1 2011-09-08T03:02:02 *** MattMaker_
2 2011-09-08T03:03:48 *** MattMaker
3 2011-09-08T03:03:49 *** MattMaker_
4 2011-09-08T05:56:52 *** Marchael
5 2011-09-08T06:43:15 *** Marchael
6 2011-09-08T07:17:47 *** raignarok
7 2011-09-08T07:32:54 *** Marchael
8 2011-09-08T07:33:40 *** raignarok
9 2011-09-08T07:52:47 <ThomasWaldmann> moin
10 2011-09-08T08:09:10 <dreimark> moin
11 2011-09-08T08:57:57 *** Marchael
12 2011-09-08T10:36:23 *** mkerrin
13 2011-09-08T14:45:46 *** greg_f
14 2011-09-08T15:11:21 *** xjjk
15 2011-09-08T15:11:33 *** xjjk
16 2011-09-08T16:07:33 *** Marchael
17 2011-09-08T17:07:27 *** Marchael
18 2011-09-08T17:35:06 *** mkerrin
19 2011-09-08T18:03:13 *** Marchael
20 2011-09-08T19:02:19 *** greg_f
21 2011-09-08T19:32:47 *** raignarok
22 2011-09-08T19:49:06 *** raignarok_
23 2011-09-08T19:49:33 *** raignarok
24 2011-09-08T20:11:53 * ThomasWaldmann continues on storage tests
25 2011-09-08T20:33:09 *** brunosmartin
26 2011-09-08T20:34:08 <brunosmartin> ThomasWaldmann: moin
27 2011-09-08T20:34:43 <ThomasWaldmann> hi brunosmartin
28 2011-09-08T20:35:18 <brunosmartin> we have and idea here that we'd like to discuss with you (3 people here)
29 2011-09-08T20:35:39 <ThomasWaldmann> sure
30 2011-09-08T20:35:44 <brunosmartin> we need the same items to be in different paths
31 2011-09-08T20:35:54 <brunosmartin> so we're thinking about a "link" content type
32 2011-09-08T20:36:33 <brunosmartin> this is because the same item will be visible in different contexts. we need different groups of people to discuss them separated, but if they edit they'll all be editing the same object
33 2011-09-08T20:36:39 <brunosmartin> item I mean
34 2011-09-08T20:36:44 <brunosmartin> does this make sense for you?
35 2011-09-08T20:37:30 <ThomasWaldmann> not yet
36 2011-09-08T20:38:32 <ThomasWaldmann> how can it be separate discussion if it is same item?
37 2011-09-08T20:38:38 <brunosmartin> well, we're using moin in a non-ortodox way :-).
38 2011-09-08T20:38:51 <brunosmartin> we use it for classes
39 2011-09-08T20:39:01 <brunosmartin> so, one teacher has an item that is content for a class
40 2011-09-08T20:39:32 <brunosmartin> and he wants to use the same item in two different classes at same time
41 2011-09-08T20:39:48 <brunosmartin> so, there is a path where all items of this class will be located
42 2011-09-08T20:39:57 <brunosmartin> and some items generated in this class will be reused in other classes
43 2011-09-08T20:40:11 <brunosmartin> some items are relevant in a general context, others are just important in context of one class
44 2011-09-08T20:40:32 <brunosmartin> so, the teacher would have his repository of reusable items and would use it in each context
45 2011-09-08T20:40:44 <ThomasWaldmann> you want symlinks or hardlinks? :)
46 2011-09-08T20:41:33 <brunosmartin> not sure yet :-). our original idea would be symlink, but it might work better with hardlink
47 2011-09-08T20:41:37 <ThomasWaldmann> btw, such stuff might get pretty easy once items are uuid based
48 2011-09-08T20:41:39 <brunosmartin> hardlink would keep all comments
49 2011-09-08T20:42:03 <ThomasWaldmann> then name is just metadata of an anon item, and noone keeps us back from stuff having multiple names at once
50 2011-09-08T20:42:20 <brunosmartin> I think that would solve our problem
51 2011-09-08T20:42:39 <brunosmartin> what's needed for this to work?
52 2011-09-08T20:42:44 <ThomasWaldmann> (what we need then is refcounting, though :)
53 2011-09-08T20:42:51 <ThomasWaldmann> uuid working
54 2011-09-08T20:42:55 <brunosmartin> tomorrow we'll have to present a development cronogram
55 2011-09-08T20:43:07 <ThomasWaldmann> storage being based on uuid, not name
56 2011-09-08T20:43:32 <brunosmartin> we can: wait for you to do it if in your plans, or do it ourselves here, or wait for you to come here and we do together
57 2011-09-08T20:43:52 <brunosmartin> what do you suggest? is this already in your short-term roadmap?
58 2011-09-08T20:44:09 <ThomasWaldmann> that stuff is twisted
59 2011-09-08T20:44:38 <ThomasWaldmann> i want this for a long time already, but every atempt yet had to be cancelled, because too much broke
60 2011-09-08T20:45:27 <brunosmartin> hmmm... maybe this should be focus of our immersion
61 2011-09-08T20:45:41 <ThomasWaldmann> my current idea is to look at storage tests, make them better, maybe rewrite them, have a basic set of simple tests to make further work more simpler
62 2011-09-08T20:46:32 <brunosmartin> ok. this will be more relevant next year, when we have second generation of classes.
63 2011-09-08T20:46:54 <brunosmartin> I think this information is enough for us to plan ourselves, I think we can work on this together
64 2011-09-08T20:47:33 <ThomasWaldmann> brunosmartin: maybe you could think a bit ahead about information flow
65 2011-09-08T20:48:00 <brunosmartin> what do you mean?
66 2011-09-08T20:48:01 <ThomasWaldmann> currently the code is a bit messed, esp. about where UUIDs are created, used
67 2011-09-08T20:48:27 <ThomasWaldmann> and also they live at item level and at revision level
68 2011-09-08T20:49:09 <brunosmartin> do you think it's realistic to consider this will be done until november, after our sprint here?
69 2011-09-08T20:49:23 <ThomasWaldmann> i hope we can work on that
70 2011-09-08T20:49:35 <ThomasWaldmann> and get it done
71 2011-09-08T20:49:48 <brunosmartin> ok
72 2011-09-08T20:50:01 <ThomasWaldmann> storage was designed without that indexing in mind that we have now in form of whoosh
73 2011-09-08T20:50:15 *** MattMaker
74 2011-09-08T20:50:52 <ThomasWaldmann> what also needs more than one mind thinking about it:
75 2011-09-08T20:51:24 <ThomasWaldmann> currently storage uses (name, revno) to address a revision (name to address an item)
76 2011-09-08T20:52:10 <ThomasWaldmann> i think (item-uuid, revno) is better to address a revision (and uuid to address an item)
77 2011-09-08T20:52:40 <brunosmartin> ok, and them there would be no uuid for revision
78 2011-09-08T20:52:43 <ThomasWaldmann> but the main question still unanswered is: do we need BOTH ways? or is the last way enough?
79 2011-09-08T20:53:26 <brunosmartin> won't the first way fail if you rename an item?
80 2011-09-08T20:53:31 <ThomasWaldmann> (the uuid it writes into revision data currently is the item-uuid, stays the same for all revisions of same item and links them together)
81 2011-09-08T20:53:50 <ThomasWaldmann> yes, that's just one of the issues of that
82 2011-09-08T20:54:20 <ThomasWaldmann> but, there might be backends which are based on names somehow (maybe vcs backends)
83 2011-09-08T20:55:07 <brunosmartin> well, but uuid can be in a common fs and work with vcs
84 2011-09-08T20:55:18 <brunosmartin> uuid is name, isn't it?
85 2011-09-08T20:55:45 <ThomasWaldmann> of course you could use a uuid as a name, but as far as vcs storage is concerned, people might rather expect names there
86 2011-09-08T20:55:57 <brunosmartin> ah, makes sense
87 2011-09-08T20:56:23 <brunosmartin> but then vcs must support hardlinks for this to work
88 2011-09-08T20:56:24 <ThomasWaldmann> otoh, if we think global and unicode, there are often issues with unicode names
89 2011-09-08T20:57:05 <ThomasWaldmann> there could be one preferred name or so
90 2011-09-08T20:57:24 <ThomasWaldmann> maybe just think about how you would expect a hg backend to work
91 2011-09-08T20:57:47 <ThomasWaldmann> because i don't think we have an issue with fs* or sqla backends
92 2011-09-08T21:00:52 <ThomasWaldmann> there is also some people who would like to "edit" wiki items via (d)vcs, but when thinking about that, it gets even worse:
93 2011-09-08T21:01:23 <ThomasWaldmann> filesystems usually store files. content = data, but metadata is too limited for moin's needs.
94 2011-09-08T21:01:38 <ThomasWaldmann> usually there is just name, mtime and not much more
95 2011-09-08T21:03:23 <ThomasWaldmann> and if you additionally consider MS windows ... :-*
96 2011-09-08T21:05:24 <brunosmartin> :-)
97 2011-09-08T21:05:41 <brunosmartin> well, people can always install linux and do it the right way :-P
98 2011-09-08T21:05:55 <brunosmartin> what if we had a .metadata file for each item?
99 2011-09-08T21:06:20 <brunosmartin> you have MyItem.rst, for example, and MyItem.rst.metadata
100 2011-09-08T21:06:33 <brunosmartin> the then history would have to consider both files
101 2011-09-08T21:07:05 <ThomasWaldmann> yeah, that would be a workaround, but suck :)
102 2011-09-08T21:07:24 <ThomasWaldmann> e.g. if you rename the one, you also have to do the same to the other
103 2011-09-08T21:07:33 <brunosmartin> yes
104 2011-09-08T21:07:52 <brunosmartin> and having multiple file paths will really be an issue. maybe this would be optional feature, depending on storage
105 2011-09-08T21:09:44 <brunosmartin> is there already support for comments in item?
106 2011-09-08T21:10:03 <ThomasWaldmann> what do you mean?
107 2011-09-08T21:10:31 <brunosmartin> discussion page
108 2011-09-08T21:10:41 <brunosmartin> it's a separated item, right?
109 2011-09-08T21:11:20 <ThomasWaldmann> yes
110 2011-09-08T21:11:41 <brunosmartin> ok
111 2011-09-08T21:12:22 <brunosmartin> well, great! now we know your plans and can do our roadmap knowing that we're converging in future
112 2011-09-08T21:12:44 <brunosmartin> thanks!
113 2011-09-08T21:14:10 <ThomasWaldmann> yw :)
114 2011-09-08T21:35:36 *** brunosmartin
115 2011-09-08T22:31:58 *** raignarok_
116 2011-09-08T22:36:12 *** MattMaker
117 2011-09-08T23:17:07 *** raignarok
118 2011-09-08T23:22:30 *** Marchael
119
