X-Git-Url: http://git.vanrenterghem.biz/git.ikiwiki.info.git/blobdiff_plain/6a7db9a46c470735b1363449510f6da1d2a61742..da494462a8bd18913247ff09a8bcfe5fe6c0f809:/doc/todo/supporting_comments_via_disussion_pages.mdwn

diff --git a/doc/todo/supporting_comments_via_disussion_pages.mdwn b/doc/todo/supporting_comments_via_disussion_pages.mdwn
index 53028c06e..e0495c8c2 100644
--- a/doc/todo/supporting_comments_via_disussion_pages.mdwn
+++ b/doc/todo/supporting_comments_via_disussion_pages.mdwn
@@ -45,6 +45,8 @@ Is this simple enough to be sensible?
 
 >>> As a side note, the feature described above (having a form not to add a page but to expand it in a formated way) would be useful for other things when the content is short (timetracking, sub-todo list items, etc..) --[[hb]]
 
+# [[MarceloMagallon]]'s implementation
+
 I've been looking into this.  I'd like to implement a "blogcomments"
 plugin.  Looking at the code, I think the way to go is to have a
 formbuilder_setup hook that uses a different template instead of the
@@ -62,13 +64,25 @@ So, I have some code, included below.  For some reason that I don't quite get it
 
 What I ended up doing is write something like this to the page:
 
-    [[blogcomment from="""Username""" timestamp="""12345""" subject="""Some text""" text="""the text of the comment"""]]
+    [[!blogcomment from="""Username""" timestamp="""12345""" subject="""Some text""" text="""the text of the comment"""]]
+
+Each comment is processed to something like this:
+
+    <div>
+        <dl>
+            <dt>From</dt><dd>Username</dd>
+            <dt>Date</dt><dd>Date (needs fixing)</dd>
+            <dt>Subject</dt><dd>Subject text</dd>
+        </dl>
 
-Each comment is processed to a <<div>> with a <<dl>> and the text inside it.  In this way the comments can be styled using CSS.
+        <p>Text of the comment...</p>
+    </div>
+
+.  In this way the comments can be styled using CSS.
 
 -- [[MarceloMagallon]]
 
-# Code
+## Code
 
     #!/usr/bin/perl
     package IkiWiki::Plugin::comments;
@@ -121,7 +135,7 @@ Each comment is processed to a <<div>> with a <<dl>> and the text inside it.  In
             $cgi->param('comments') : '';
         my $comment=$cgi->param('blogcomment');
 
-        $content.=qq{[[blogcomment from="""$name""" timestamp="""$timestamp""" subject="""$subject""" text="""$comment"""]]\n\n};
+        $content.=qq{[[!blogcomment from="""$name""" timestamp="""$timestamp""" subject="""$subject""" text="""$comment"""]]\n\n};
         $content=~s/\n/\r\n/g;
         $form->field(name => "editcontent", value => $content, force => 1);
     } # }}}
@@ -147,4 +161,55 @@ Each comment is processed to a <<div>> with a <<dl>> and the text inside it.  In
         return $r;
     } # }}}
     
-    1;
\ No newline at end of file
+    1;
+
+# [[smcv]]'s implementation
+
+I've started a smcvpostcomment plugin (to be renamed to postcomment if people like it, but I'm namespacing it while it's still experimental) which I think more closely resembles what Joey was after. The code is cargo-culted from a mixture of editpage and inline's "make a blog post" support - it has to use a lot of semi-internal IkiWiki:: functions (both of those plugins do too). It doesn't fully work yet, but I'll try to get it into a state where it basically works and can be published in the next week or two.
+
+My approach is:
+
+* Comments are intended to be immutable after posting (so, only editable by direct committers), so they go on internal pages (*._comment); these internal pages are checked in to the RCS (although later I might make this optional)
+
+* ?do=smcvpostcomment (in the CGI script) gives a form that lets logged-in users (later, optionally also anonymous users) create a new comment
+
+* \[[!smcvpostcomment]] just inserts a "Post comment" button into the current page, which goes to ?do=smcvpostcomment - it's intended to be used in conjunction with an \[[!inline]] that will display the comments
+
+* The title (subject line), author and authorurl are set with \[[!meta]] directives, just like the way aggregate does it (which means I'll probably have to disallow the use of those \[[!meta]] directives in the body of the comment, to avoid spoofing - obviously, spoofing can be detected by looking at RecentChanges or gitweb, but the expectation for blog-style comments is that the metadata seen in the comment can be trusted)
+
+* The initial plan is to have comments hard-coded to be in Markdown, with further directives not allowed - I'll relax this when I've worked out what ought to be allowed!
+
+I've also updated Marcelo's code (above) to current ikiwiki, and moved it to a "marceloblogcomment" namespace - it's in the "marcelocomments" branch of my repository (see <http://git.debian.org/?p=users/smcv/ikiwiki.git;a=log;h=refs/heads/marcelocomments>). I had to reconstitute the .tmpl file, which Marcelo didn't post here.
+
+--[[smcv]]
+
+OK, the postcomment branch in my repository contains an implementation. What
+do you think so far? Known issues include:
+
+* The combination of RSS/Atom links and the "post new comment..." button is
+  ugly - I need a way to integrate the "new comment" button into the feed links
+  somehow, like the way inline embeds its own "new blog post..." feature
+  (I don't think the current way really scales, though)
+
+* There are some tweakables (whether to commit comments into the VCS, whether
+  wikilinks are allowed, whether directives are allowed) that are theoretically
+  configurable, but are currently hard-coded
+
+* The wikilink/directive disarming doesn't work unless you have
+  prefixdirectives set (which I just realised)
+
+* \[[!smcvpostcomment]] now displays the comments too, by invoking \[[!inline]]
+  with suitable parameters - but it does so in a very ugly way
+
+* Start-tags in a comment with no corresponding end-tag break page formatting
+  (unless htmltidy is enabled - inline and aggregate have the same problem)
+
+* There is no access control, so anonymous users can always comment, and so
+  can all logged-in users. Perhaps we need to extend canedit() to support
+  different types of edit? Or perhaps I should ignore canedit() and make the
+  access control configurable via a parameter to \[[!smcvpostcomment]]?
+  I'd like to be able to let anonymous (or at least non-admin) users comment
+  on existing pages, but not edit or create pages (but perhaps I'm being too
+  un-wikiish).
+
+--[[smcv]]