Re: [Gnash-dev] non-conformance to HTTP standards

[Eric Hughes]

At 09:38 AM 5/17/2007, Martin Guy wrote:
> Usually "lax with what you accept, strict with what you emit" is the
> byword so right on.

Thanks.  I wasn't worried I was doing a bad thing, just one worthy of 
making note of.  There are times when masking bad behavior in other 
software is A Bad Thing.  I don't think that's the case here.

> However I'm surprised if there's a hole in the
> HTML standard.

Don't be.  The formal grammars in RFC's are not of particularly high 
quality, in my experience, full of surprises, generally-small holes, 
ambiguities, and the like.  Putting a compiler to work would greatly help 
the situation.

So here's another one.  The new URI specification, RFC 3986, does _not_ 
accept lowercase hexdigits in percent-quoted characters.  "%7F" is legal 
but "%7f" is not.  This was not the case in the previous revision.  I am 
guessing this would be a surprise to most people, including some of the 
authors.  (FWIW, I'm accepting lowercase.)

And while I'm on the subject, the SMTP standard, even after its restatement 
RFC 2821, is not up to snuff.  For example, it doesn't even fully define 
the accepted-vs.-nonaccepted line ending grammar!  Sure, if everything is 
CR-LF, you're safe, but the handling of naked CR and naked LF is 
incompletely specified.  At least HTTP doesn't have _that_ problem.

Eric