3 * Missing conflict detection, just overwrites changes and does not svn up
6 How to resolve this? Would this work when saving a page?
9 2. svn up (so it can merge with any other changes)
10 3. check for conflicts
11 4. if conflicts, read in page w/conflict markers, svn revert page, and display resolution form
12 5. otherwise, svn commit
13 6. if commit fails, perhaps we were raced and there are conflicts still; so check for conflicts and if found, goto 2.
15 No, it won't work, because the other committer updates the srcdir when they make their commit, so svn does not get a chance to merge in the changes on update in step 2. Hmm..
17 Well, it could keep track of the version number of the page it started editing, and if it's changed during the edit, ask svn to merge all changes from that version number to head into the page after writing it out. Doing that without races with other svn committers seems tricky.
19 If each commit operated in a separate copy of the wc, that was created for the commit, the steps above would work. But it would be very expensive to make and maintain all those copies.
21 Note that unless we use `svn lock`, we really can't "warn" about concurrent edits and such the way other wikis do, which is ok, since that system sucks anyway. Using svn up for smart merging will be much nicer most of the time.
23 * Eventually, might want page deletion.
24 * Eventually, might want file upload.
28 * Should support RSS for notification of new and changed pages.
30 This can be a static rss file that is generated when the moo
31 is built. (As long as all changes to all pages is ok.)
33 * Should support mail notification of new and changed pages.
35 Hmm, should be easy to implement this.. it runs as a svn post-coommit hook
36 already, so just look at the userdb, svnlook at what's changed, and send
37 mails to people who have subscribed.
40 1. [[Joey]] mentioned that being able to subscribe to globs as well as
41 explicitly named pages would be desirable.
42 2. I think that since we're using Perl on the backend, being able to
43 let users craft their own arbitrary regexes would be good.
45 Joey points out that this is actually a security hole, because Perl
46 regexes let you embed (arbitrary?) Perl expressions inside them. Yuck!
48 It would also be good to be able to subscribe to all pages except discussion pages or the SandBox: `* !*/discussion !sandobx`, maybe --[[Joey]]
50 3. Of course if you do that, you want to have form processing on the user
51 page that lets them tune it, and probably choose literal or glob by
54 The first cut, I suppose, could use one sendmail process to batch-mail all
55 subscribers for a given page. However, in the long run, I can see users
56 demanding a bit of feature creep:
58 4. Each user should be able to tune whether they see the actual diff parts or
60 5. Each user should be able to set a maximum desired email size.
61 6. We might want to support a user-specified shibboleth string that will be
62 included in the email they receive so they can easily procmail the messages
67 ## pluggable renderers
69 I'm considering a configurable rendering pipeline for each supported
70 filename extension. So for ".mdwn" files, it would send the content through
71 linkify, markdown, and finalize, while for ".wiki" files it might send it
72 through just a wiki formatter and finalize.
74 This would allow not only supporting more types of markup, but changing
75 what style of [[WikiLink]]s are supported, maybe some people want to add
76 [[CamelCase]] for example, or don't like the [[SubPage/LinkingRules]].
78 The finalize step is where the page gets all the pretty junk around the
79 edges, so that clearly needs to be pluggable too.
81 There also needs to be a step before finalize, where stuff like lists of pages
82 that linked back to it could be added to the page. However, doing linkbacks
83 also needs to tie into the main logic, to determine what pages need to be
84 renered, so maybe that won't be a plugin.
88 Being case insensative is handy, but it does make the [[BackLinks]] a bit
89 ugly compared to other links. It should be possible to support pagenames
90 that have uppercase, while still allowing them to be linked to using any
95 Make the html valid. Add css.
99 Need a way to sign name in page that's easier to type than "--\[[Joey]]"
100 and that includes the date.
102 What syntax do other wikis use for this? I'm considering "\[[--]]" (with
103 spaces removed) as it has a nice nmemonic.
105 OTOH, adding additional syntax for this would be counter to one of the
106 design goals for ikiwiki: keeping as much markup as possible out of the
107 wiki and not adding nonstandard markup. And it's not significantly hard to
108 type "--\[[Joey]]", and as to the date, we do have page history.
110 ## recentchanges links to commit diffs
112 Would take a bit more viewcvs integration, let the be a "[diff]" link in
113 recentchanges that goes to the diff for any listed change.
115 ## recentchanges more than 100
117 Possibly add "next 100" link to it, but OTOH, you can just use svn log if
122 Need a toned down version of this wiki with a basic frontpage, sandbox and
123 docs to use as a seed for new wikis.
127 * full text (use third-party tools?)
128 * list of all missing pages
129 * list of all pages or some kind of page map
133 Might be nice to support automatically generating an index based on headers in a page, for long pages. The question is, how to turn on such an index?
137 Some wikis will need the abiity to lock a page, or the whole wiki, so that only admins can edit them. Probably using the same globbing as for recentchanges mails to determine what to lock.
139 Probably it's ok if locking is only supported for web commits.