0.8.4 2019-06-27
      - Various fidelity and robustness improvements, along with more test cases
      - Improved source locators for debugging and other purposes.
      - More flexible backmatter, allowing multiple bibliographies, indices, etc.
      - Improvements to math lexer
      - Support for Chocolatey package manager
      - More generalized support for lists of (contents, tables, figures, etc)
      - New bindings: aipcheck.tex, aipproc.cls, bibunits.sty, blindtext.sty,
        calrsfs.sty, chapterbib.sty, chngcntr.sty, csquotes.def, csquotes.sty,
        fancyheadings.sty, footnote.sty, mathdots.sty, nopageno.sty, PoS.cls,
        quantumarticle.cls, remreset.sty, sidecap.sty, subfiles.(cls|sty),
        tcilatex.tex, thm-restate.sty, twoopt.sty, underscore.sty, undertilde.sty,
        varioref.sty, varwidth. styvmargin.sty

      Thanks: Deyan Ginev, Tom Wiesing, Ben Firsch, Michael Kohlhase,
        adhilton, angerhang, Takuto Asakura, asmaier, Luis Berlioz,
        Alex Bobrikovich, Christoph Hauert, giovanni111, goska,
        Andre Greiner-Petter, Sam Kacer, Bernard Kleine, Jakub Kukul,
        Raffaele Mancuso, molke-productions, Alexander Nozic,
        Hans Olsson, raspatan, Red-Quill, Moritz Schubotz, seuliang,
        Scott Walter, Frederic Wang, Dietmar Winkler, Sigfried Zoetzsche

0.8.3 2018-07-30
      - Way too long between releases...
      - Too many bug fixes, fidelity improvements and enhancements to list.
      - (nominal) support for TEI format 
      - New Bindings: aastex61.cls, aastex6.cls, accents.sty, acmart.cls,
        aecompl.sty, algorithm2e.sty, appendix.sty, cases.sty,
        cleveref.sty, csquotes.sty, empheq.sty, endnotes.sty, enumitem.sty,
        etex.sty, etoolbox.sty, flowchart.sty, flushend.sty,
        french.ldf, gensymb.sty, icml2018.sty (& others), ieeeconf.cls, IEEEtran.cls,
        jheppub.sty, lastpage.sty, lipsum.sty, makecell.sty,
        microtype.sty, mleftright.sty, mnras.cls, moderncv.cls,
        newfloat.sty, newtxmath.sty, newtxtext.sty, pdfpages.sty,
        pgfmath.code.tex, pzd.fontmap, ragged2e.sty, revtex4-1.cls,
        sectsty.sty, siunitx.sty, siunitx.sty.bak, stfloats.sty,
        subcaption.sty, svjour3.cls, tablefootnote.sty, textcase.sty,
        thmtools.sty, titlesec.sty, titling.sty, todonotes.sty,
        tracefnt.sty, turing.sty, utf8x.def, wiki.sty, xkeyval.sty,
      - 2 Experimental features:
          * xpath source locators: for locating the origin of the XML in the TeX sources.
          * math lexemes: textual encoding of mathematics for data mining studies.
          Please consult the mailing list for more information.

     Thanks: Deyan Ginev, Tom Wiesing, Michael Kohlhase;
       And thanks for reports from Siegfriedzoetzche, Grief, MikePetullo, 
       matteosecli, goska, teepeemm, Fred Wang, Ben Firsch, Joe Cornelli,
       Bernard Kleine, eshellman, swalter62, asmaier, charlyms, senthilmm,
       cooperstevenson, Andre G-P, blahah, nthiery, webbery, evoludolab, ilovezfs, manwar,
       Moritz Schubotz, wspr, skellat, giovanni111, Thanathan-k, HansOlsson,
       angerhang, Stefan Becuwe, IvanRadianStd, daroloson, sumandoc, gpetty,
       SabinSnorlax, DeathByRegex, QiangF, aimsky, dominique-unruh, SephidehAlassi,
       casio7131, davpoole, LuiSlacker, kaagithub, ashinkarov, ianhbell, jolutti,
       dragonxtek & laszlokv.

0.8.2 2016-07-10
     - Numerous bug, robustness, fidelity fixes;
     - fixes to amsmath, amsthm, keyvals, listings, tikz/pgf, color, xcolor, supertabular packages;
     - fixes to bibliography, OpenMath, MathML, svg generation;
     - accomodate sizing of math, better position adjustments;
     - JATS/NLM format;
     - Colorized error logging;
     - New bindings: aas_macros.sty, algc.sty, algcompatible.sty, algmatlab.sty, algorithm.sty,
       algorithmic.sty, algorithmicx.sty, algpascal.sty, algpseudocode.sty, amsgen.sty,
       apjfonts.sty, attachfile.sty, authblk.sty, balance.sty, braket.sty, breakurl.sty,
       colordvi.sty, deluxtable.sty, emulateapj5.sty, esint.sty, fix-cm.sty, flafter.sty,
       grffile.sty, ifluatex.sty, import.sty, lmodern.sty, marvosym.sty, mathbbol.sty,
       mathtools.sty, nameref.sty, newlfont.sty, parskip.sty, pdfsync.sty, psfig.tex,
       pslatex.sty, rotate.sty, srcltx.sty, showkeys.sty, slashed.sty, soul.sty, subfig.sty,
       subfloat.sty, stmaryrd.sty, svg.sty, t1enc.sty, tabulary.sty, threeparttable.sty,
       transparent.sty, wasysym.sty.

    Thanks: Deyan Ginev, Kim Philipp Jablonski, Lukas Kohlhase, Michael Kohlhase,
       Viacheslav Zholudev;
       And angerhang, Stefan Becuwe, Stephane Binarez, casio7131, Howard Cohl, Joe Cornelli,
       Boyd Duffee, Giovanni, goska, Bernhard Kleine, neos21, Grant Petty, Bob Plantz,
       Paolo Prandoni, Tim Prescott, Matteo Seclì, M. Senthilkumar, Moritz Schubotz, Raniere Silva,
       Cooper Stevenson, Tom Tczyczko, Scott Walter, Frederic Wang, Simon Winter, Tom Wiesing,

0.8.1 2015-02-17
     - Generate "span soup" for inline blocks, when otherwise would produce invalid HTML5.
     - Can use svg as math ouput format.
     - listings from the listings package provides link to downloadable raw data 
     - implemented low-level TeX input/ouptut primitives
     - more consistent handling of math spacing, delimiter sizes,
       equation positioning and alignment; improved linebreaking layout
     - parallel math markup w/cross-referencing
     - math grammar enhancements
     - language support (xml:lang & (slightly) internationalized number recognition)
     - font size handling more consistent with CSS
     - New fallback for unknown classes "OmniBus.cls" mimics common frontmatter markup.
     - improved package option handling
     - improved documentation
     - improved portability (Windows, Perl versions)
     - various other bugs & fidelity improvements
     - New bindings:
       a0poster.cls, a0size.sty, braket.sty, crop.sty, epsf.tex, ellipsis.sty,
       fancyhdr.sty, floatpag.sty, ifthen.sty, lineno.sty, listingsutf8.sty,
       pgfplots.sty, relsize.sty, type1cm.sty, xargs.sty

    Thanks: Deyan Ginev, Michael Brade, Joe Cornelli, dpantele, Giovanni,
       Kim Phillip Jablonski, Bernhard Kleine, Lukas Kohlhase, Michael Kohlhase,
       Silviu Opreah, Antonio Sanchez, Moritz Schubotz, Rainere Silva, Frederic Wang & Others.

0.8.0 2014-05-05
     - Too many changes to enumerate...
     - Generates HTML5, ePub
     - new Color objects allowing better & more accurate, and extensible
       color models; binding for the xcolor package.
     - RDFa support; Thanks Christoph Lange
     - consistent error reporting, in both conversion & post-processing
       that supports automated build systems; Thanks Deyan Ginev, Heinrich Stammerjohanns
     - uses dvipng (IF available) for converting tex code to images;
       MUCH faster than going via eps.
       HOWEVER, it wont handle embedded postscript,
       so we only use it for converting math to images.
     - Reorganize various non-perl data files into a (hopefully) more manageable arrangement,
       and with more consistent naming:
       lib/LaTeXML/resources contains various resources for running the program
       * DTD, RelaxNG for holding various schema
       * XSLT, CSS, javascript for holding various styling & script resources
     - Mechanism for bindings to request resources (css, javascript) that will
       be included in generated output if appropriate (eg. if html)
     - more consistent naming schemes
       * for classes typically start ltx_<element>, ltx_font_<font>, ltx_role_<role> etc
       * LaTeXML.css for core.css
       * package specific css files: ltx-<package>.css
       * special purpose css files: LaTeXML-<feature>.css
     - XSLT uses <xsl:element> exclusively so same modules cope with both xhtml & html5;
       More easily extended and overridden by author customizations.

0.7.9xx 2012-08-01
     - some slight efficiency improvements
       (reports Deyan Ginev, Joe Corneli)
     - fixed bug in Post reading from STDIN (thanks Josh Bialkowski)
     - various robustness
     - I/O reorganization, which includes
       * an extensible pseudo-protocol for sources of TeX and data,
       	 including file:<filename>, string:<explicit-string-data>
       * Migration of control of I/O to functions defined in Package.pm
       	 InputFile, InputDefinitions, etc
       * Obsolete $GULLET->input, $GULLET->expandTokens
       * Tokens lists are now immutable (do NOT implement Mouth API)

0.7.9xx 2012-07-01
     - more consistent handling of math spacing;
       treat \quad and wider as punctuation
     - better support for adding ID to all elements
     - parallel math markup will establish node-to-node cross-referencing
       Particularly pmml+cmml uses id & xref to connect related pmml & cmml nodes.
     - generate MathML mrow instead of mfenced.
     - experimentally, do NOT define non-control sequences as mathactive
       (in order to assign XMTok attributes), but store such properties separately.
     - Have MathBox store the XMTok properties of simple symbols;
       create Primitives for simple symbols, rather than Constructors.
     - switched comparison files from dvi to pdf; more portable

0.7.9xx 2012-06-02
     - MathFork updates
     - use LaTeXML.catalog more consistently;
       define URN's for RelaxNG modules and XSLT modules

0.7.9xx 2011-06-01
     - various changes and small completions to LaTeX that help to
       process raw TeX style & definition files from the LaTeX distribution.
     - Support LaTeX's input encoding mechanism by finding, reading and 
       implementing the encoding definitions
     - Support LaTeX's fontencoding mechanism and TeX's \char, etal,
       by implementing FontMap's that map input codepoints to unicode.
       Also read in font encoding definitions from the LaTeX distribution
     - Support babel by implementing core support and reading individual
       language definitions from the LaTeX distribution.

0.7.x 2011-03-16
     - Bindings for floatfig, floatflt, JHEP,JHEP2,JHEP3
     - improved model, attributes and conversion for floating objects
     - Have \caption increment the appropriate counter, rather than figure, table, etc
       This avoids many problems with subfigures, and so forth.

0.7.x 2011-02-28
     - Bindings for llncs.cls, rsfs.sty, multicol.sty, enumerate.sty, xspace.sty,
       caption.sty, subfigure.sty, upgreek.sty
     - Initial support for AmSTeX (AmSTeX.pool)
     - Schema changes: new ltx:inline-para, fixed ltx:classification, positionable ltx:p,
       ltx:bib-data holds original bibentry
     - improved parallel markup
     - support for Unicode Plane 1 as alternative to mathvariant
     - various improved code, implenting \fracwithdelims, \DeclareMathAlphabet,
       \obeylines, \hypersetup, \centering, \raggedright, \raggedleft, \let,
       \rotatebox, \reflectbox, \scalebox, \qopname, \Sb, \Sp, \afterassignment,
       \sidecaption, \@vec, \minCDarrowwidth, \beginsection, \proclaim,
       \AtBeginDocument, \AtEndDocument,
       rules within tables
     - new Box() function
     - \iflatexml in latexml.sty
       namespaced attributes
     - improved table header heuristics
     - better handling, and distinguishing broup, begingroup
     - better ID handling; many elements get ID's, equations always do.
     - better MathML conversion, even when parsing fails.
     - support for --icon provides a favicon
     - more test cases (but not enough)
     - More careful distinction between ToString vs Stringify vs UnTeX
       Stringify is for debugging, ToString should return string form, NOT neccesarily for getting TeX
       many changes to definitions and usage of these.
     - Many constructors converted to Primitives to support turning accents or Unicode conversions
       directly to boxes, not requiring constructors; this allows them to be unambiguosly converted
       to Strings (using ToString) and thus can be put in attributes.
     - avoid introducing doubled slashes in pathnames, so that (eventually) might
       use the TeX standard interpretation that treats them as recursive wildcard.
       (that is, expandable directories)
     - INCOMPATIBLE: Localization; remove assumed names & symbols from formatting titles, sections, etc.
       Let style files determine prefixes or formatting of reference numbers.
       Put these within the ltx:title (eg.) and wrapped in ltx:tag.
       This is incompatible as different attributes are used, and ltx:tag is used to contain the reference number.
     - more flexible TOC creation, support for list-of-figures and similar.
     - Fuller support for seealso, heuristics to try to connect the terms to actual entries.
     - Safer encode/decode of objects being stored in the ObjectDB.

0.7.x 2009-06-19 (ref 983)
     - fixed typos in aa_support & sv
     - added wrapfig.sty, llncs.cls, rsfs.sty, multicol.sty, enumerate.sty,
       xspace.sty, caption.sty, subfigure.sty
     - initial implementation of AMSTeX (amstex.tex & amsppt.sty)
     - enabled better CSS styling via XSLT.
      Thanks Lee Worden

0.7.0  2009-06-16 (rev 964)
     - Release 0.7.0

0.6.x  2009-06-15 (rev 959)
     - mostly complete listings.sty
     - new classes sv, svmult
     - better compatibility-mode support
     - made indices case sensitive and fixed sorting order correspondingly
     - many small fixes throughout
     - improved manual
     - Thanks to Deyan Ginev, Michael Kohlhase, Lee Worden for reports/patches.

0.6.x  2009-05-07 (rev 899)
     - added various macros to existing packages.
     - new packages/classes: elsart, iopart, iopams, mn, mn2e
     - added less obnoxious Info message for things not as severe as Warning.
     - changes to centering, flush commands; get rid of centering element,
       try to use class and css for same effect.
     - fix typos in latexmlpost (Thanks Jason Blevins)
     - reduced load time, primarily for latexmlmath
0.6.x  2009-03-15 (rev. 824)
     - efficiency improvements
     - documentation and error message improvements
     - various minor fixes, extra macros, 
     - heuristics to handle misused environments
     - support to "lock" macros from being inappropriately
       redefined within TeX documents
     - new packages: paralist, eurosym

0.6.x  2009-01-14 (rev. 740)
       Preparation for release 0.7.0
     - New packages:
       revtex and revtex4 classes and styles
       (Thanks Deyan Ginev <d.ginev@jacobs-university.de>
       and Catalin David <c.david@jacobs-university.de>)
       aas styles, amscd, lscape
     - latexml now processes BibTeX files
     - latexmlmath; new command for creating images or MathML
       for individual math expressions.
     - Improvements to MathGrammar for bra-ket notations,
       assignment, and other presentation markup like odd sub/superscripts
     - Improved option handling
     - More consistent math meaning "ontology"
     - Some documentation improvements
     - Rearranged Makefile & dependencies to port to Centos, MacOS
     - Many additional macros, robustness fixes
     - Clarified progress and error messages.

0.6.0  2008-04-09 (rev 485)
     - Reorganized site and manual building.
       Unfortunately, the manual isn't as extensively rewritten
       as I'd like.
     - reorganized build & installation.
       Should be able to generate rpm's now.

0.5.x  2008-03-18
     - Added implementation for supertabular and longtable
       (the latter isn't perfect; it produces empty rows after
     - random minor fixes
0.5.x  2008-03-03
     - Added support for class & package options
       with new exported functions in LaTeXML::Package
       (DeclareOption, PassOptions, ProcessOptions, ExecuteOptions).
     - Modular XSLT: individual XSLT files correspond
       to the Schema modules and are assembled into
       composite html and xhtml versions.
     - Option to latexmlpost to suppress section numbering.
     - new package ifpdf
     - more consistent date formatting
     - switch to using id in html instead of <a name=...
     - make "\ " generate no-break-space (more portable)
     - random other minor fixes.
     - documentation improvement

0.5.x  2008-01-16
     - Major check-in of accumulated changes
     - Added a MathFork/MathBranch construct to represent
       math with alternative structures, especially alignment.
       This allows representing eg. eqnarray as a group
       of equations. Each equation has a MathFork where
       the main branch is the complete equation (hopefully semantic),
       and the branch contains 1 or more tr each containing 3 td.
       Thus both the meaning and the structured alignment can
       be preserved.
     - Change to use xml:id instead of id.
     - Changed common attribute label to labels (plural)
       along with support in postprocessing, etc.
       This is since LaTeX in principle allows multiple labels
       to attach to the same place.
       The labels are also stored as "LABEL:text" instead of just "text",
       so that the same mechanism can be leveraged for other
       crossreferencing later (eg. acronyms,etc).
     - Some generic exported support functions to ease the above
       CleanLabel, CleanIndexKey, CleanBibKey, CleanURL, GenerateID

0.5.x  2007-12-01
     - Conversion to RelaxNG Schema.
       (currently only recognizes XML format)
       Use RelaxNGSchema($schema) to declare the schema to use.
     - Continued refinement and simplification of model.
     - Use xml:id instead of id.

0.5.x  2007-10-10
     - Began revising documentation.
       Added documentation of DTD; this led to refining
       and making the model and attributes more consistent.
     - Yet another alignment/math fix.
     - more complete hyperref.
0.5.99  2007-09-26

0.5.x  2007-09-20
     - Added Conditionals to deal with encoding changes in
       XML::LibXML; LaTeXML should work with 1.61 or newer.
     - Better detection of latex (vs. plain tex)
     - Modularized CSS; allows multiple css stylesheets,
       separated supplied stylesheet into several options
       core, navbar-left, navbar-right, theme-blue, amsart
     - Lots of more fixes due to arXMLiv project testing.
     - Initial implementation of ams classes
       (amsart,amsproc,amsbook, gen-j-l,gen-m-l,gen-p-l),
       and upref.
     - Improved interaction between alignments, text and math modes.
     - improved manual.

0.5.x  2007-09-05
     - registered missing handlers in Post::Scan (thanks Harald Soleng)

0.5.9  2007-06-20
     - Hmmm, I haven't been keeping the changelog uptodate.
       Things _have_ been happening...
     - Lots of changes, including:
     - Improvements to handling of various kinds of Alignments,
       including multrow, rowspan, tabularx packages
     - enhancements to postprocessing to support splitting single
       documents into multiple output pages, as well as crossreferencing
       between multiple source documents.
     - Improvements to manual.
     - Updated for XML::LibXML version 1.62
       No longer explicitly requires XPathContext.
     - Experimental math linebreaker 
       (see --linelength option to latexmlpost)
     - Various minor correctness & robustness patches.
       [Thanks to the DLMF and ArXMLiv projects for testing and patches]

0.5.1  2006-04-27

0.5.0  2006-03-22

0.5.x  2006-08-03
     - Changes to how frontmatter is handled, and represented.
       Frontmatter gets accumulated and added to <document>
       when it is absorbed, whether or not there's a \maketitle.
       <author> => <creator role='author'><personname>...
       <affliation> => <contact role='affiliation'>... inside of <creator>
       Similarly for <email>
       This construct gives more flexibility for representing
       editors, translators, reviewers, etc,
       And also for various kinds of info about them (address, etc).

0.5.x  2006-07-24
     - Fixed a sneaky bug in \def parameters, and gullet->readMatch
       where TeX collapses multiple spaces! Thanks Ioan!

0.5.x  2006-07-23
     - Fixed the TeX font command to recognize more fonts.
     - Redefined sectioning commands to go via \@startsection,
       so more author customizations will still work.
     - added report & book classes. 
     - Other random (minimal) packages: eulervm, yfonts, a4wide

0.5.x  2006-07-22
     - Dealing with some font issues, and adding minimal
       implementations: fixltx2e, textcomp, exscale
       mathptmx, mathpazo, charter, utopia, chancery, helvet, avant,
       courier, bookman, newcent, times, palatino, mathptm, mathpple,
       latexsym, beton, euler, ccfonts, concmath, cmbright,
       luximono, txfonts, pxfonts, fourier,

       And a start at pifont...

0.5.x  2006-07-22
     - Went through the TeX book and implemented bunches of
       Appendix B (plain) --- still not complete, but better.
0.5.x  2006-07-21
     - Added implementation of amsthm; thanks Ioan Sucan.
       along with tests, implementation of LaTeX's newtheorem,
       and a few other tweaks to make it cleaner.

0.5.x  2006-07-17
     - Revised handling of sub/superscripts:
       Abandoned the SUBSUPERSCRIPT combinination of sub+super
       generated by the parser; now sub and superscripts
       (including prescripts) are nested in the (seemingly)
       given order; ams's sideset is done similarly.
       This imho gives a more sensible semantic structure.

       The stackscripts attribute was renamed to scriptpos
       and generalized to track the position (pre, mid, post)
       and the bracket nesting level where the script was created.
       This allows the presentation mathml module to determine
       sensible nesting and positioning so that it can
       determine under/over vs sub/sup as well as subsuper combinations,
       and mprescripts according to the given tree.

       A few other enhancements were made to mathml generation, as well.

     - Added class attribute to <text> and defined new (generic)
       block element with class attribute.  These can serve
       as fallback elements for testing purposes.

0.5.x  2006-06-24
     - With some trepidations...
       Converted math arrays to use XMArray/XMRow/XMCell
       so there's more sharing with tabular, and it can
       handle the lines, headings, etc.
       Instead of the more abstract XMApp/[role=ARRAY], etc.

     - Added a meaning attribute to math.
       The idea is:
         role : grammar or presentation info.
	 name : a name for the token, probably from the cs,
	        but not necessarily completely semantic.
	 meaning: a hopefully semantic enough name for the token.	
	 	This would be used for content conversions.

     - Recent runs show the Perl function bound as an XPath
       extension are very costly.  Recoded the font match
       handling to avoid the perl function, using a set of
       contains calls.  Vastly faster!

0.5.x  2006-06-16
     - Fixed typos in DTD parsing
       Thanks Ioan Sucan <i.sucan@iu-bremen.de>
     - Reworked pathname_find to be a little clearer about
       seaching for files that come with the installation
       (but can be overridden by SEARCHPATHS).
       Added pathname_findall which finds all matching files,
       and used it so that all available catalogs are loaded,
       in particular any in the SEARCHPATHS.

       In the process noted a bug that if the environment variable
       XML_DEBUG_CATALOG is set, XML::LibXML bombs
       (seeking advice from mailing list).
     - Fix to \varintjlim (Ioan Sucan)_

0.5.x  2006-06-01
     - Simplifications to PMML; use roles more consistently
       (which means there are some roles that never appear in Grammar,
       but which represent presentational structure).

0.5.x  2006-05-20
     - Sorted out (hopefully finally) the Unicode nonsense
       w/chars in 80--FF; Perl is dumm; you really need to use pack;
       exported handy UTF(hex) from LaTeXML::Package to help.
       I had been (over)using Unicode::Normalize::NFC to patchup
       after the fact, but this has screwy effects
       (translates \langle to something in chinese block !?!)

0.5.x  2006-04-28
     - A number of initializations, typos, missing \and fixed.
       Thanks Christopher B. Hamlin <chamlin@aip.org>
     - Fix to eqnarray numbering bug.
       Thanks Eduardo Tabacman <eduardot@dessci.com>

0.5.1  2006-04-27
     - Release 0.5.1

0.5.x  2006-04-24
     - Refined the math grammar, added some test cases.

0.5.x  2006-04-09
     - Fixed up some math grammatical quirks, 
       redefined default role for :, \mid

     - Corrected handling of \left.,\right.

     - Fixed up Presentation MathML handling of unsuccessfully
       parsed math.

0.5.x  2006-03-28
     - straightened out some namespace mismatches in DTD's

     - Updated documentation to reflect current commands & API's

0.5.0  2006-03-22
     - Release 0.4.0

0.4.x  2006-03-18
     - Defined \LXDeclMath for "Math Declarations" in latexml package.
       These declarations can be embedded in the TeX Source.
       Basically these define patterns to match to scoped portions
       of the generated document tree (using Rewrite rules),
       and add declarative attributes to support the math parsing.

0.4.x  2006-03-01
     - Modularized the DTD, along with lots of cleanup.

0.4.x  2006-02-09
     - Cleaned up Makefile.pm: made ImageMagick optional
       (tho' without any clear failure mode when used, yet);
       Safer XSL style file generation.
       Should be close to able to install on windows.
     - Wrote some Test::Builder support code, reorganized
       the test suite, and started adding new tests.

0.4.x  2006-01-27
     - Essentially backtracking on changes for 0.2.0,
       I'm concerned about the number of globals and exports,
       and formalizing extensible readers for control sequences.
       Thus, yet another incompatible change in the parameters
       to code blocks defining macros, primitives and constructors.
       constructor($document,@args, %properties)
       For Tag code blocks:
0.4.x  2006-01-22
     - Yet another rewrite of tabular processing.
       Now allows @-expressions.

     - New Tag option: whitspaceTrim; this trims
       leading & trailing whitespace from the direct
       text content of these tags.

     - Significant namespace cleanup. 
       There are 2 prefix/namespace mappings.
       (1) the one used in code (eg.ltxml) for constructors,etc.
       (2) the one used to interpret the DocType (dtd).
       Constructors should always specify a namespace prefix for names,
       unless they are in the null namespace (NOT default namespace).
       In fact, there is no longer a notion of default namespace, as such,
       and RegisterNamespace no longer takes that 3rd argument.
       DocType takes as extra args prefix=>namespaceURI mappings
       to be used in interpretting the DTD, and the resulting document
       will be constructed using those same prefixes.

0.4.1  2006-01-09
     - Relase 0.4.1

0.4.x  2006-01-09
     - Experimental tabular transformation.
       More faithful reproduction of latex tabular in html,
       via CSS.  Heuristics for table headers.

0.4.x  2005-12-15
     - Fixed some namespace usages, so that
       constructors containing "<foo:bar>..." will
       work, provided foo was registered (RegisterNamespace)
       Added a too-simple testcase.

     - If "-" is used on latexml|latexmlpost command line,
       they reads the TeX|XML, respectively, from STDIN.

0.4.x  2005-09-27
      - Added missing test result file keyval.xml
      - Patch to postprocessor: only mung LaTeXML's DTDs.

0.4.0  2005-09-26
      - Release 0.4.0

0.3.x  2005-09-xx
      - Hopefully harmless simplifications in DTD regarding text.
        Combined the <textstyle> and <hbox> with the <text> element.
	Changed model for XMath to only allow <text> rather
  	than %Simple.class; and made <text> auto-open so that
	all non-obviously math things will be wrapped in <text>.
      - More DTD (and generation) modification to better support
        a logical versus physical paragraph  structure.
	  <paragraph> is a possibly numbered & labeled element
	     generated by the \paragraph command.
	  <para> represents a logical paragraph; It contains block
	     elements, in particular it can contain sequences of
	     <p> and <equation> that represent a logical paragraph.
	     It can have a refnum and label, although it does
	     NOT get the label assigned by \label.
	  <p> represents a physical paragraph --- a block of text.
      - implemented various missing plain macros.

0.3.x  2005-08-xx
      - Added support for LaTeX's picture environment and
        pstricks (along with pst-node).
	A Postprocessing module converts the resulting XML into SVG!
	Thanks very much to Ioan Sucan!!

      - Reverted the attribute xml:id to id on math nodes (XM*),
        since XMRef's idref attribute should only refer to XMath nodes.
	This also avoids conflict with other uses of xml:id that
 	a developer might need to make.

0.3.x  2005-07-xx
       - INCOMPATIBLE changes.
         In order to make constructors more flexible, I'm incorporating
	 the possibility to invoke arbitrary functions within constructors.
	 So, something like:
	 <foo bar='&Func(#1,'a')'/>
	 would set the attribute bar on the element foo to
	 be the result of applying the function Func to the first argument,
	 and the string 'a'.
	 Note that even w/o args,parens are required (so maybe entities still work).

	 The Incompatibility is due to absorbing previous ad-hoc functionality:
	   ?IfMath is now ?#isMath
	   (since isMath is an internal property of all Whatsits)
           Accessing bound values is
                   VALUE('name') => &LookupValue('name')

         A new constructor pattern triggered by '%' is defined
	 such that %value adds a _set_ of attributes to an element,
         where value would be something like #1, #foo, &KeyVals(#1)
 	 such that the value returns a hash reference.
       - The above also allows KeyVals to be better encapsulated
         and pulled out from the core of LaTeXML.
	 The functionality of keyvals will now only be available if
	 you \usepackage{keyvals} or RequirePackage('keyval')
	 The Parameter specification for KeyVals is now of the form;
	 or OptionalKeyVals:name
	 where name is the name of the keyval set.  The first expects
	 keyvals wrapped in the usual {} pair, whereas the second
	 expects optional args wrapped in [], if present.

	 Furthermore, the constructor patterns have been redefined in a more
	 general framework:
	 Accessing KeyVal data:
   	    #1{key} is now &KeyVal(#1,'key')
	 Accessing all keys would now be
	    <foo %&KeyVals(#1)/>
	 instead of simply <foo #1/>
      -  Similar change to argument type semiverb.  Instead of 
         you should now write
         this reads an {} delimited argument, but with most catcodes turned off.

      - Conditional patterns in constructors now properly balance the delimiting
        parentheses.  Thus conditionals can now be nested, and function calls
	used within the patterns should work.

      - Revamped and regularized Parameter specs, making them more extensible.
        {KeyVals:foo} => RequiredKeyVals:foo
        [KeyVals:foo] => OptionalKeyVals:foo
        Flag:*       => OptionalKeyword:*

0.3.x  2005-06-xx
      - Fixed Subtle bug with conditionals and \else; 
         Special case: \else doesn't get expanded while
 	 the conditional test is being expanded! (See TeX: The Program)
        (Thanks Kohlhase for pointing it out)
      - Fixed \underline, \overrightarrow, \overleftarrow to work in textmode.

0.3.2  2005-05-16
      - Cleanup of LaTeXML.dtd; to be make sure all elements get
        appropriate attributes defined (should validate mostly).
      - More tweaks & tuning for more understandable error messages.
      - implemented \raggedright, \raisebox,\buildrel,\stackrel
      - New functionality in Constructor patterns:
        VALUE('keyword') can be used where a value is expected
        to lookup a value in the state. Also allows args & such,
        So \ref ends up defined as <ref labelref='#1'>VALUE('LABEL@#1')</ref>
        This also means that the constructors \@VALUE and \REF are no longer
        needed, so they're removed.

0.3.1  2005-05-10
      - Improved mismatched environment reporting.
      - More faithful implementation of verbatim & comment environments
      	with fixes to mouth's readRawLines.
      - Fix in Stringify for XML nodes; apparently a documentation bug
        in XML::LibXML::Namespace ? (it doesn't implement getValue)
      - imcremental improvements in latexmlfind

0.3.0  2005-05-06
      - Release 0.3.0
      - Some speculative code on handling the picture environment,
        along with pstricks, but not yet settled.
      - More exports from Package for common operations there,
        and hopefully reduce the usage of global $STOMACH, etc.
      - Improved and updated documentation.
      	Still need to document the new Rewrite facilities
	(but would like to make API more concise)

0.2.99x 2005-04-13
      - Allow * flag (ignored) on \newcommand, et.al.
      - Fixed some problems with fake environments
        (ie. \begin{small}...\end{small})

0.2.99 2005-04-07
      - Released as 0.2.99 so the Bremen folks can get
        some work done.  Documentation update is needed
	for 0.3.0 release.

0.2.xx 2005-03-17
      - Bigger changes, increment version.
      - Modified DocType; don't add namespace, use RegisterNamespace
      - Intestine now creates XML::LibXML structures directly.
        Module LaTeXML::Node is removed.
	In fact, Intestine essentially represents the Document itself
	and thus is now renamed LaTeXML::Document.
      - Removed global exported Font() and MathFont()
      - Made more definitions scopable, cleaned up stash & scope
        implementation.  Renamed: methods {de}activateStash
      - Implemented Rewrite rules that act on the constructed document.
        They also allow rules defined in terms of TeX strings
	(tokenized, digested, converted	to document fragments
	and then XPath statements, as needed).
	These rules can be used to effectively declare variable
	or symbol's Grammatical roles.
	Math Parsing is now part of the latexml script
	and removed from latexmlpost.

0.2.3 2005-01-xx
      - Fixed a problem where misplaced egroups could
        inadvertently change the mode. Mode is no longer
	affected by the TeX stack; they must be explicitly
	start/finished (even though they also introduce grouping).

      - Fixed counting of `raw' lines read for
      	"comment"ed environments.  Line numbers for
	errors were getting skewed.

      - Moved sectional attribute declarations inside the
      	%define.structure; block to ease defining extension DTD's

0.2.2 2005-01-11
      - Random minor bug fixes and improvements to error
0.2.1 2005-01-10
      - Bug fixes to stylesheet LaTeXML-xhtml,
      	Thanks Yann Golanski
      - A few rearrangements and renamings to make a simple
        top-level 'digest from string' alternative.
 	[  $latexml->readAndDigestString($string) ]
	Also, renamed the slightly misnamed Stomach methods:
          readAndDigestChunk => digestNextChunk
          readAndDigestBody => digestNextBody
      - A few typos in Stomach fixed
      - Almost complete implementation of the various AMS
        Still need to complete and test the various alignment environments

0.2.0 2004-12-25
      Extensive changes, so incrementing minor version, but
      not robust enough for major version!

      - Added version info to latexml, latexmlpost help output.
        separated --debug and --trace options.
      - Removed mathConstructor option to various DefXXs
      	Use new constructor conditional "?IfMath(..)(...)"

      - POS is an annoying acronym. Role is better and
        upon reflection, doesn't conflict with OpenMath's ussge.
	Hence, partOfSpeech and POS have been replaced by role
	to describe the grammatical role (or `part of speech')
	of tokens to be interpreted during a math parse.

      - Reduced introduction of new `name' attributes for
      	math tokens, especially when they add little value.
	Most greek & math characters are just replaced
	by thier unicode equivalent; In most cases, a name
        is synthesized from the control sequence when needed.
        The intestines will now create an XMTok, if required.
	Also, it will automatically manage the font
	and assign a `cs' attribute to record the macro
	used to create the token.

      - DefSymbol is deprecated (removed in fact)
      - DefMath (new) covers what DefSymbol did,
      	and more: handles the common form for functions
	taking arguments. When the macro takes args and
	the replacement presentation text involves #1,
	it generates an XMDual using the replacement as
	an expansion, but also creates the content form.
        To avoid duplicating the arguments, the XMArg's
        containing the arguments in the content branch
        are marked with an id; in the presentation branch
	<XMRef idref='id..'/> is used.  Corresponding
	code in postprocessor looks up the referred node
	when needed.  XMRef can also be used on it's own:
	see the macros \@XMArg and \@XMRef in TeX.ltxml.

      - Constructors take property arguments which supply
      	properties to the whatsit (which can be CODE evaluated
	at digestion time).  These properties can be used
	in the constructor pattern.

      - Refactoring of Intestine & DOM; most interaction
        with Model is done in Intestine.  Renamed
        DOM to Node and renamed it's subclasses.

      - "In for a dime, in for a dollar":
        Since I'd found it necessary to use global variables
	to access the stomach and intestine from strange places, 
	then I might just USE the darn'ed things!
	Consequently, most places were a $stomach, $intestine
	(or $gullet and $model) were passed around as arguments, 
        no longer do. Now, just use the globals, which the inlines
	As a side effect, the `0-th' argument to CODE implementing
	control sequences is generally the definition (for whatever
	use that might be), or the Whatsit for constructors.

      - Made DOM construction more forgiving by using SalvageError
        when constructing a tree that doesn't conform to DTD.
	The result may not be valid, but continues processing.
	This led to major rewriting & cleanup of error reporting,
	and storing a `locator' on all data objects that record
	where in the source file they were created.
	[Thanks to suggestions from Kevin Smith]
      - Cleanup of math parsing, presentation mathml generation.
      - Added postprocessing module for generating OpenMath.
        It is insufficient, but a starting point.
      - latexml.sty & latexml.ltxml
        A start at providing special purpose macros that make sense
	in LaTeX, and do even more interesting things in LaTeXML.
	Currently, define some silly macros like \XML, \LaTeXML, etc,
	and provide LaTeX bindings for things like DefMath!

      - Sadly, I gave up on "overload".  Nice idea, but 
        for a big package, it's tricky to get right.
	The magic creation of methods can lead to hard-to-find
	performance issues, if you try to do to much with Stringify.
	So, Object doesn't use overload.  To stringify or compare,
	consider the (newly exported functions in Global):
	Stringify($ob), ToString($ob) and Equals($a,$b).

      - added latexml.sty which should get installed in the local branch
        of the standard texmf directories.  Not yet documented, but
	it provides (or will) LaTeX bindings to interesting LaTeXML
	declarations, eg. defining math commands.

0.1.2 2004-09-02
      - Some experiments to reduce namespace redundancy.
      	C14N is too severe, use of $node->addNewChild
  	is non-portable and awkward.
	Kludge: leverage the namespace cleanup on _parsing_ !!
	(which means, write to string & reparse!!!)
	(acknowledged need on libxml2 end, but not done)

      - Portability fixes to LaTeXML::Util::Pathname
        Should work in Windows, thanks Ioan Sucan.

      - Modifications to Constructor patterns
	* Changed the `property' value pattern to '#name'
	  (eg. #body instead of %body).        
	  (Gratuitous, but simplifies the grammar)
        * Values in patterns, #1 and #name can now
	  be followed by {key}, for KeyVal arguments,
	  to access the value associated with a given key.
	* Conditional expressions now recognize general values:
	  ?<value>(...) and also accept an else clause
	* The NOT conditional, !<value>(...), is removed;
	  Use ?<value>()(...) instead.
	* Prefixing the constructor pattern with '^'
	  allows the generated XML to `float up' to a parent
	  node that is allowed to contain it, according to
	  the document type.  The floats keyword for definitions
	  is also removed.
	* The untex strings for constructors that shouldn't appear
	  in the math TeX string (used for image generation)
	  should now be empty, '', instead of using the floats keyword.

0.1.1 2004-06-15
      - Packages:
        * Made package loading more robust; doesn't re-load;
	* Crude access to options
	* Added several missing definitions to TeX & LaTeX
	* Implemented comment, acronym packages;
	  initial (mostly empty) amssymb

      - General:
        * Catch filters that don't actually change the input.
	* Refactoring: New module Global.pm carries all exported
	  constants and constructors to simplify coding.
	* Refactoring: name changes & code movement of methods
	  confusingly called `digestFoo' and similar.
	* Refactoring: Moved macro parameter handling to new
	  module Parameters.pm (and as side-effect had to
	  rename parameters to Register (DefRegister, etc))
	* More careful Token equals method, so newline can convert to
	  a T_SPACE w/ newline inside; this means the output nominally
	  preserves lines! (but STOMACH->setValue('preserveNewLines',0);
	  disables it).
	* New constructor \@VALUE fetches values from stomach during
	  absorbtion in intestines. This (or similar) used to put
	  reference numbers in \ref, like 2nd LaTeX pass.

      - Math:
        * Introduced new element <Math> which can contain the various
	  alternative representations of math, such as XMath, m:math, ...
	  Moved most of XMath's attributes to Math.

      - Error/Warning Messages
        * Added messages to show progress during processing (unless -quiet)
	* New SalvageError message for things that in principle are errors,
	  but we're going to try to proceed; added some things to this category,
	  like unknown macros, and such.
	* Added source locator to Whatsit to improve error messages.

      - PostProcessing:
        * Fixed xml catalog so it finds mathml dtd and entity files
	* Fixed latexmlpost and LaTeXML::Post to recognize html and xhtml
	  output formats; refined the stylesheets (LaTeXML-html.xsl and
	  LaTeXML-xhtml.xsl (both of which include LaTeXML-base.xsl))

      - Put LaTeXML tags in thier own namespace:
	And first pass at fixing postprocessors to recognize this
	[probably introduced bugs, and in any case, namespace normalization
	 is pretty crummy]

      - NEED TO DO:
      	* Implement alltt package
	* Question: Should XMath be duplicated before parsing?
	  (ie w/different status=tokenized|parsed|partially-parsed|....)
	  This would allow more inference and then re-run the parser.
	* Extend constructor syntax to work with KeyVals, apply random functions?
	* Refactor DOM? eg. use XML::LibXML, move more analysis to Intestine
	  this needs (at least) resorting Font reduction.

      Thanks to Michael Kohlhase <m.kohlhase@iu-bremen.de> for comments & examples
      leading to many of these patches.

0.1.0 2004-05-10
      Initial (pre)release