> > And most browsers can handle forgetting to put the URI in
> > quotes, but they
> > will get hopelessy confused when only one of the quote marks
> > is missing.
> Netscape doesn't seem to mind, so my program shouldn't mind. And if
> it doesn't mind, then my program should also accept it. But, since that
> clearly is illegal according to the official standards, the official
> standards are not a reliable guide to what my program must accept.
Hmmm. We in the Ada community are accustomed to a standards mentality
that defines both the minimum _and_ the maximum sets of legal programs.
Ada 83 was, famously, like this, of course.
We tend to forget that - in practice - most standards like this just
define the minimum set of legal programs. That is, "if a program is
legal you must accept it." They don't say much about the illegal ones,
that is, they do not say "if a program is illegal you must reject it."
Implementers may (and often do) move programs freely from the rejected
set to the accepted set. It's called proprietary "extensions.":-)
Even GNAT is moving this way, that is, extending the set of syntactically
and semantically accepted programs beyond the RM-legal set.
Quiz: who knows what the GNAT extensions are? I know at least one.:-)