                TMX Specifications
                ==================
                Nov-25-1997


This package contains the following files:

tmx.htm          TMX Specifications.
tmx.dtd          SGML Document Type Declaration for TMX.
tmx.gif          TMX logo.
lisasig.gif      LISA-sig logo.

sample1.tmx      Short sample file in TMX format.
sample2.tmx      Sample of TMX with English, Japanese and Hungarian data.
sample3.tmx      Sample of TMX with Thai and English data.
sample4.tmx      Sample of TMX with English and Turkish data.

parsetmx.bat     Batch file to process a TMX file with NSGMLS.
sgmlxml.dec      SGML Declaration file for SGML processing of an XML file.
tmxcatlg.txt     TMX catalog file for NSGMLS
readme.txt       This file.


NSGMLS
------

In order to run the ParseTMX.bat file from a Windows 95 DOS prompt, you must 
create a BIN subdirectory in the directory where ParseTMX.bat resides and 
place in it the following executables from the SP SGMLS parser package 
(available from ftp.jclark.com in the sp directory):

NSGMLS.EXE
MSVCRT.DLL
SP120.DLL   (Certain versions of the package will use a file called 
            SP110.DLL in the place of SP120.DLL. This version will
            also work.)

Then, for example, issue the following command from the DOS prompt:

parsetmx sample1.tmx

and check the parse.err file. If it is zero bytes long, then the parse was
fully successful. Otherwise, it will contain error messages. If your error
file returns errors of type 'x???? is not a valid function' then the parser
does not support hexadecimal character references and you will need to
perform a global search and replace, as described in the following section.


NUMERIC CHARACTER REFERENCES
----------------------------

&#x00AB; and &#xAB; are both valid character references in XML. But
depending on the parser you use they may be not implemented correctly and 
may require exactly four hex digits.

In SGML as first standardized in 1986, only decimal character references 
are allowed (&#nnnn;). However, there is a proposed modification to SGML 
allows hex references as well, for compatibility with XML (see 
http://www.ornl.gov/sgml/wg8/document/1929.htm section heading: "Hex
character reference", which is a cryptic description of the proposed
modification to SGML to allow hex character references as defined in the
XML specification).

Note that a particular SGML declaration is also used in this package to
allow XML empty tags that end with a slash, such as <META/>.

In the meantime, until SGMLS parsers are updated to support XML features,  
there are several workarounds. One is to set the -E (error) parameter to a 
high number and look for errors besides the "not a function" errors 
generated by hex character references. Another is to do a global search and
replace on the TMX file, changing "&#x" to "$#x" before parsing and then 
back as it was after parsing.


VIEWING TMX DOCUMENTS 
---------------------

The hexadecimal character references described above can be hard to
decipher, especially if a TMX file has CJK or other non-Latin character
language information. It would be much simpler to view the document as true
Unicode, with all charters displayed as they were intended. A utility to
facilitate this is available for download on the TTT.ORG website, in the
Technology section. This utility, called UHA2UCS, converts any UHA-8 file
(8-bit Unicode Hexadecimal characters and ASCII) to standard 16-bit UCS-2
Unicode. These files may then be viewed and edited in any Unicode-capable
text editor. Information about one such program available for trial use
can be found at Duke University's Humanities Computing Facility website:
http://www.lang.duke.edu/faq/faq.htm



-end-
