XOM API for Processing XML with Java

["McEniry, Michael" <mmceniry AT itsc.uah.edu>]

> client. Possibly I can do it on the server by marking the files as 
> binary. Though then patches and diffs don't really work. However, 
> that may be the best solution.

On one project, we stored some test data files in CVS as regular text so
that we could use patches and diffs, but used a cheap perl script as
part of the test prep work to convert them to a specific pseudotext
format. Of course, we had to test the converter script itself.

Some other formats were very binary, but we used the same trick,
revisioning text files and generating the binary files. In some cases,
floating point implementations broke the tests on certain platforms
(Ultrix was too precise?). Same general problem though.

CVS is primarily for (ASCII!) text files, so its default behavior really
is the "right" one in most cases. It's not so good with binary
artifacts, and text files that need to be treated as binary just confuse
matters. Sometimes it's easier to have a separate archive for binary
artifacts, as one would for compiled distribution tarballs.