-My field-etc branch in git://git.pseudorandom.co.uk/git/smcv/ikiwiki.git (gitweb:
-<http://git.pseudorandom.co.uk/smcv/ikiwiki.git?a=shortlog;h=refs/heads/field-etc>)
-has some fixes for compatibility with old YAML modules, mostly done by imitating
-Joey's code in IkiWiki::Setup::Yaml. Please consider merging :-) --[[smcv]]
+**Update:** I've submitted a patch, [rubykat/ikiplugins pull request #5](https://github.com/rubykat/ikiplugins/issues/5).
-> I would if I could *find* it. I checked out the "field-etc" branch, but I can't find the plugins in question under IkiWiki/Plugin; am I looking in the wrong place, or what?
-> --[[KathrynAndersen]]
+I have just opened [rubykat/ikiplugins issue #4](https://github.com/rubykat/ikiplugins/issues/4)
+regarding the fact that ymlfront doesn't seem to delete any old pagestate when fields have been
+removed in an edit. The fields are stuck there with their old values until a full rebuild. Seems
+to me ymlfront should just clear out all of the `{ymlfront}` pagestate before parsing the new
+stuff - including in the case where the new page has no ymlfront section at all.
->> Sorry, I accidentally removed `field-etc` by pushing with `--mirror` from a
->> different checkout. I've put it back; it's a branch from your `ikiplugins.git`,
->> so yes, the code should be in `IkiWiki/Plugin`. --[[smcv]]
+I discovered another slightly-different-but-related issue where simply _changing_ a field value
+in the YAML section doesn't always cause the generated HTML to be updated. Oddly, ikiwiki will
+_say_ it's building the page, but when you look at the HTML output, it's the old content.
+
+Could this involve some clever optimization where ikiwiki looks at the content (that's left over
+after ymlfront stripped out the YAML) and sees it hasn't changed? Does ymlfront need to do
+something more to indicate there is a change? Does the _template_ need to somehow be declared
+to depend on more stuff?
+
+As I said, the log does have a line for 'building' the page, so whatever optimization is happening
+must come later than the determination of what pages to 'build'.
+
+I'm mentioning it here because I'm not sure whether this or the issue on github will be seen
+first - there's a pretty old one open there. This seems to be quite
+potentially useful stuff that never quite got finished - is [[KathrynAndersen]] still
+interested? -- [[jcflack]]
+
+----
+Previous discussion re: delimiters
+
+Now that I have implemented a \[[!ymlfront ...]] directive, I would like to remove support for the old "---" delimited format, because
+
+* it is fragile (easily breakable)
+* it is non-standard
+
+Any objections?
+
+> Well, I don't have much standing since I have been too lame to integrate
+> ymlfront into ikiwiki yet. Buy, my opinion is, I liked the old
+> format of putting the YAML literally at the front of the file. It
+> seemed to allow parsing the file as YAML, using any arbitrary YAML
+> processer. And it was nice how it avoided boilerplate. --[[Joey]]
+
+>> The old delimited format also has the advantage of being remarkably similar to the
+>> [MultiMarkDown](http://fletcherpenney.net/multimarkdown/users_guide/multimarkdown_syntax_guide/)
+>> way of including metadata in documents. The only difference is that MMD doesn't expect the
+>> triple-dash separators, but I'm thinking about submitting a patch to MMD to actually support
+>> that syntax. --GB
+
+>>> Yes, the idea was to allow the file to be parsed as YAML, you're right. I just found that I tended to have problems when people used "---" for horizontal rules. However, I have also found that trying to keep it solely as an IkiWiki directive doesn't work either, since sometimes the meta-data I need also contained "]]" which broke the parsing of the directive.
+>>> So I have decided to go for a compromise, and make the delimiter configurable, rather than hardcoded as "---"; the triple-dash is the default, but it can be configured to be something else instead. I haven't pushed the change yet, but I have written it, and it seems to work. -- [[KathrynAndersen]]
+
+>>>> I'm not sure about what kind of problems you're meeting with "---" being used
+>>>> for horizontal rules: isn't it sufficient to just check that (1) the triple-dash
+>>>> is the first thing in the page and (2) there are only YAML-style assignments
+>>>> (and no blank lines) between the two markers? Check #2 would also be enough to
+>>>> support MMD-style metadata, which means (a) no start marker and (b) empty line
+>>>> to mark the end of the metadata block. Would this be supported by the plugin?
+>>>> --GB
+
+>>>>> Since I allow all legal YAML, the only way to check if it is legal YAML is to use the YAML parser, by which time one is already parsing the YAML, so it seems a bit pointless to check before one does so. -- KA