To: MX%"[log in to unmask]"
>Can anyone comment on Omnimark vs. SGMLS.pm as a tool for doing SGML
>transformations and down-translations?
Here's my personal subjective off the top of pate comparisons:
price Large Zero
feature set large but closed unlimited 
learning curve steepish flat for Perlers;
steep for others
performance excellent good
SGML-support full full 
ESIS access full (?) limited 
 Omnimark uses its own programming language which, to my eyes, looks
a little strange. SGMLS-pm uses perl5 which to some eyes looks even
stranger, but which has the virtue of being very very widely supported.
 Both products incorporate fully conformant SGML parsers. Omnimark's
engine is proprietary but very very fast; SGMLS-PM uses James Clark's sp
(though it could presumably be tweaked to operate on the ESIS from any
 SGMLS-pm only keeps track of the ESIS tree as it steps sequentially
through the document. So forward references (e.g.) require special
treatment -- using either a supplied set of stack-handling routines or
multiple passes through the file. I believe that Omnimark does the same,
but packages it more transparently.
Another product which you should take into account is BALISE from AIS.
This is at least as powerful as Omnimark, (and just as expensive) but
has a programming language which (to my eyes at least) is rather less
weird. It uses J Clark's SP.
There was a session comparing these and other transformation tools at the
SGML94 conference in Montreux which got rather heated...