A. Pagaltzis wrote:
>[snip]
> RFC4287 defines the relation value in terms of IRIs, but does not
> require that relations be compared as such, and then constrains
> the set of values to a subset of IRIs. This constrained set is
> more amenable to simple string comparison. My interpretation of
> these facts is that string comparison is explicitly expected.
>
+1. I don't believe any reasonable implementation should be expected to
do anything more than a simple case-sensitive string comparison of the
IRI. If an implementation chooses to be forgiving and perform some form
of normalized comparison, then that's the implementations choice. In
Abdera, I went ahead and checked in a change that normalizes the IRI
then compares the result with the
"http://www.iana.org/assignments/relation/alternate"; form.
> Given then that all implementations that I have read the source
> of do indeed compare relation values as strings, my conclusion is
> that while you are free to compare them as IRIs, doing so is
> unwise; likewise, while you are free to specify registered
> relation values in your published feeds as absolute IRIs
> including the http://www.iana.org/assignments/relation/ base,
> doing so is unwise.
>
I wouldn't say comparing as IRI's in unwise as much as it is simply very
forgiving. Publishing a feed that uses rel values that are not in a
normalized form is unwise as your pretty much guaranteeing that folks
won't be able to process your feed properly.
- James
