Re: [xri] Property, rel comparison

[Breno de Medeiros <breno@google.com>]

>
> If it's application-specific, then it's not scheme-specific either, it's
> just application-specific.

Do we really need this level of generality?

>
>> Okay, and do you want to define rules for these in the spec? Or rely
>> on existing scheme-based normalization?
>
> Either there's an existing reference or there isn't. Eran was saying earlier
> there isn't, but I think it's moot if the rules are NOT in fact
> scheme-specific but left to applications to define.
>
>>> Relation types and property types URIs on the other hand are strings
>> structured like URIs. They should follow the same rules as XML namespaces
>> and should not be normalized. There is no reason what-so-ever for anyone
> to
>> use different variations of these URIs - that's where interop breaks.
>>
>> Okay, in this case, I propose we have rules on valid names that make
>> normalization unnecessary.
>
> I'm not sure what you're suggesting there. The set of valid names is simply
> the set of absolute URIs; it can't be restricted further, can it? It's not
> good practice to include certain constructs in such URIs, like ports, but it
> can't really be made illegal.

According to Eran, these are not URIs, but XML namespaces in URI format.