2006-03-25T10:37:41 *** birkenfeld has joined #pocoo 2006-03-25T10:39:12 moin birkenfeld 2006-03-25T10:39:43 moin 2006-03-25T10:45:07 birkenfeld: was ist pocoo.pkg.pages.core? 2006-03-25T10:45:23 mitsuhiko: momentan nur die IndexPage 2006-03-25T10:45:39 evtl. kann man das später in pocoo.pkg.core verschieben, dann hat man nur eine bltin package 2006-03-25T10:45:57 jup, weil pocoo.pkg.core ist momentan relativ leer (leeres init file) 2006-03-25T10:46:15 hm? 2006-03-25T10:46:25 ich hab jetzt nÃ¤mlich pocoo.pkg.tests fÃ¼r interaktive tests angelegt, die man nicht mit unittests Ã¼berprÃ¼fen kann/will 2006-03-25T10:46:49 du meinst pocoo.pkg.pages? 2006-03-25T10:46:51 birkenfeld: argh. nicht pocoo.pkg.core sondern pocoo.pkg.pages 2006-03-25T10:51:29 so, verschoben 2006-03-25T10:53:18 birkenfeld: wie findet der momentan die builtin pages? 2006-03-25T10:53:48 mitsuhiko: ich hab BUILTIN_PACKAGES rausgenommen, weil die alle in pocoo.pkg.core liegen sollten 2006-03-25T10:53:57 die tests sollen ja nicht immer geladen werden 2006-03-25T10:54:01 stimmt 2006-03-25T10:54:16 dann lad ich die, wenn debug true ist 2006-03-25T10:54:30 du brauchst noch ein __init__.py 2006-03-25T10:54:43 birkenfeld: wo? 2006-03-25T10:54:51 in tests/ 2006-03-25T10:55:05 argh. svn add vergessen 2006-03-25T10:55:16 ah, wunder mich schon wo ITestProvider herkommt ;) 2006-03-25T10:55:27 so. jetzt issts oben :) 2006-03-25T10:57:53 ich frag mich immer noch warum am anfang alles zweimal geladen wird 2006-03-25T10:58:13 jup. ich auch. ich bin noch am forschen :) 2006-03-25T10:59:37 du hattest doch gesagt, es liegt nicht am reloader= 2006-03-25T11:00:13 ich habs hier nochmal getestet 2006-03-25T11:00:16 wenn der aus ist gehts 2006-03-25T11:00:20 aber damals nicht? hm 2006-03-25T11:00:22 bei mir auch gerade 2006-03-25T11:01:46 ich werd mir den reloader nochmal genauer ansehen 2006-03-25T11:04:21 ach ja, ist klar 2006-03-25T11:04:38 erst wird pocoo importiert, dabei alles geladen 2006-03-25T11:04:52 dann der reloader gestartet, der den prozess nochmal startet 2006-03-25T11:05:39 argh. stimmt 2006-03-25T11:06:12 ich teste mal was 2006-03-25T11:33:21 mitsuhiko: einverstanden damit, dass der reloader jetzt via "manage.py runserver --no-reloader" abgeschaltet wird? 2006-03-25T11:33:42 ok :) 2006-03-25T11:33:50 keine einwende meinerseits 2006-03-25T11:40:22 ok 2006-03-25T11:56:53 mitsuhiko: ist eigentlich pocoo.utils.cli auf dem neuesten stand? 2006-03-25T11:57:18 birkenfeld: jep 2006-03-25T11:57:32 weil geht bei mir nicht -.- 2006-03-25T11:57:39 birkenfeld: welcher fehler? 2006-03-25T11:58:17 Request object has no attribute "write" 2006-03-25T11:59:18 hm. also eine write methode gibts ja seit der trennugn von request/response nicht mehr 2006-03-25T11:59:26 ich pass da mal den docstring an 2006-03-25T11:59:36 ohne write gehts aber auch nicht 2006-03-25T11:59:56 ich Ã¼berarbeite das mal 2006-03-25T12:11:51 done 2006-03-25T12:22:51 exec_request gibts aber nicht mehr... 2006-03-25T12:23:08 birkenfeld: das heiÃŸt jetzt execure_response 2006-03-25T12:23:16 weil am request objekt klebt ja jetzt nix mehr drauf 2006-03-25T12:23:38 mitsuhiko: aber in create_request wird das noch zugewiesen 2006-03-25T12:23:58 oh. das hab ich Ã¼bersehen. weg damit 2006-03-25T12:24:18 und weg isses 2006-03-25T12:40:29 hui 2006-03-25T12:40:35 * mitsuhiko hat sich gerade phpbb3 installiert 2006-03-25T12:40:40 und? 2006-03-25T12:40:50 noja. der code ist nach wie vor unbenuztbar 2006-03-25T12:40:55 hehe 2006-03-25T12:40:58 neue features? 2006-03-25T12:41:10 subforen und jabber benachrichtigungen 2006-03-25T12:41:15 und so einen bot support 2006-03-25T12:41:22 was das ist schau ich gerade 2006-03-25T12:41:40 editbots? 2006-03-25T12:41:52 ney. er erkennt jetzt suchbots 2006-03-25T12:42:06 ich glaub die bekommen irgendwie andere seiten als normale user 2006-03-25T12:42:36 jup. man kann den bots eigene themes und sprachen spendieren. 2006-03-25T12:42:53 ahso 2006-03-25T12:43:26 aber das admin interface ist wirklich gut geworden. sehr aufgerÃ¤umt 2006-03-25T12:43:55 viel arbeit? 2006-03-25T12:44:19 phpbb3 hat momentan 90.000 zeilen programmcode. das ist verhÃ¤ltnismÃ¤ÃŸig wenig 2006-03-25T12:44:34 aber das admin interface afaik 30% davon 2006-03-25T12:44:52 * mitsuhiko wird einfach mal mit den GNOME leuten reden, die haben ahnung von usability *scnr* :) 2006-03-25T12:45:19 aber mein plan fÃ¼rs admin interface ist ja gar kein admin interface :) 2006-03-25T12:45:49 wÃ¤re viel cooler, wenn man einfach auf das element doppelklickt und "inline" bearbeitet 2006-03-25T12:46:05 beispielsweise doppelklick auf den forentitel, es erscheint eine textbox, man gibt text ein und drÃ¼ckt enter 2006-03-25T12:48:27 hm. browser? 2006-03-25T12:49:00 das dynamische fÃ¼r alle ajax browser (ie4.0 oder hÃ¶her) und dann noch ein CLI tool fÃ¼r ernsthafte server admins :) 2006-03-25T12:49:24 und alle anderen beißen ins gras? 2006-03-25T12:50:03 hm 2006-03-25T12:50:28 man mÃ¼sste irgendwie eine abstraktionsschicht machen, dass er immer eine ajax und eine statische lÃ¶sung herbeizaubert 2006-03-25T12:55:32 aber unÃ¼bersichtlich ist es das phpBB3. (nicht das admin panel) 2006-03-25T12:55:35 wirkt viel zu Ã¼berladen 2006-03-25T12:56:39 ich glaub das mit dem admin panel lassen wir mal bis die grundfunktionen da sind 2006-03-25T12:56:57 ja 2006-03-25T13:21:22 *** jedie has joined #pocoo 2006-03-25T13:23:10 moin jedie 2006-03-25T13:23:12 lange nicht gesehen 2006-03-25T13:24:45 jup 2006-03-25T13:24:47 arbeit 2006-03-25T13:49:21 uff 2006-03-25T13:54:43 mitsuhiko: kde und usability *tss* 2006-03-25T13:56:15 xorAxAx: hast du den post von linus nicht gelesen? :-) 2006-03-25T13:56:26 mitsuhiko: ach den 2006-03-25T13:56:35 er war gegen gnome oder? 2006-03-25T13:56:42 jup 2006-03-25T13:56:52 er hat die gnome entwickler als "interface nazis" hingestellt 2006-03-25T13:56:57 das war zu viel des guten :) 2006-03-25T13:57:19 wann kommt ubuntu/freebsd? 2006-03-25T13:57:22 hey, das ist doch schon ein paar Wochen her... gibt's was neues? 2006-03-25T13:57:45 jedie: ney, es war nur gerade vorher eine diskussion Ã¼ber das pocoo admin panel 2006-03-25T13:57:50 * mitsuhiko wird einfach mal mit den GNOME leuten reden, die haben ahnung von usability *scnr* :) 2006-03-25T13:58:01 ich glaub um das gings gerade 2006-03-25T15:30:21 *** jedie is now known as jedie|afk 2006-03-25T15:39:22 * birkenfeld sucht mal wieder ne aufgabe 2006-03-25T15:39:46 birkenfeld: ich glaub da gibts noch genug :) 2006-03-25T15:39:55 * mitsuhiko wollte jetzt mal mit dem user objekt anfangen 2006-03-25T15:40:12 wenn das existiert kann man anfangen da settings anzulegen 2006-03-25T15:40:22 mitsuhiko: pocoo.pkg.core.user 2006-03-25T15:40:24 dann sollte gettext und die templates funktionieren 2006-03-25T15:40:36 schon fertig? uff 2006-03-25T15:40:43 fertig nicht unbedingt 2006-03-25T15:41:31 argh. das ist ja wirklich schon da :) 2006-03-25T15:41:34 * xorAxAx hört gerade till kampeter zu 2006-03-25T15:41:41 xorAxAx: wer ist das? 2006-03-25T15:43:13 birkenfeld: hast du eine idee, wie die Packages templates bereitstellen? 2006-03-25T15:43:17 +kÃ¶nnen 2006-03-25T15:43:18 mitsuhiko: druck,scan guru von mandriva,paris 2006-03-25T15:57:42 * mitsuhiko ist jetzt mal weg 2006-03-25T18:23:21 re 2006-03-25T18:32:30 birkenfeld: ping :) 2006-03-25T18:41:34 mitsuhiko: p*ng 2006-03-25T18:41:43 perfekt 2006-03-25T18:41:57 req.user ist ja momentan ein UserObj 2006-03-25T18:42:07 ja 2006-03-25T18:42:25 wÃ¼rde es nicht mehr sinn machen da einfach UserObj von User abzuleiten und zusÃ¤tzlich einfach noch ein identified attribut zu setzen? 2006-03-25T18:42:32 so hÃ¤tte man direkt zugriff auf den user 2006-03-25T18:43:02 der anonyme wÃ¼rde dann einfach die defaults aus der board config laden 2006-03-25T18:43:11 mitsuhiko: hab ich mir auch überlegt. man müsste aber dann sicherstellen dass das objekt nicht zufällig geändert wird 2006-03-25T18:43:35 jep 2006-03-25T18:44:00 aber es wäre sicher eine idee schneller 2006-03-25T18:46:18 der groÃŸteil der plugins muss nÃ¤mlich immer auf user settings zugreifen und da wÃ¤re ein req.user.settings['template'] sehr praktisch :) 2006-03-25T18:46:51 ja, warum nicht 2006-03-25T18:47:01 wie willst du die settings speicehrn? 2006-03-25T18:47:28 man kÃ¶nnte natÃ¼rlich fÃ¼r jedes ein feld in der db machen, oder einfach ein gepickeltes dict 2006-03-25T18:48:35 mitsuhiko: letzteres ist vielleicht besser, wegen erweiterbarkeit 2006-03-25T18:48:48 * mitsuhiko ist mal schnell essen 2006-03-25T18:48:55 wenn man nicht danach selecten muss 2006-03-25T18:49:04 birkenfeld: man mÃ¼sste irgendwie noch die board defaults aus der config noch nachladen 2006-03-25T18:49:08 da Ã¼berleg ich mir noch was 2006-03-25T18:50:41 okay 2006-03-25T18:59:34 *** jedie|afk has quit IRC 2006-03-25T19:02:36 re 2006-03-25T19:02:46 das nenn ich mal schnelles essen :) 2006-03-25T19:04:54 *** tabellar has joined #pocoo 2006-03-25T19:04:56 mal ein allgemeines problem 2006-03-25T19:05:41 wenn wir die ganzen configs in der pocoo.conf ablegen sind die zwar sehr schnell erreichbar, aber eine verÃ¤nderung der daten braucht einen neustart der anwendung, da viele componenten die config values direkt in der __init__ auslesen 2006-03-25T19:05:45 moin tabellar 2006-03-25T19:05:57 servus ... 2006-03-25T19:09:41 wie wÃ¤r's mit einer zentralen (Admin) reload() Function der componenten. Jede Componente ladet beim entsprechenden Trigger seine INIT Daten nach ... 2006-03-25T19:10:32 for comp in components: ... comp.reload() ... oder so Ã¤hnlich... 2006-03-25T19:10:58 sowas wÃ¤re denkbar 2006-03-25T19:14:24 WayToJavascript ist gut geworden :) 2006-03-25T19:14:56 tjo. aber javascript nervt immer noch :) 2006-03-25T19:29:01 argh 2006-03-25T19:29:13 da will wieder mal jemand Teile von pocoo unter der MIT lizenz 2006-03-25T19:29:14 ? 2006-03-25T19:50:33 * mitsuhiko ist mal mit der Familie karten spielen 2006-03-25T21:10:20 re 2006-03-25T21:10:34 wb birkenfeld 2006-03-25T21:10:50 birkenfeld: http://tryruby.hobix.com/ <-- sowas brauchen wir fÃ¼r python :) 2006-03-25T21:12:02 mitsuhiko: ich leite das mal weiter 2006-03-25T21:13:33 argh. immer diese scheiÃŸ lizenzfragen :-/ 2006-03-25T21:13:43 wieso? 2006-03-25T21:14:11 birkenfeld: weil ich gerade eine kontra GPL diskussion in #pythonpaste hatte/habe 2006-03-25T21:14:24 Windle will den pocoo db layer gerne MIT haben 2006-03-25T21:14:37 hehe 2006-03-25T21:16:47 *** mitsuhiko changes topic to "pocoo Developement Channel / Entwicklungskanal | http://trac.pocoo.org/ | wer Deutsch spricht, spricht Deutsch / try to avoid german words" 2006-03-25T21:16:50 ^^ 2006-03-25T21:17:07 mitsuhiko: Development 2006-03-25T21:17:24 argh. nicht schon wieder 2006-03-25T21:17:28 *** mitsuhiko changes topic to "pocoo Development Channel / Entwicklungskanal | http://trac.pocoo.org/ | wer Deutsch spricht, spricht Deutsch / try to avoid german words" 2006-03-25T21:23:39 *** Windle_ has joined #pocoo 2006-03-25T21:23:45 moin Windle_ 2006-03-25T21:23:52 hey there :) 2006-03-25T21:24:02 figured it'd make more sense to keep pocoo related questions here :) 2006-03-25T21:24:15 hehe 2006-03-25T21:24:16 hi Windle_ 2006-03-25T21:24:54 was looking over the db layer, I'm just wondering why the choice was made to go with a table dict, rather than just declaring in the class, stuff like: threads = many(backref='forum') 2006-03-25T21:25:03 and having the metaclass pull it together 2006-03-25T21:25:33 hi there birkenfeld 2006-03-25T21:25:37 hm 2006-03-25T21:25:44 truely i don't know :) 2006-03-25T21:26:23 a lot of the info in the table and column stuff could be defaulted, and then just let you override if necessary 2006-03-25T21:27:07 Windle_: I didn't want to add too much magic, and I wanted to stay close at SQLAlchemy 2006-03-25T21:27:54 but it's worth thinking about it 2006-03-25T21:29:10 well, having a more comprehensive metaclass would definitely require more objects to assemble them 2006-03-25T21:29:20 so it'd prolly be quite a bit more work than the current solution 2006-03-25T21:30:53 --> http://trac.pocoo.org/wiki/DatabaseApi 2006-03-25T21:31:07 SQLAlchemy attaches to an in-memory SQLite database as follows: e = create_engine('sqlite://filename=:memory:') 2006-03-25T21:31:10 zzzeek 2006-03-25T21:31:22 what does this mean? 2006-03-25T21:31:26 that's from "anonymous", not me 2006-03-25T21:31:32 hm 2006-03-25T21:31:36 I wondered myself 2006-03-25T21:32:10 hm. because it works quite well the way were using it ^^ 2006-03-25T21:32:32 mitsuhiko: I know you liked the declarative style of Django, which had me wondering if anyone else would be interested in working on a metaclass which uses similar syntax for SA 2006-03-25T21:32:58 Windle_: there is the ActiveMapper... 2006-03-25T21:33:02 this seems like a good first step, though obviously going farther would make the definition look quite different from the object you actually get in return 2006-03-25T21:33:18 yea, I definitely understand the desire to stay truer to SA 2006-03-25T21:33:43 this seems like a good compromise on making it easier to declare your Model all in one place, while still keeping close to SA 2006-03-25T21:35:01 oh, regarding the rss feed stuff in Pocoo 2006-03-25T21:35:30 that stuff is in WebHelpers (copied from Django) 2006-03-25T21:35:42 birkenfeld: argh. HttpResponse doesn't get the charset from the application. how should we pass that to the __call__ method? 2006-03-25T21:36:00 the only object knowing anything about the charset is the request object currently 2006-03-25T21:36:19 mitsuhiko: HttpResponse? 2006-03-25T21:36:36 --> http://trac.pocoo.org/browser/colubrid/trunk/colubrid/response.py 2006-03-25T21:37:09 Did you see my changes to it for the paste WSGIResponse? 2006-03-25T21:37:34 Windle_: jup. i've noticed that you're doing quite the same :) 2006-03-25T21:37:49 both look like django :) 2006-03-25T21:37:59 yup :) 2006-03-25T21:38:59 I use a StackedObjectProxy module global to retain encoding information though 2006-03-25T21:39:02 for charset and such 2006-03-25T21:39:44 i'm not that happy with global objects 2006-03-25T21:40:16 that would mean that you can't run two applications the same time within the same python instance 2006-03-25T21:40:28 you can with StackedObjectProxy's :) 2006-03-25T21:40:43 and its fully WSGI safe for multiple instances of the same app in the same thread 2006-03-25T21:41:04 Windle_: and if you have two applications? 2006-03-25T21:41:10 no prob 2006-03-25T21:41:22 its built for that 2006-03-25T21:41:54 hm 2006-03-25T21:45:27 http://svn.webwareforpython.org/Paste/trunk/paste/registry.py 2006-03-25T21:51:47 that would be one solution 2006-03-25T21:52:06 its very conveinent usability wise 2006-03-25T21:52:24 since you can just use module globals and when the app is running, they'll be correct 2006-03-25T21:52:25 jep. but i'm still no fan of those threading local context thingies :) 2006-03-25T21:52:41 exactly, thats why I was hoping to make it so ppl didn't need to worry about that 2006-03-25T21:52:59 you feed the Registry a normal object, you never need to worry about thread-locals 2006-03-25T21:54:55 I'm using it heavily in the full wsgi update to Pylons, here's a better example of using it, http://pylonshq.com/project/pylonshq/browser/Pylons/branches/fullwsgi/pylons/wsgiapp.py 2006-03-25T21:55:32 line 58 and below, where I create the request object, register it and some others. then the user can just use "request" without worrying about it. the usability is very nice 2006-03-25T21:55:50 it also helps for registering charsets and such, so that the Response object can use the proper charset encoding 2006-03-25T21:56:36 i will think about that 2006-03-25T22:01:06 in that same batch of code, line 51-56 will transform WSGIResponse objects, not sure if that'd help in Colobrid, I'd assume you're doing something similar already 2006-03-25T22:01:45 and line 61 registers the default charset encoding 2006-03-25T22:02:45 as far as I can tell, you have to have some module globals somewhere, just for usability 2006-03-25T22:02:59 Django has its django.settings stuff or whatever that it pulls for the charset 2006-03-25T22:04:22 yeah. in pocoo we have them too 2006-03-25T22:04:27 pocoo.settings.conf 2006-03-25T22:04:36 is that object request-safe? 2006-03-25T22:04:49 Windle_: no. since it don't have to be 2006-03-25T22:04:52 ie, if you run a pocoo wsgi app underneat a pocoo wsgi app, will pocoo.settings.conf use a stack? 2006-03-25T22:05:16 currently pocoo requires an own python instance per pocoo instance 2006-03-25T22:05:27 ahh, thats not very wsgi-friendly ;) 2006-03-25T22:05:30 i know 2006-03-25T22:06:07 and thats not ideal 2006-03-25T22:06:15 but no idea so far for fixing that :) 2006-03-25T22:06:28 well, Registry would fix it ;) 2006-03-25T22:06:34 for the module globals at least 2006-03-25T22:06:47 but not for the component system 2006-03-25T22:06:53 is there anything else thats not safe per instance? 2006-03-25T22:07:06 jup. the whole component system 2006-03-25T22:08:00 that's the main problem 2006-03-25T22:08:07 it would be possible to fix 2006-03-25T22:08:14 but this would make the application much slower 2006-03-25T22:08:42 what would? 2006-03-25T22:09:36 Windle_: putting components into thread local "dont-know-how-to-call-them" 2006-03-25T22:10:02 ah, what kind of components are we talking about? 2006-03-25T22:10:23 the StackedObjectProxy just proxies in, its fast enough that I doubt it'd be a bottleneck in a program using it heavily 2006-03-25T22:10:59 Windle_: http://trac.pocoo.org/browser/pocoo/trunk/pocoo/pkg/tests/json.py 2006-03-25T22:11:01 components like those 2006-03-25T22:11:16 .oO(although it's just a test page currently) 2006-03-25T22:12:01 that would be hard to make non instance related :( 2006-03-25T22:12:03 I don't see anything in there thats not instance-safe 2006-03-25T22:12:05 why? 2006-03-25T22:12:37 Windle_: for performance reasons classes inheriting from pocoo.api.Component get instanced directly after creation 2006-03-25T22:13:02 hmm 2006-03-25T22:13:27 well, if the Pocoo wsgi app creates the Components, it can hold references to them, and register them with the Registry during execution 2006-03-25T22:13:48 would that work? 2006-03-25T22:13:53 Windle_: pocoo currently holds references to them in pocoo.api.ComponentRegistry 2006-03-25T22:14:01 but not per thread 2006-03-25T22:14:06 what kind of object is that? 2006-03-25T22:14:13 Windle_: it's a metaclass 2006-03-25T22:14:51 so why not have the wsgi app itself hold the references? 2006-03-25T22:15:01 this won't be a problem 2006-03-25T22:15:10 that would make them instance safe 2006-03-25T22:15:18 but since components can depend from each other i have to import them 2006-03-25T22:15:36 for example the @remotecall decorator requires the RemoteCall component 2006-03-25T22:16:00 and since those components arn't thread safe as such this works :) 2006-03-25T22:16:16 they only get passed the request object when they want to interact with a user request 2006-03-25T22:16:56 well, if you register the same object on every call, its as if they were global, even though they're instance-safe 2006-03-25T22:17:35 for example, in Pylons, there's a 'g' object, anything attached is 'global'. That is, its available in all threads 2006-03-25T22:17:49 i know the pylons structure :) 2006-03-25T22:18:17 right, so in that file, line 37, and line 64. it retains a reference to the globals, while making the same object available everywhere 2006-03-25T22:18:30 its like your globals, but instance-safe. wouldn't that work with the Component stuff? 2006-03-25T22:19:01 for example you want to run two pocoo instances in the same python instance (one on /p1 and one on /p2) 2006-03-25T22:19:10 right 2006-03-25T22:19:22 pocoo would have to create all components twice (no problem) 2006-03-25T22:19:42 and safe them in one componenty registry (no problem) 2006-03-25T22:19:43 but :) 2006-03-25T22:19:50 why one component registry? 2006-03-25T22:19:57 why can't each instance have its own component registry? 2006-03-25T22:19:59 Windle_: that's because of the Interface system 2006-03-25T22:20:03 a 2006-03-25T22:20:04 no 2006-03-25T22:20:10 that would be possible 2006-03-25T22:20:11 but 2006-03-25T22:20:13 not faster :) 2006-03-25T22:20:21 faster than what? 2006-03-25T22:20:28 since each access to a component would result in a bunch of getattr redirects 2006-03-25T22:20:42 just a single one 2006-03-25T22:20:58 the StackedObjectProxy holds the real object, so its just a single getattr redirect 2006-03-25T22:21:05 hm 2006-03-25T22:21:22 birkenfeld: ping :) 2006-03-25T22:21:37 mitsuhiko: pong? 2006-03-25T22:22:04 birkenfeld: current discussion is about the problem with one python instance per pocoo instance 2006-03-25T22:22:10 you mentioned that some days before 2006-03-25T22:22:15 yep 2006-03-25T22:22:37 Windle_ worked one a magic object called StackedObjectProxy with might solve that problem 2006-03-25T22:23:03 do you think that will work? 2006-03-25T22:23:18 the Pylons fullwsgi branch is using it quite nicely if you want an actual working example 2006-03-25T22:24:09 ie, here's all the module "globals": http://pylonshq.com/project/pylonshq/browser/Pylons/branches/fullwsgi/pylons/__init__.py 2006-03-25T22:24:24 and the other file I mentioned earlier registers them all during the WSGI call 2006-03-25T22:25:00 multithreaded applications are quite complex :) 2006-03-25T22:25:09 Windle_: so the StackedObjectProxy dispatches access to the current app's objects? 2006-03-25T22:25:12 wsgi is worse than just multi-threaded 2006-03-25T22:25:27 since within a single thread, its possible you could have two differently configured instances of the same WSGI app 2006-03-25T22:25:58 birkenfeld: right, it internally uses a thread-local list, and proxies access to the current app's object 2006-03-25T22:26:16 birkenfeld: that way you don't need to worry about thread-safe stuff at all, just register a object, and its handled 2006-03-25T22:26:32 I wonder about performance though 2006-03-25T22:26:40 * mitsuhiko too :( 2006-03-25T22:26:42 its only a single additional getattr call 2006-03-25T22:26:58 Windle_: and a call for the currentThread 2006-03-25T22:27:15 yea, but the thread local is implemented in C if you're using Python 2.4 2006-03-25T22:27:57 if a single additional getattr call plus a thread-local is a performance bottleneck in your application, I'd be extremely impressed ;) 2006-03-25T22:28:06 bbiab, heading out for the day 2006-03-25T22:28:24 Windle_: thanks 2006-03-25T22:32:35 .oO(Da steh ich nun ich armer Thor und bin so klug als wie zuvor...) 2006-03-25T22:49:06 mitsuhiko: tja 2006-03-25T22:49:33 irgendwie wÃ¤rs mir lieber wsgi wÃ¼rde wie FastCGI funktionieren 2006-03-25T22:49:42 dann wÃ¤re das mit den mehreren instanzen nicht das problem 2006-03-25T22:51:13 ich hasse solche __getattr__ dinger 2006-03-25T22:51:18 ich auch 2006-03-25T22:51:25 sie zerstÃ¶ren dir() 2006-03-25T22:51:37 und langsam sind sie auch nocht 2006-03-25T22:51:38 -t 2006-03-25T22:51:47 genau 2006-03-25T22:51:49 -.- 2006-03-25T23:36:35 mitsuhiko: weißt du wie man einen apachen in ein chroot sperrt? 2006-03-25T23:37:07 hm 2006-03-25T23:37:27 auf welchem system? 2006-03-25T23:37:55 auf meinem :) 2006-03-25T23:38:03 linux nehm ich an 2006-03-25T23:38:10 ja 2006-03-25T23:38:31 ich hab ehrlich gesagt noch nie einen apachen in einem chroot gehabt 2006-03-25T23:39:51 birkenfeld: http://wiki.sprayen.de/index.php/Apache_2_in_Chroot_Umgebung 2006-03-25T23:40:00 vielleicht hilft das 2006-03-25T23:40:27 hm 2006-03-25T23:40:36 muss es ein ganzer apache sein? 2006-03-25T23:40:46 mitsuhiko: es reicht ein Python CGI-Skript 2006-03-25T23:41:15 einen python webserver wie paster/BaseHTTPServer oder sowas kann man eignetlich direkt in einem chroot starten 2006-03-25T23:42:15 hm hm 2006-03-25T23:43:11 ich hab nämlich hier ein so ein interaktives Python nachgebaut 2006-03-25T23:43:28 cool 2006-03-25T23:43:39 :) 2006-03-25T23:44:00 birkenfeld: ich auch, aber erst den client teil :) 2006-03-25T23:44:31 mitsuhiko: mit hilfe von urwid geht das recht fix 2006-03-25T23:45:34 argh. auf sowas komm ich mal wieder nicht :) 2006-03-25T23:46:32 birkenfeld: ist das eine reine CGI anwendung? 2006-03-25T23:46:35 oder wsgi? 2006-03-25T23:46:44 mitsuhiko: nö, CGI 2006-03-25T23:46:57 und das geht einigermaÃŸen schnell? 2006-03-25T23:47:31 das problem ist dass er jede taste einzeln sendet, das muss ich noch ändern 2006-03-25T23:47:54 birkenfeld: die ruby demo sendet die tastendrÃ¼ck erst bei enter 2006-03-25T23:48:00 eben 2006-03-25T23:48:21 ich portiers dir dann gerne auf WSGI :) 2006-03-25T23:48:42 birkenfeld: was hast du bei dir laufen? debian, suse, mandrake? 2006-03-25T23:48:50 gentoo 2006-03-25T23:49:10 birkenfeld: ich hab eigentlich geplant gehabt evalexception anzupassen und das frontend zu verbessern 2006-03-25T23:49:33 das ist auch sinnvoller 2006-03-25T23:49:57 nur das frontend in ein richtiges terminal zu verwandeln ist sau schwer 2006-03-25T23:53:36 birkenfeld: mir fÃ¤llt gerade ein, das es kein chroot braucht 2006-03-25T23:53:48 man muss ja eigentlich nur das ganze mit SuExec als nobody laufen lassen 2006-03-25T23:54:29 mitsuhiko: lesen kannst du trotzdem 2006-03-25T23:54:57 hm