Re: scoping the IRI processing spec (was Re: status of IRI work)

On 2011-09-25 19:27, Chris Weber wrote:
> On 9/20/2011 5:21 PM, Bjoern Hoehrmann wrote:
>> I think there are two extremes: one is that "HTML5" can say "Oh, URI and
>> IRI stuff, check this RFC, that's all" and the other is that "HTML5" has
>> no need to contradict what we specify but is not aided much beyond that.
>> I think the working group should make a specification that does not have
>> to be contradicted; anything beyond that is quite optional as far as the
>> "'HTML5' requirement", or any other "user"-specific requirement goes.
>
> You think it would be a mistake for this WG to put out an IRI processing
> spec that helps to define, or align with, HTML, or HTTP-specific
> behavior. What do other members of the WG think?

How does HTTP get into this discussion?

> It seems to me that some share your view while others have been in favor
> of moving this idea forward in an attempt to align, or control, the
> outcome. Maybe that's because HTML5 as a reference could have a stronger
> influence than 3986/3987 in how other protocols would process URIs/IRIs?
>
> Bjoern do you consider it a contradiction or an override that HTML5
> would redefine the <unreserved> and other aspects of the ABNF definition?

I think that requirement comes from the wish to be able parse and 
resolve references that are invalid.

One way to do that is to simply point people to RFC 3986 (regexp-based 
extraction plus the resolution mechanism) and be done. We also could 
work on 
<http://greenbytes.de/tech/webdav/draft-reschke-ref-parsing-latest.html>, or 
integrate it into something else.

Alternatively, we could expand the ABNF to more or less match 
"everything", so doing what IRIbis does for LEIRIs, just doing a bit more.

I don't think it makes a huge difference.

> Do you see it as a contradiction or override that HTML5 would replace
> all U+005C REVERSE SOLIDUS (\) characters with U+002F SOLIDUS (/)
> characters?

As Björn observed, this only affects handling of invalid URIs/IRIs, so 
it's ok to write down rules for these *if* the UA vendors actually agree 
on the rules (my recollection is that they do not).

> ...

Best regards, Julian

Received on Monday, 26 September 2011 13:47:05 UTC