1 2011-09-20T01:14:45 *** MattMaker
2 2011-09-20T01:26:45 *** MattMaker
3 2011-09-20T03:49:18 *** vpv
4 2011-09-20T03:49:25 *** vpv
5 2011-09-20T04:53:50 *** moinBot`
6 2011-09-20T04:54:05 *** moinBot
7 2011-09-20T10:43:38 *** mkerrin
8 2011-09-20T11:23:57 <ThomasWaldmann> moin
9 2011-09-20T11:24:11 <ronny> moin ThomasWaldmann
10 2011-09-20T11:24:40 *** greg_f
11 2011-09-20T11:25:13 <ThomasWaldmann> rename is missing
12 2011-09-20T11:26:36 <ronny> ThomasWaldmann: as in changing the name metadata of a item?
13 2011-09-20T11:27:11 <ronny> ThomasWaldmann: i tihnk there is a need for general utilities to take a revision metadata item and change it into one suitable for the next version
14 2011-09-20T11:27:12 <ThomasWaldmann> well, considering router, that can have consequences
15 2011-09-20T11:28:09 <ronny> hmm
16 2011-09-20T11:28:18 <ronny> i think we might have a can of conner cases there
17 2011-09-20T11:28:41 <ronny> also today im busy
18 2011-09-20T14:10:00 <ThomasWaldmann> http://etherpad.osuosl.org/moin2-storage < thoughts about rename and trash, please comment
19 2011-09-20T15:44:37 <ronny> ThomasWaldmann: looks sane at first glancem its true that we can have the same item uuid in different backends, and the latest one will win
20 2011-09-20T15:44:45 <ronny> bbl, relocating
21 2011-09-20T15:46:28 *** xorRaxRax
22 2011-09-20T15:47:59 *** xorAxAx
23 2011-09-20T16:04:38 *** xorRaxRax
24 2011-09-20T18:13:24 *** marchael
25 2011-09-20T18:21:56 *** marchael
26 2011-09-20T18:58:26 *** mkerrin
27 2011-09-20T19:07:58 *** greg_f
28 2011-09-20T19:10:24 * ThomasWaldmann killed the trash backend
29 2011-09-20T19:19:18 <ronny> ThomasWaldmann: you mean no different backends for trash vs normal?
30 2011-09-20T19:19:39 <ThomasWaldmann> yes, as i said in etherpad
31 2011-09-20T19:19:50 <ronny> i just got back, need to reopen
32 2011-09-20T19:20:12 <ronny> ThomasWaldmann: imho there should be a way to move a complete item from one backend to another later on tho
33 2011-09-20T19:20:15 <ThomasWaldmann> no new stuff there sind today noon
34 2011-09-20T19:20:24 <ThomasWaldmann> ronny: for?
35 2011-09-20T19:22:35 <ronny> ThomasWaldmann: inter-backend rename shouldnt loose history from the other backend when one resuffls
36 2011-09-20T19:22:51 <ThomasWaldmann> i mean if the backends are for really different stuff, like content vs. userprofiles, you don't need to move stuff between them
37 2011-09-20T19:24:02 <ThomasWaldmann> and if you just want to get all your stuff into a new, differently implemented backend, you can serialize/deserialize
38 2011-09-20T19:24:35 <ronny> brb, dinner
39 2011-09-20T19:26:29 <ThomasWaldmann> hmm, if we used hierarchical acls and backends are mounted at some specific places, one item at the mountpoint could replace the per-backend acl
40 2011-09-20T19:38:20 *** MattMaker
41 2011-09-20T19:38:40 *** MattMaker
42 2011-09-20T20:00:24 *** MattMaker
43 2011-09-20T20:00:49 *** MattMaker
44 2011-09-20T20:07:19 <ronny> ThomasWaldmann: depends on the amount of work you want to have involved
45 2011-09-20T20:07:42 <ronny> (cpu work that is)
46 2011-09-20T20:08:52 <ThomasWaldmann> right, one lookup more
47 2011-09-20T20:14:53 <ThomasWaldmann> hmm, i guess we rather want acls on parts of the namespace
48 2011-09-20T20:15:01 <ThomasWaldmann> (configured acls)
49 2011-09-20T20:15:16 <ThomasWaldmann> like for UserProfiles/*
50 2011-09-20T20:15:35 <ThomasWaldmann> then we don't need a separate backend just for that reason
51 2011-09-20T20:15:51 <ThomasWaldmann> simple setups then do not even need the router
52 2011-09-20T20:19:54 <ronny> ThomasWaldmann: we could create a index of inherited acl's
53 2011-09-20T20:20:06 <ronny> where we really just have acl + names
54 2011-09-20T20:20:32 <ronny> the question is integrate with normal index or have another index
55 2011-09-20T20:20:39 <ronny> we should test what makes more sense
56 2011-09-20T20:24:29 <ThomasWaldmann> ronny: we have the item acls in index
57 2011-09-20T20:25:03 <ThomasWaldmann> and if we compute "effective acls" (that consider parents, config, etc.), we have a dependency / consistency problem
58 2011-09-20T20:25:35 <ThomasWaldmann> e.g. if you update a acl in config or in the wiki, you'ld have to recompute all childs also
59 2011-09-20T20:26:22 <ronny> ThomasWaldmann: which is why i would lazyly put it into a different index, and do search+destroy on parent update
60 2011-09-20T20:26:53 <ThomasWaldmann> ah, you mean just as a cache? yes, that makes sense.
61 2011-09-20T20:27:09 <ThomasWaldmann> we could also use caching system for that
62 2011-09-20T20:27:18 <ronny> true
63 2011-09-20T20:27:27 <ronny> we need efficient invalidation tho
64 2011-09-20T20:27:50 <ronny> whenever parent acl content changes, all childs need to be killed from the cache
65 2011-09-20T20:28:00 <ronny> which is why shoosh index came to mind
66 2011-09-20T20:28:15 <ThomasWaldmann> or just make them to be not found
67 2011-09-20T20:29:02 <ThomasWaldmann> access key could be a *parents_and_myself_revids tuple
68 2011-09-20T20:30:31 <ronny> ThomasWaldmann: wouldnt that require to access all whoosh latest docs for all parents anway
69 2011-09-20T20:31:04 <ronny> ThomasWaldmann: if we need to fetch the docs for the parents to calculate the access key, we dont need to cache the effective acl
70 2011-09-20T20:31:47 <ronny> (with parent i mean relation in the tree of items)
71 2011-09-20T20:35:52 <ThomasWaldmann> ronny: it would need index access for revid determination, but would save recomputing acl
72 2011-09-20T20:36:03 <ThomasWaldmann> maybe not worth...
73 2011-09-20T20:36:13 <ThomasWaldmann> we just try without caching I guess
74 2011-09-20T20:37:22 <ronny> ThomasWaldmann: i wonder if we can effectively ask whoosh for the docs for all parts of a item name
75 2011-09-20T20:49:35 <ThomasWaldmann> ronny: shall we try implanting today?
76 2011-09-20T20:51:14 <ThomasWaldmann> ronny: could you fix the memcached store tests like the others?
77 2011-09-20T20:53:07 <ThomasWaldmann> i guess memcached should be rather considered a specific protocol to talk to a server than a memory-only server implementation
78 2011-09-20T20:55:08 <ronny> ThomasWaldmann: memcached cant be a real backend anyway, can we kill it?
79 2011-09-20T20:55:57 <ThomasWaldmann> ronny: the kyoto stuff (for which we have an own backend now has also a memcached protocol similation AND persistence)
80 2011-09-20T20:56:14 <ThomasWaldmann> so the memcached protocol does not necessarily mean == RAM
81 2011-09-20T20:56:38 <ThomasWaldmann> also, who known, some day we might have persistent RAM :D
82 2011-09-20T20:56:46 <ThomasWaldmann> knows*
83 2011-09-20T20:58:42 <ThomasWaldmann> brunomartin: you may need some new serializer soon :)
84 2011-09-20T21:02:28 <brunomartin> ThomasWaldmann: don't care with us... you and Ronny are doing a great work! ;-)
85 2011-09-20T21:03:13 <ronny> ThomasWaldmann: but memcached has no list keys
86 2011-09-20T21:03:43 <ThomasWaldmann> ah, right. I'll check again...
87 2011-09-20T21:06:00 *** grzywacz
88 2011-09-20T21:06:01 *** grzywacz
89 2011-09-20T21:08:22 <ThomasWaldmann> there is a way, but rather dirty.
90 2011-09-20T21:08:31 <ThomasWaldmann> ok i kill memcached store
91 2011-09-20T21:12:08 <ThomasWaldmann> pushed
92 2011-09-20T21:33:28 *** raignarok
93 2011-09-20T21:43:06 <dreimark> brunomartin: please write a bit more on your homepage on MM
94 2011-09-20T21:43:16 <dreimark> and in english
95 2011-09-20T21:46:51 <brunomartin> dreimark: yeah, I will do it... I just created it... I was testing the notifications when you subscribe some item! ;-)
96 2011-09-20T21:51:11 <dreimark> you won't get a notification for your owns
97 2011-09-20T21:56:05 <brunomartin> dreimark: is there someone working on notifying for moin2?
98 2011-09-20T22:02:17 <ThomasWaldmann> brunomartin: afaik no
99 2011-09-20T22:02:34 <ThomasWaldmann> there is some mail code in there, and some event system
100 2011-09-20T22:04:55 <ThomasWaldmann> mail.sendmail
101 2011-09-20T22:05:22 <ThomasWaldmann> signalling.*
102 2011-09-20T22:14:10 <brunomartin> ThomasWaldmann: I think signalling fits for now... thanks...
103 2011-09-20T22:18:30 <brunomartin> ThomasWaldmann: I will work on notification for changes to be shown when the user log in, not just by mail... I show you when it's done.. ;-)
104 2011-09-20T22:18:32 <ThomasWaldmann> brunomartin: make sure you use the index :)
105 2011-09-20T22:19:09 <ThomasWaldmann> what's the point of that?
106 2011-09-20T22:19:29 <ThomasWaldmann> if you want to know what's new after logging in, you could look on global history
107 2011-09-20T22:21:54 <brunomartin> what if the user wants to see what has changed just on subscribed items, not all wiki?
108 2011-09-20T22:24:14 <ThomasWaldmann> ok, maybe useful for a busy wiki
109 2011-09-20T22:24:53 <ThomasWaldmann> instead of events and remembering them, one could also just filter global history btw.
110 2011-09-20T22:26:42 <brunomartin> ThomasWaldmann: great idea! :-)
111 2011-09-20T22:28:52 <ThomasWaldmann> you have to think and discuss about efficient ways for that
112 2011-09-20T22:31:05 <brunomartin> ThomasWaldmann: there is no global history on moin2 yet, right?
113 2011-09-20T22:31:15 <dreimark> brunomartin: it is
114 2011-09-20T22:31:24 <dreimark> see History
115 2011-09-20T22:31:37 <dreimark> http://test.moinmo.in/+history
116 2011-09-20T22:31:56 <dreimark> brunomartin: that was a project of my gsoc student
117 2011-09-20T22:34:48 <brunomartin> dreimark: Oh! sorry... I found it! :-D
118 2011-09-20T22:42:12 <brunomartin> dreimark: now I know what the user.bookmark does....
119 2011-09-20T22:43:29 <dreimark> dreimark: :)
120 2011-09-20T22:43:47 <brunomartin> dreimark: well, I need to do some filters, what do you think I can improve there? I saw a todo here....
121 2011-09-20T22:43:50 <brunomartin> ;-)
122 2011-09-20T22:45:22 <dreimark> ThomasWaldmann has added an image with some ideas to make it not consuming so much lines
123 2011-09-20T22:45:41 <dreimark> that may also simplifier special pagination for history
124 2011-09-20T22:48:09 <dreimark> this must be somewhere on the issue tracker
125 2011-09-20T22:49:22 <dreimark> https://bitbucket.org/sinha/moin2/issue/5/history-improvements
126 2011-09-20T22:49:28 <dreimark> looks at images
127 2011-09-20T22:50:24 <dreimark> s/simplifier/simplify/
128 2011-09-20T22:51:40 <dreimark> brunomartin: http://etherpad.osuosl.org/moin-2-ui line 50
129 2011-09-20T22:52:01 <dreimark> we had some ideas but it was not implemented
130 2011-09-20T23:10:31 <dreimark> gn
131 2011-09-20T23:16:09 * ThomasWaldmann implements a sqla store
132 2011-09-20T23:18:08 <brunomartin> dreimark: great job! :-) I will study all this now and come back later with doubts and purposes...
133 2011-09-20T23:25:48 * ThomasWaldmann transplants storage-ng now :)
134 2011-09-20T23:33:28 *** raignarok
135 2011-09-20T23:49:18 *** raignarok
136
