Dear Godfried,
But even then, I would be hard-pressed to see how XSLT would be better than Perl or a similar language when it comes to processing data in which which are not essentially to do with formatting XML documents for display.
I find it particular handy to operate on XML data in a code that is XML itself. As well XPath expressions might be way shorter (and clearer?) as Perl constructs? But it's a matter of taste.
If I wanted to, say, develop a program to lemmatise text marked up in paragraphs<p>, or parse Roman calendar dates marked up with<date>, is that something for which you would recommend using XSLT? XSLT would easily let you find that data, but I am not sure it would be easy to then process it in XSLT.
I would dare to say "They become more and more similar", e.g. with Perl coping better with XML while XSLT starts operating on the file system? As long as both are Turing-complete, it should be possible to implement every operation in both languages?
Not having associative arrays or more complex data structures in XSLT makes that particularly hard, not?
You aren't talking about XSLT2, are you? Because there we have e.g. temporary trees that make up for complex data structures and speed up the processing alot. Although processing data with a certain type of loop would also make up for that?
Best, Torsten