> Tim Wegner writes:
> > > Seriously, though, I would suggest that a well-written browser should
> > > gracefully ignore any characters that they don't recognize (possibly
> > > issuing a warning as they do so).
> >
> > This has been said by others, so at the risk of being repetitious,
> > this isn't good enough. Everything after ^z should be ignored.
>
> Fair enough. I've updated the list of proposed changes
> (http://sunee.uwaterloo.ca/~broehl/vrml/changes.html) to spell that out.
Bernie and all, I'm not sure my advice about ^z belongs in the VRML
spec. It's more under the rubric of robust behavior for worldly-wise
VRML-esading software. Because of the historical circumstance that ^z
signalled the end of the file in prehistoric operating systems, a
smart VRML file parser might choose to handle a ctrl-z as though it
really does mark the end of the file, and only consider the
characters before the ^z as part of the "file". As someone pointed out, there
might well be circumstances in the future where ^z takes on some
other meaning in some character set context. If this happens then the
VRML reader will have to be even smarter, but in any case the ^z
situation should be handled and not simply cause a crash.
Perhaps the VRML spec needs a "recommendations" section such as we
put in the PNG spec. The PNG team doesn't consider the recommendations part of the
spec proper, but more along the lines of helpful advice for implementors.
Having said that, if VRML did have its own end-of-vrml-stream marker it
might make this sort of issue easier for VRML readers to handle,
unless it is already clear to parsers when the VRML is finished.
Tim