Re: bad code Re: Subject: ChatGPT results are "subject

> Michael, I wonder what "nasty accidents" you are thinking of -- some XSLTs
used in particular industries with real-world safety issues?


I came across an example where XSLT was being used to display the results of a
clinical study. Anomalous results were supposed to be shown in red. That was
controlled by a boolean test such as (@value > @threshold). In XSLT 1.0 that
did a numeric comparison, which was correct, but in XSLT 2.0 it did a string
comparison, and failed to highlight the anomaly. And they didn't have good
enough tests to spot it, so they went live with the bug. That's certainly a
potential safety issue.

Michael Kay
Saxonica