Lists Home |
Date Index |
On Sun, 2002-03-03 at 12:45, Elliotte Rusty Harold wrote:
> I think it most clearly does affect logical equivalence, especially
> after the latest XInclude working draft. Now that fallback elements
> can stand in for missing resources, and provide very different
> content, the actual content of a post-include document may vary
> depending on whether somebody's dropped a backhoe through various T-1
> lines connecting different parts of the Internet, and which parts of
> the Internet are relatively disconnected at which times. In such a
> world, XInclude resolution makes no sense as part of canonicalization.
I see what you mean, but I'd expect the canonicalization to have already
performed that resolution, and halted if it encountered a fall back.
The canonicalized document sent to me would have no risk of "backhoe
errors" modifying the meaning of the document, much as canonicalization
currently removes the risk of external entities suffering backhoe
> A document that includes a list of links and fallback content is
> simply not logically the same as a document in which all the links
> have been replaced by their remote resources. It's the difference
> between a phone book and all the people listed in the phonebook.
> They are related, but they are not the same thing.
I suspect that in the case of canonicalization, I'd rather have the
people than the numbers that may or may not work. I'm not sure a
digitally-signed phonebook is much good if the numbers are all broken.
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!