Re: When correct software meets illegal data (was: Ada The Best...)

[tmoran@acm.org]

>But the whole worldwide web situation is a tribute to software that
>does what one means rather than rigidly following a spec.  What happens
>...
>but I know that mine aren't.  There are all kinds of HTML peccadillos
>that are ignored or worked with by the successful browsers.  If
>everything had to be correct, the worldwide web would be the worldwide
>wasn't.
  Being fault tolerant is nice.  Depending on fault tolerance is not
so nice.  The lack of an "as built" spec makes life very difficult
for anyone attempting to write code that accepts the same unknown set
of inputs as "the successful browsers".  It also makes life hard for
someone trying to write a web page that will be rendered very
similarly by *all* "the successful browsers".  As earlier pointed
out, if the tolerance was accidental rather than on purpose, it often
leaves a security hole.  Some tolerance is hard to imagine except
on purpose, e.g. <a href="http://domain.name/file<br>name>
and it's hard to tolerate on purpose if you don't know (from some
spec, if only an extracted one) that such a thing should be handled.