Check in the libxslt roll script.

third_party/libxslt/TODO

-                  ********
-                 *        *
-                 *  TODO  *
-                 *        *
-                  ********
-
-Lifetime of result Value Tree, make sure we keep pointers. Exslt
-handling of node set values is especially nasty in this respect,
-lots of potential mem leaks...
-
-Pattern tester:
-  -> try to optimize for ID scan and tests.
-
-Pattern scanner:
-  -> add error checks on all returns
-
-Sorting:
-  -> add lang and case-order
-  -> add foreign sorting functions (interfaces ?).
-
-                  ********
-                 *        *
-                 *  DONE  *
-                 *        *
-                  ********
-
-Design:
-  - should transforms for a given stylesheet be thread clean,
-    -> the precompilation now occur only at stylesheet processing
-       time (except the binding for named templates and extension
-       functions which need to be computed once at run-time).
-       Multiple threads should be able to reuse the same stylesheet
-       now.
-
-Embedding Stylesheets:
-  - example in 2.7 would force to validate, we do it by default now
-
-ID and Key support:
-  -> Done
-
-Extra functions:
-  -> document() should not be a problem since Result Tree Fragments are
-     implemented
-  => Done
-
-Templates:
-  -> check the built-in template rule for attributes
-  -> make sure @xxx matches are applied
-
-Contextual error reporting:
-  -> provide a couple of functions providing context analysis, not urgent
-
-Validity:
-  -> should we add validation by default ? Make this an option
-  -> redirrect validity errors
-  => done added a special parsing mode
-
-Import:
-  -> parse them
-  -> provide functions to circulate in the import tree of stylesheets
-  -> make sure we use the cascade wherever it's needed
-
-Extra functions:
-  -> make a separate module.
-  => done functions.[ch]
-
-Support Attribute value templates:
-  -> starts to be urgent. Design it in flexible ways but try to optimize
-     to handle most of it at the stylesheet parse time ...
-  => Done for the most part need to check all attributes in XSLT constructs
-     using them and use the dedicated readin function.
-  -> optimization by checking their existence at stylesheet parse time.
-  => done when building the preproc function
-
-Sorting:
-  -> add support for imbricated sorts
-  => done but not well tested.
-
-Separate util module:
-  -> macros, config, verbosity ?
-  => xsltutils.[ch]
-
-Support for disable-output-escaping="yes":
-  -> looks problematic, libxml has no support for anything like this,
-     and unless adding a new node type :-( or tweaking text node and
-     output routines this is gonna be messy ... must be handled at libxml
-     level.
-  => Done with a trick, text node name is different, requires > 2.2.11
-
-Pattern scanner:
-  -> compute priority
-  -> handle unions
-  -> support for mode
-  => done
-
-Pattern tester:
-  -> also put fast lookup for "text()", "comment()", "node()"
-     based patterns lists.
-  => done
-
-Support Attribute value templates:
-  -> namespace support for attribute value templates is not done, need
-     a small API redesign
-
-Doc:
-  - put a page at http://xmlsoft.org/XSLT/
-  - generate/transform the DocBook to HTML
-  - add HTML to package
-  - manpage and doc for xsltproc
-
-
-Error handling:
-  -> check the version stuff, design a separate module for error interfacing
-     and default handling, parsing vs. runtime, fatal / compat / warning,
-     and lack of optionnal features.
-  -> reports context
-
-ID and Key support:
-  -> done but namespace support in keys is not defined
-  -> make sure keys are recomputed on new document input
-
-Profiler:
-  -> added looks good enough
-  -> autocorrection of initial calibration loop