summaryrefslogtreecommitdiff
path: root/doc/users/jasonblevins.mdwn
blob: 0030a30d3fd15937e324f3b0508a2ec34efc2713 (plain)

[[!meta title="Jason Blevins"]]

I'm currently hosting a private ikiwiki for keeping research notes which, with some patches and a plugin (below), will convert inline LaTeX expressions to MathML. I'm working towards a patchset and instructions for others to do the same.

I've setup a test ikiwiki here where I've started keeping a few notes on my progress. There is an example of inline SVG on the homepage (note that the logo scales along with the font size). There are a few example mathematical expressions in the sandbox. The MathML is generated automatically from inline LaTeX expressions using an experimental plugin I'm working on.

My (also MathML-enabled) homepage: http://jblevins.org/ (still using Blosxom...maybe one day I'll convert it to ikiwiki...)

Current ikiwki issues of interest:

  • [[bugs/recentchanges_feed_links]]
  • [[bugs/HTML_inlined_into_Atom_not_necessarily_well-formed]]
  • [[plugins/toc/discussion]]
  • [[todo/BibTeX]]
  • [[todo/svg]]
  • [[todo/Option_to_make_title_an_h1?]]
  • [[bugs/SVG_files_not_recognized_as_images]]

Plugins

These plugins are experimental. Use them at your own risk. Read the perldoc documentation for more details. Patches and suggestions are welcome.

  • mdwn_itex - Works with the mdwn plugin to convert inline LaTeX expressions to MathML using itex2MML.

  • h1title - If present, use the leading level 1 Markdown header to set the page title and remove it from the page body.

  • code - Whole file and inline code snippet syntax highlighting via GNU Source-highlight. The list of supported file extensions is configurable. There is also some preliminary documentation. See the FortranWiki for examples.

  • metamail - a plugin for loading metadata from email-style headers at top of a file (e.g., title: Page Title or date: November 2, 2008 11:14 EST).

  • pandoc - Markdown page processing via Pandoc (a Haskell library for converting from one markup format to another). LaTeX and reStructuredText are optional.

  • path - Provides path-specific template conditionals such as IS_HOMEPAGE and IN_DIR_SUBDIR.

MathML and SVG support

So far, I've made some notes on sanitizing MathML and SVG via htmlscrubber on the [[todo/svg]] todo item.

I've also worked out some content-negotiation issues. First of all, one needs to modify the default templates to use the XHTML+MathML+SVG doctype (see e.g., this patch). For most browsers, the content type of the pages should be application/xhtml+xml. The solution is easy if you want to just send application/xhtml+xml to everybody: just change the content type of .html files across the board.

However, if you want to support browsers that don't accept application/xhtml+xml (and those that will but say they don't, such as IE with the MathPlayer plugin), then one needs a mod_rewrite rule like the following:

RewriteCond %{HTTP_ACCEPT} application\/xhtml\+xml [OR]
RewriteCond %{HTTP_USER_AGENT} (W3C.*Validator|MathPlayer)
RewriteRule \.html$ - [T=application/xhtml+xml]

This solves the problem of MathML and inline SVG in static pages but some additional work is required for dynamically generated pages, like page previews, that are generated by ikiwiki.cgi. We need to allow ikiwiki.cgi to set the content type dynamically based on the HTTP_CONTENT_TYPE environment variable (e.g., with the following patch). Then, the following rewrite rules can pass the correct content type to ikiwiki:

RewriteCond %{HTTP_ACCEPT} application\/xhtml\+xml [OR]
RewriteCond %{HTTP_USER_AGENT} (W3C.*Validator|MathPlayer)
RewriteRule ikiwiki.cgi$ - [T=application/xhtml+xml]

One final critical issue is that a production-ready setup needs to implement some sort of on-the-fly error handling. If a user submits an invalid LaTeX expression or SVG code (not malicious, just invalid) and saves the page, then browsers like Firefox will halt processing of the page, preventing any further viewing or editing. A less than optimal solution is to force users to preview the page before saving. That way if someone introduces invalid XHTML then they can't save the page in the first place (unless they post directly to the right URL).