pandoc/TODO

# TODO

*   Revisions for building with windows under cygwin:
    Cabal under windows produces 'pandoc.exe', and some of the scripts
    expect 'pandoc'.

*   Consider making section headers block titles rather than blocks.
    Instead of:  [Header 1 "My title", Block1, Block2, Block3],
    Section "My title" [Block1, Block2, Block3].
    This seems cleaner and would facilitate a docbook writer.
    It might also simplify the rst reader.

*   pandoc's HTML output fails to validate completely (w3c).
    There are a few quirks:
    + HTML doesn't like the \> at the end of <meta tags.
      But if we remove them, we'll have trouble with S5 output,
      which seems to need the xhtml header?
    + There's also a problem with the email obfuscation scheme.
      <noscript> isn't allowed inside <p> blocks.  <script> is
      allowed!  Options:
          - come up with another scheme, perhaps more like markdown.pl's
          - ignore the validation problems
          - others?

*   Consider adding support for acronyms.
    Perhaps like this:  [AAAS]
      [AAAS]: "American association for the advancement of science"
    <acronym title="American association for the advancement 
    of science">AAAS</acronym>

*   Consider changing footnote syntax so that all footnotes in markdown
    are embedded (and automatic).^[Like this.  Here's a footnote.  It
    is parsed like a block, so you can have embedded code blocks:

         like this { code }

    ] That was the end of the note.  This means having block elements
    embedded in inline elements, which is possible.
    Advantage:  Much easier to write.  You don't have to pick a label,
    move down to type your note, move back up.
    Disadvantage:  Perhaps slightly harder to read.  (But HTML and LaTeX
    output will still be easy to read.)

*   Consider scrapping most of the wrapper scripts in favor of having
    symlinks to pandoc.  Modify pandoc so that it changes its defaults
    depending on the name of the calling program (getProgName).
    This would eliminate a lot of complexity and allow better handling
    of options (eliminating the need for a separation between wrapper
    and pandoc options, for example).

    If we do this, we should change option parsing in pandoc to allow
    options after arguments.  This will preserve backward-compatibility
    with the present wrapper system.  We'd also want to add an -o
    option to pandoc (output file).  When -o foo is specified, pandoc
    should print "Created foo" to stderr on success (unless --quiet
    is specified).

    A disadvantage is that we'd lose iconv conversion.  But maybe this
    isn't needed anymore; UTF-8 seems to be standard on most systems now. 

    The tricky wrappers to replace are markdown2pdf and html2markdown.

    markdown2pdf: 

        save working_directory
        create tempdir
        if markdown2latex "$@" >tempdir/output 2>tempdir/logfile; then
           extract output-file from logfile (this will be foo.pdf)
           if output-file found:
              mv foo.pdf tempdir/foo.tex
            else:
              mv tempdir/output tempdir/foo.tex
           cd tempdir
           run pdflatex on foo.tex to produce foo.pdf
           mv foo.pdf working_directory/foo.pdf     
        else:
           display logfile to inform user
        on exit:
           get rid of tempdir
 
    html2markdown:  needs to run the HTML through tidy (mainly because
    pandoc's html parser requires closing tags, etc.)  So we probably
    need something like the existing wrapper script here.  roktas
    suggests perhaps keeping html2markdown simple and using a separate
    script, web2markdown.   note:  we also need iconv here, since web
    pages may not be in UTF8.