Thanks to all who offered suggestions on solving my problem with an
unterminating transform. I took your suggestions and tried running the
transform under a few different processors, which much success. Here are
the results: (all times are in seconds)

Sample Size:      LotusXSL       XT (Oct)     Saxon(Oct)
Saxon(July)
50kb                 2.5                   7
5.0                  3.9
100kb               8.3                   7
12.1                 5.1
200kb              42.2                  9
20.0                 7.2
400kb             348.9                11
62.8               12.8
800kb                **                  15
216.0               25.8

Conclusions:
While this test was by no means rigorous, it shows that the LotusXSL
processor is good for small samples but may not be suitable for large
samples. By comparison, the XT processor is very well suited for large
samples. Interestingly, the only processor which showed clear linear
behavior was the older version of Saxon.

A few notes about the test:
IBM's XML4J parser was used in all cases. Times listed for XT are
estimates, because XT doesn't seem to print out the processing time.

The XML sample is much broader than it is deep (many branches off the
root node, only three or four levels deep in places). The number of
branches off the root node grows pretty much linearly with the file
size. Thus, a 50Kb file has about 50 branches off the root node, etc...

Greg


 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list