Ian Hickson wrote:
If so, that doesn't really buy much as far as forwards compatibility
goes. We have to be backwards compatible with what UAs accept, not what
validators accept.
Right, the parsing behaviour defines what UAs accept, as far as I can
tell.
However doing something like what Maciej suggests, of stopping the url
parser at the first whitespace character, sounds like it would solve the
forwards compat issue.
Agreed.
Ok, so we need to make that change to the spec once Anne gets back.
However, if the HTML5 algorithm only considers the same URLs valid as
RFC 3986 does, is there a reason not to point directly to RFC 3986
instead? Seems like there is no reason to have more relaxed error
handling here than needed?
As you said, we have to be backwards compatible with what UAs accept, not
what validators (and RFCs) accept.
This doesn't answer the question of what the win is of pointing
Access-Control to HTML5 rather than to RFC 3986 for the url parsing
algorithm.
/ Jonas
