Re: for telecon: ISSUE-23 @profile order

This all makes perfect sense. :)

On Wed, May 26, 2010 at 9:45 AM, Toby Inkster <tai@g5n.co.uk> wrote:
> Currently RDFa Core 1.1 says that when @profile includes multiple
> profiles, e.g.
>
>        profile="http://example.com/a http://example.net/b"
>
> these are processed from left to right. This means that if a particular
> term or prefix is defined in both profiles, then the definition in the
> latter profile "wins".
>
> If we were defining RDFa in isolation, this might make perfect sense,
> but @profile is an attribute from HTML 4 and is (at least
> theoretically, though often not in practise) used by microformats, so
> we need to take into account this historical baggage.
>
> HTML 4 defines @profile as a list but says that only the first URI is
> deemed to be significant, the rest being ignored. If we make allow
> later profiles to overrule the first profile, then this seems at odds
> with HTML 4 - we're saying that later profiles are more significant
> than the first.
>
> XMDP, the profile format used by microformats, extends the definition
> of @profile by saying that profiles are listed in descending
> significance. This means that when a term is defined in two profiles,
> the first profile wins. If we do this the other way around, it may
> prove an annoyance for people wishing to publish hybrid
> RDFa/microformats documents.
>
> It seems apt to mention GRDDL if only in passing. In GRDDL it doesn't
> matter which order you process profiles in. That's fine.
>
> Lastly, let's think about human psychology. We tend to employ a
> technique called cognitive frontloading - i.e. we get the most important
> things out of the way first. That means, if there's a bunch of profiles
> I'm going to type into a list, chances are that I'm going to type in the
> most important ones first. Given that, it makes sense for us to
> allow earlier profiles to override later ones.
>
> So my proposal is for RDFa Core 1.1 to state that profiles are handled
> from right to left.
>
> As an aside, we may wish to mention somewhere that implementations
> should be free to download and parse the profiles in any order, or
> in parallel. What's important is that the terms and prefixes defined in
> them are added to the lists of terms and prefixes in the correct order.
>
> --
> Toby A Inkster
> <mailto:mail@tobyinkster.co.uk>
> <http://tobyinkster.co.uk>
>
>
>

Received on Wednesday, 26 May 2010 09:22:57 UTC