this extra bit world be determining from itemrevsisontype, storage location, implimented uniquely for each storage layer
2006-07-25T23:01:42  <Kepplar> the question is.. is the first example bollocks?
2006-07-25T23:01:53  <Kepplar> because you can have two things of the same mimetype, but different
2006-07-25T23:02:04  <Kepplar> which goes back to the ultimate question
2006-07-25T23:02:18  <Kepplar> how does one handle user object really
2006-07-25T23:02:20  <Kepplar> ThomasWaldmann: ?
2006-07-25T23:02:49  <Kepplar> what other data is there apart from page/atachment (and its subclasses) and user?
2006-07-25T23:02:53  <Kepplar> site?
2006-07-25T23:03:53  <ThomasWaldmann> currently there is nothing else
2006-07-25T23:04:14  <ThomasWaldmann> user has a profile, bookmarks per wiki and 1 trail
2006-07-25T23:04:43  <Kepplar> hows it stored? (i know i can check and if you think its quicker to decypher it than you explaining it just say)
2006-07-25T23:05:35  <xorAxAx> ThomasWaldmann: how about delay-loaded request classes as i pointed out in the mail? we would need to bundle the demandload module (< 10 kb) but that should be ok
2006-07-25T23:05:44  <ThomasWaldmann> that mimetype to storage implementation map is just needed for 1.5, not as a general feature
2006-07-25T23:06:03  <Kepplar> um
2006-07-25T23:06:06  <Kepplar> there isnt one/
2006-07-25T23:06:09  <Kepplar> its mimetype to item type
2006-07-25T23:06:14  <Kepplar> and itemtype to itemsavinglocation
2006-07-25T23:06:43  <Kepplar> (itemsavinglocation being a path, or say an sql table or just how /where to save a partiular itemtype)
2006-07-25T23:07:25  <ThomasWaldmann> that's a feature we dont need for future storage
2006-07-25T23:07:56  <Kepplar> btw, something non SOC related, is there a plugin that does a "voting" system, aka make a "page" and every registered user gets a vote once on it?
2006-07-25T23:08:03  <Kepplar> ThomasWaldmann: how so? which bit?
2006-07-25T23:08:46  <ThomasWaldmann> Kepplar: look on the market pages. and btw: voting has many problems.
2006-07-25T23:09:44  <ThomasWaldmann> future storage can handle all item mimetypes in the same way
2006-07-25T23:09:57  <ThomasWaldmann> s/handle/store/
2006-07-25T23:10:40  <ThomasWaldmann> except maybe some minor differences like switching off revisioning where not needed
2006-07-25T23:12:02  <Kepplar> your jumping from one side to the other I think
2006-07-25T23:12:10  <Kepplar> you've missed out the fact its itemrevisions its dealing with
2006-07-25T23:13:36  <ThomasWaldmann> maybe make your problem more clear
2006-07-25T23:14:28  <Kepplar> problem is i have a visual mind and cant really put it all down in words like this
2006-07-25T23:14:39  <Kepplar> what i really need is a whiteboard and being in the same room:P
2006-07-25T23:15:19  <Kepplar> would be really good is moin can have pages where people login and they can do collaborated "whiteboard discussions" using ajax or something
2006-07-25T23:15:26  <Kepplar> and it saves the minutes
2006-07-25T23:15:36  <Kepplar> corps would love you
2006-07-25T23:16:52  <ThomasWaldmann> if you cant put it into a paragraph, you maybe also cant draw it
2006-07-25T23:17:00  <Kepplar> im sure it could
2006-07-25T23:17:06  <Kepplar> _I_ on the otherhand cant :)
2006-07-25T23:17:09  <xorAxAx> ThomasWaldmann: does no answer mean "no"?
2006-07-25T23:17:33  <Kepplar> parsers are in, formatters are out right?
2006-07-25T23:17:59  <ThomasWaldmann> no answer means no answer. and i guess i didnt read what you are talking about.
2006-07-25T23:18:36  * xorAxAx wonders why
2006-07-25T23:18:44  <xorAxAx> Kepplar: yes
2006-07-25T23:19:28  <Kepplar> formatters are locked to certain parsers@
2006-07-25T23:19:29  <Kepplar> ?
2006-07-25T23:19:37  <xorAxAx> Kepplar: no
2006-07-25T23:19:43  <xorAxAx> thats the point :)
2006-07-25T23:19:51  <Kepplar> how can your html formatter parse er pypaintytoolthing?
2006-07-25T23:20:16  <xorAxAx> Kepplar: what do you mean?
2006-07-25T23:20:23  <xorAxAx> colorise source?
2006-07-25T23:20:41  <xorAxAx> colouri[sz]e ...
2006-07-25T23:20:54  <Kepplar> i just fail to see how without a generic abstract datatype any formatter will know how to handle any random piece of information from any parser
2006-07-25T23:21:19  <xorAxAx> Kepplar: the trick is a common interface
2006-07-25T23:21:28  <xorAxAx> which is called the formatter interface
2006-07-25T23:21:44  <xorAxAx> which is rather old and seems to show its limits
2006-07-25T23:22:30  <Kepplar> i know that parsers call formatter interface
2006-07-25T23:22:56  <xorAxAx> so where is this "undefinedness"?
2006-07-25T23:22:57  <Kepplar> would be kind enough to talk me through an example (Again i know)
2006-07-25T23:23:05  <xorAxAx> that you were expecting
2006-07-25T23:23:18  <xorAxAx> of how the interface is used?
2006-07-25T23:23:20  <Kepplar> me or ThomasWaldmann?
2006-07-25T23:23:27  <xorAxAx> you
2006-07-25T23:23:57  <Kepplar> well I think its critical to know how data is coming into the storage engine in how to develop the specialised architecture inside it
2006-07-25T23:24:10  <Kepplar> At the momment there are two "specialisations" that exist in the dataengine
2006-07-25T23:24:15  <Kepplar> first, we have mimetype, nuff said
2006-07-25T23:24:21  <xorAxAx> thats another direction ... the formatter side is the output
2006-07-25T23:24:29  <xorAxAx> which is rather unrelated to the storage engine
2006-07-25T23:25:01  <Kepplar> Secondly we have item type, which at one level isjust things like page/user/attachments, but also drill down into storage types, each one being handled possibly in a different way by the actual storage implementation
2006-07-25T23:25:14  <Kepplar> atm we loosely have mimetypes mapping to item types
2006-07-25T23:25:17  <Kepplar> which is deadwrong
2006-07-25T23:25:47  <Kepplar> i think
2006-07-25T23:26:04  <xorAxAx> note that this is irrelevant for formatter interaction
2006-07-25T23:26:18  <Kepplar> maybe formatter, but not parser
2006-07-25T23:26:24  <xorAxAx> just the parser has to be chosen ... which is mainly about knowing the mimetype of the item
2006-07-25T23:26:27  <Kepplar> I think
2006-07-25T23:26:36  <Kepplar> but what "mimetype" is a user?
2006-07-25T23:26:48  <xorAxAx> a custom internal one, that was thomas idea at least
2006-07-25T23:26:54  <Kepplar> ah
2006-07-25T23:26:54  <Kepplar> ok
2006-07-25T23:26:57  <Kepplar> making more sense
2006-07-25T23:27:06  <xorAxAx> or does he wants to attach it to the homepages? no idea, ask him :)
2006-07-25T23:27:24  <xorAxAx> twiki has a namespace for users, and the metadata on the homepages builds their settings
2006-07-25T23:27:37  <Kepplar> yea, ive used twiki
2006-07-25T23:27:55  <Kepplar> i think its all a bit overhyped though
2006-07-25T23:28:24  <Kepplar> hmm
2006-07-25T23:28:36  <Kepplar> so if its amimetype there a direct map
2006-07-25T23:28:41  <Kepplar> which is good
2006-07-25T23:30:10  <ThomasWaldmann> xorAxAx: see moin-1.6 for the multiconfig problem solution
2006-07-25T23:30:28  <xorAxAx> ThomasWaldmann: yes, i didnt even expect that you fix it now that i had the idea
2006-07-25T23:30:43  <Kepplar> but, if assetaning the itemtype is soley on the mimetype, whats the point of having an item type, just keep it closely coupled with the mimetype?
2006-07-25T23:30:58  <xorAxAx> so we dont have to ask the people to change their config now
2006-07-25T23:31:20  <xorAxAx> maybe i find a possiblity to output deprecation information on usage of the old name
2006-07-25T23:32:59  <ThomasWaldmann> there is no point in delaying such stuff. it is usually fixed in a minute and there is even a workaround now in case the software installer cant edit the config.
2006-07-25T23:34:11  <xorAxAx> at least the user has a helpful message now ...
2006-07-25T23:34:12  <ThomasWaldmann> now even people who dont read docs will get it easily
2006-07-25T23:34:29  <xorAxAx> but you mean that all kind of deprecation information for such things is useless?
2006-07-25T23:34:54  <xorAxAx> normally it is considered good practice to warn the user/third party dev before changing things
2006-07-25T23:35:14  <ThomasWaldmann> there is a warning, in the docs
2006-07-25T23:35:55  <ThomasWaldmann> and I don't see me warning people a year ago just because I change some minor thing
2006-07-25T23:36:10  <xorAxAx> yeah, but that warning says that the user has to change it now. normally deprecationwarnings say "fix it before you do the next upgrade (1.7 in our case)"
2006-07-25T23:36:40  <xorAxAx> i think it doesnt make sense for major changes, and just for minor changes :)
2006-07-25T23:36:52  <ThomasWaldmann> for typing 5 chars into a file, this procedure is ridiculous
2006-07-25T23:37:00  <ThomasWaldmann> sorry, 7 :)
2006-07-25T23:37:27  <xorAxAx> its all about user experience :)
2006-07-25T23:38:10  <ThomasWaldmann> if you tell "fix it before 1.7", what do you think will happen in most cases?
2006-07-25T23:38:20  <ThomasWaldmann> it will work, so no need to fix it now.
2006-07-25T23:38:41  <ThomasWaldmann> and those who never read docs, will then have a hard time at 1.7
2006-07-25T23:38:45  <xorAxAx> people will see the deprecationwarning, edit their config and then upgrade
2006-07-25T23:38:55  <ThomasWaldmann> so you just delay the problems
2006-07-25T23:38:56  <xorAxAx> and if they complain then, _i_ dont care
2006-07-25T23:38:58  <xorAxAx> at all
2006-07-25T23:39:05  <xorAxAx> no, i just promise a seamless upgrade
2006-07-25T23:50:09  <ThomasWaldmann> Kepplar: what does assetaning mean?