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

[Eric Hughes]

At 07:12 AM 5/17/2007, strk wrote:
> Which normative document says that a query string makes it *not* an abs_path ?
> I'd think both the following are abs_path:
>
>         /file
>         /file?v=1

Yeah, I thought both of them would product abs_path when I started.  Then I 
dug into the grammar definition.

The magic line is in RFC 2616, Section 5.1.2
        Request-URI    = "*" | absoluteURI | abs_path | authority

The definition of abs_path and authority are in RFC 2396.
        abs_path      = "/"  path_segments
        authority     = server | reg_name

Neither of these admits "?".   So for your example "/file" is indeed 
abs_path, but "/file?v=1" is not, because abs_path cannot contain 
"?".  Hey, it's not my rule.

As far as I'm concerned, this is a defect in the specification.

And now, a rant.  You'd think that after twenty years of writing standards 
documents with BNF and ABNF and the like that there would be software that 
would (1) check a grammar for consistency, (2) check a grammar for 
completeness and (3) compile into a recognizer that could be used to verify 
test vectors.  And much less standard conformance suites.  A standard 
that's not operational in these ways is far weaker than one that's just on 
paper.  Formal grammars are hard enough to understand without assistance.

Eric