Souke/vendor/ezyang/htmlpurifier/TODO

TODO List

= KEY ====================
    # Flagship
    - Regular
    ? Maybe I'll Do It
==========================

If no interest is expressed for a feature that may require a considerable
amount of effort to implement, it may get endlessly delayed. Do not be
afraid to cast your vote for the next feature to be implemented!

Things to do as soon as possible:

 - http://htmlpurifier.org/phorum/read.php?3,5560,6307#msg-6307
 - Think about allowing explicit order of operations hooks for transforms
 - Fix "<.<" bug (trailing < is removed if not EOD)
 - Build in better internal state dumps and debugging tools for remote
   debugging
 - Allowed/Allowed* have strange interactions when both set
 ? Transform lone embeds into object tags
 - Deprecated config options that emit warnings when you set them (with'
   a way of muting the warning if you really want to)
 - Make HTML.Trusted work with Output.FlashCompat
 - HTML.Trusted and HTML.SafeObject have funny interaction; general
   problem is what to do when a module "supersedes" another
   (see also tables and basic tables.)  This is a little dicier
   because HTML.SafeObject has some extra functionality that
   trusted might find useful.  See http://htmlpurifier.org/phorum/read.php?3,5762,6100

FUTURE VERSIONS
---------------

4.6 release [OMG CONFIG PONIES]
 ! Fix Printer. It's from the old days when we didn't have decent XML classes
 ! Factor demo.php into a set of Printer classes, and then create a stub
   file for users here (inside the actual HTML Purifier library)
 - Fix error handling with form construction
 - Do encoding validation in Printers, or at least, where user data comes in
 - Config: Add examples to everything (make built-in which also automatically
   gives output)
 - Add "register" field to config schemas to eliminate dependence on
   naming conventions (try to remember why we ultimately decided on tihs)

5.0 release [HTML 5]
 # Swap out code to use html5lib tokenizer and tree-builder
 ! Allow turning off of FixNesting and required attribute insertion

5.1 release [It's All About Trust] (floating)
 # Implement untrusted, dangerous elements/attributes
 # Implement IDREF support (harder than it seems, since you cannot have
   IDREFs to non-existent IDs)
     - Implement <area> (client and server side image maps are blocking
       on IDREF support)
 # Frameset XHTML 1.0 and HTML 4.01 doctypes
 - Figure out how to simultaneously set %CSS.Trusted and %HTML.Trusted (?)

5.2 release [Error'ed]
 # Error logging for filtering/cleanup procedures
 # Additional support for poorly written HTML
    - Microsoft Word HTML cleaning (i.e. MsoNormal, but research essential!)
    - Friendly strict handling of <address> (block -> <br>)
 - XSS-attempt detection--certain errors are flagged XSS-like
 - Append something to duplicate IDs so they're still usable (impl. note: the
   dupe detector would also need to detect the suffix as well)

6.0 release [Beyond HTML]
 # Legit token based CSS parsing (will require revamping almost every
   AttrDef class). Probably will use CSSTidy
 # More control over allowed CSS properties using a modularization
 # IRI support (this includes IDN)
 - Standardize token armor for all areas of processing

7.0 release [To XML and Beyond]
 - Extended HTML capabilities based on namespacing and tag transforms (COMPLEX)
    - Hooks for adding custom processors to custom namespaced tags and
      attributes, offer default implementation
    - Lots of documentation and samples

Ongoing
 - More refactoring to take advantage of PHP5's facilities
 - Refactor unit tests into lots of test methods
 - Plugins for major CMSes (COMPLEX)
    - phpBB
    - Also, a FAQ for extension writers with HTML Purifier

AutoFormat
 - Smileys
 - Syntax highlighting (with GeSHi) with <pre> and possibly <?php
 - Look at http://drupal.org/project/Modules/category/63 for ideas

Neat feature related
 ! Support exporting configuration, so users can easily tweak settings
   in the demo, and then copy-paste into their own setup
 - Advanced URI filtering schemes (see docs/proposal-new-directives.txt)
 - Allow scoped="scoped" attribute in <style> tags; may be troublesome
   because regular CSS has no way of uniquely identifying nodes, so we'd
   have to generate IDs
 - Explain how to use HTML Purifier in non-PHP languages / create
   a simple command line stub (or complicated?)
 - Fixes for Firefox's inability to handle COL alignment props (Bug 915)
 - Automatically add non-breaking spaces to empty table cells when
   empty-cells:show is applied to have compatibility with Internet Explorer
 - Table of Contents generation (XHTML Compiler might be reusable). May also
   be out-of-band information.
 - Full set of color keywords. Also, a way to add onto them without
   finalizing the configuration object.
 - Write a var_export and memcached DefinitionCache - Denis
 - Built-in support for target="_blank" on all external links
 - Convert RTL/LTR override characters to <bdo> tags, or vice versa on demand.
   Also, enable disabling of directionality
 ? Externalize inline CSS to promote clean HTML, proposed by Sander Tekelenburg
 ? Remove redundant tags, ex. <u><u>Underlined</u></u>. Implementation notes:
    1. Analyzing which tags to remove duplicants
    2. Ensure attributes are merged into the parent tag
    3. Extend the tag exclusion system to specify whether or not the
    contents should be dropped or not (currently, there's code that could do
    something like this if it didn't drop the inner text too.)
 ? Make AutoParagraph also support paragraph-izing double <br> tags, and not
   just double newlines.  This is kind of tough to do in the current framework,
   though, and might be reasonably approximated by search replacing double <br>s
   with newlines before running it through HTML Purifier.

Maintenance related (slightly boring)
 # CHMOD install script for PEAR installs
 ! Factor out command line parser into its own class, and unit test it
 - Reduce size of internal data-structures (esp. HTMLDefinition)
 - Allow merging configurations.  Thus,
        a -> b -> default
        c -> d -> default
   becomes
        a -> b -> c -> d -> default
   Maybe allow more fine-grained tuning of this behavior. Alternatively,
   encourage people to use short plist depths before building them up.
 - Time PHPT tests

ChildDef related (very boring)
 - Abstract ChildDef_BlockQuote to work with all elements that only
   allow blocks in them, required or optional
 - Implement lenient <ruby> child validation

Wontfix
 - Non-lossy smart alternate character encoding transformations (unless
   patch provided)
 - Pretty-printing HTML: users can use Tidy on the output on entire page
 - Native content compression, whitespace stripping: use gzip if this is
   really important

    vim: et sw=4 sts=4