Re: Tools / process for building the MathML test suite?

Hi Ian,

> I am wondering what process the MathML WG adopted 
> to create the test suite. For instance, was there a 
> particular process followed to submit/approve tests? 
> What tools were used to create the test files or
> to generate the test suite?

Neil Soiffer at Wolfram Research and I mostly put together the MathML
test suite, with a few other people contributing a test here and
there. Unfortunately, the tools we used are not portable.

I wrote an ad hoc CGI-based system for submitting tests, either one at
a time, or zipped together in a batch.  We defined a little XML
vocabulary for a test -- author, description, spec section, rendering,
MathML code, etc.  Then I wrote some scripts for adding, deleting and
modifying tests.  I hooked up DSI's WebEQ MathML software on the back
end to generate images of renderings if they weren't submitted as part
of the test.  Finally, there were some scripts for regenerating the
table of contents, for zipping the whole suite up, etc.

It was all custom Perl code, very much tied to our specific test
suite structure, and I could only swing it because I hosted it on one
of our company servers where I had complete control over permissions,
etc.  There was also plenty of hand work with Perl, find, grep, etc
that I just coordinated with modifications to the scripts.

It worked okay, though it suffered from growing pains, since I was
building the management system at the same time as the test suite.

Unfortunately, the system is pretty much useless now.  Once the suite
was "done" we moved it to the W3C servers, where there was essentially
no hope of getting enough access to continue to maintain the scripts
-- for example WebEQ needs access to an X server for graphics calls,
and there were a bunch of other insecure aspects to it.  I think it
would really have to be done by one of the web admin guys there.

Further, once the suite was at W3C, we started having to make changes
by hand, to handle bugs and errata, and it just got out of control to
the point where it would take some significant effort on my part to
reimport into the development environment.

However, Neil and I are tasked with doing a major round of
additions/corrections/reformatting soon, and so I have been wringing
my hands about how to best go about it.  I really just don't have time
to do the programming for the forseeable future, but I would gladly
zip up the stuff I have and send them off to anyone who is interested
in taking over and putting in the time to make it a really robust,
general purpose system.  If you pulled out the WebEQ part, and just
stuck with requiring any images to be submitted along with the
associated test, that would clear out one of the big obstacles.

Let me know.


Robert Miner                          
MathML 2.0 Specification Co-editor                    651-223-2883
Design Science, Inc.   "How Science Communicates"

Received on Thursday, 20 September 2001 13:59:28 UTC