> On 11 Nov 2016, at 08:20, Michael Kay <mike@saxonica.com <mailto:mike@saxonica.com>> wrote: > >> >> Fallback to codepoint collation on every uca uri? >> >> >> Seems the feature is optional after all. >> > > The working group often reminds itself in such situations that conformance and fitness-for-purpose are not the same thing, and that fitness-for-purpose is entirely the resposibility of the vendor. > > Michael Kay > Saxonica > Another example of this effect, which I have often had occasion to remind people of, is that it is perfectly conformant for a processor to reject every call on doc() with the error "document not found". The product would be conformant but unsaleable (unless perhaps it's designed exclusively for embedding in non-internet-aware toasters). As for collations, we made the decision quite deliberately. We want users to have an interoperable way of requesting the collation they want; whether they choose to run the application with a product that only offers a rough approximation to this collation is up to them. We've adopted this policy consistently when designing features that control localisation of output: the request should be interoperable even if the output isn't. We've worded the spec to strongly encourage implementers to provide the feature, but we do know there are environments where implementation is challenging (like in the browser!) Michael Kay Saxonica
Received on Friday, 11 November 2016 08:31:18 UTC