Re: "Web addresses in HTML 5" for review (ISSUE-56 urls-webarch)

Hi,

looking at the "Web Addresses in HTML 5" draft, two quick comments:

1. It seems that in section 3, step 6, the UseSTD3ASCIIRules flag  
should *not* be set:  These are rules to be followed when constructing  
a zone file; it seems unadvisable to enforce them when trying to  
resolve a URI reference.  (Note that RFC 3490 leaves the choice to  
implementations.)

2. While I frankly haven't spent the time to study the parsing  
algorithm in depth (and would much prefer to see this specification  
layer on top of the existing IRI and URI specs), I'm a bit puzzled by  
this remark in the beginning of appendix B:
> Known discrepancies:
>
> The algorithm below allows square brackets in the <host> field; the  
> algorithm in section 2 above does not.
It appears that the algorithm in section 2 includes special casing for  
IPv6 address literals (which use square brackets), as does RFC 3986,  
as does the old algorithm.  This might just be an editorial issue with  
that particular note, or it might point to a deeper problem.

PS: Please keep me CCed explicitly on any follow-up messages.

Hope this is useful,
--
Thomas Roessler, W3C  <tlr@w3.org>