Re: Is about="[talk:]" a valid CURIE?

Thanks both to Manu and to Shane

I must admit that I was mixed up by the RFC text (I looked at it right
after Shane's mail). What got me mixed up is a stupid syntax thing. I
was used to the '|' character for (E)BNF notations, so when I saw

irelative-part = "//" iauthority ipath-abempty
                  / ipath-absolute
                  / ipath-noscheme
                  / ipath-empty

I was not sure whether this was the same as

irelative-part = "//" iauthority ipath-abempty
                  | ipath-absolute
                  | ipath-noscheme
                  | ipath-empty

And I had then to run to another meeting:-)

(As an aside: it is really strange that RFC-s are still written in ASCII
and not in HTML:-)

Anyway, issue solved! Thanks!

Ivan

Manu Sporny wrote:
> Shane McCarron wrote:
>> Manu Sporny wrote:
>>> curie :=   [ [ prefix ] ':' ] reference
>>>
>>> Which, I believe, means that this:
>>>
>>> about="[talk:]"
>>>
>>> is invalid, even if talk is specified as a valid prefix earlier in the
>>> document.
>>>   
>> No, that's wrong. Reference can be an empty string according to the
>> associated RFC.
> 
> Just got off the phone with Shane... He was kind enough to explain
> exactly how someone could understand this from reading the spec. There
> were two key points I was missing:
> 
> 1. The CURIE statement is written in eBNF[1] - I thought it was written
>    in some sort of W3C modified regex syntax.
>    (eBNF statement) curie := [ [ prefix ] ':' ] reference
> 
> 2. Not specifying anything for reference is technically valid per the
>    IRI spec[2]. Here's how:
> 
>    reference      := irelative-ref
>    irelative-ref  = irelative-part [ "?" iquery ] [ "#" ifragment ]
>    irelative-part = "//" iauthority ipath-abempty
>                        / ipath-absolute
>                        / ipath-noscheme
>                        / ipath-empty
>    ipath-empty    = 0<ipchar>
> 
> Therefore /reference/ can be /ipath-empty/ and thus, you can have:
> 
> prefix:
> 
> Just to see if this was just confusing me, or would confuse future
> implementers, I asked three of our engineers to briefly look at the
> CURIE section of the RDFa spec and state whether or not the following
> were valid CURIEs forms (Y means "Yes", N means "No", ? means "Don't know"):
> 
> Y Y N prefix           (Correct answer: N)
> N N N prefix:          (Correct answer: Y) *current discussion topic
> Y Y Y prefix:reference (Correct answer: Y)
> N N N :                (Correct answer: N)
> Y Y ? :reference       (Correct answer: Y)
> Y Y Y reference        (Correct answer: Y)
> 
> This means that:
> 
> 1. We need a test case clarifying this, which I'll put together.
> 2. Ivan gets to keep his implementation as-is... and librdfa will have
>    to change :)
> 
> -- manu
> 
> [1] http://en.wikipedia.org/wiki/Extended_Backus–Naur_form
> [2] http://www.ietf.org/rfc/rfc3987.txt
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf