Lists Home |
Date Index |
Karl Waclawek wrote:
> I read this as meaning that such prefixes are not illegal per se.
> But I am confused about how to treat declarations.
> "xml" may, but does not need to be declared.
> "xmlns must not be declared.
> But what about the other prefixes with leading characters "xml"?
> I would tend to think they should be treated like "xml", that is,
> declaration optional, but must be to the correct namespace.
> If no mapping to a namespace defined yet (in specifications), then
> do not reject, but feel free to issue a warning.
the only exceptions made to the requirement that a prefix must be be bound are those concerning the prefixes xml and xmlns. the phrase "processors must not
treat [all other prefixes beginning with the three-letter sequence x, m, l] as fatal errors" permits the appearance of such prefixes as the ncname of a
prefixedattname and as a prefix in a qname. there is nothing in the passage which parallels the assertion which is made with respect to the xml prefix, this is,
that "[i]t may, but need not, be declared," which would mean that the exceptions do not govern the appearance of such an unbound prefix. which mean the the
normal constraints would apply to them.