Re: Illustratation of prefix interaction details

Dave Beckett wrote:
> On Mon, 25 Oct 2004 11:51:10 +0100, Dave Beckett <dave.beckett@bristol.ac.uk> wrote:
> 
> 
>>Here's a pathological example of when prefixes and QNames can be confusing.
> 
> 
> I should have said what I'd propose.
> 
> 1. Remove the second PrefixDecl term from the Query term, so there
> is only one place prefixes are defined; before the ReportFormat term.

+1

> 
> 2. Change FromSelector to only allow URIs. Edit to be ::= QuotedURI
> or alternatively, put it inside FromClause, delete FromSelector.

-0 (this is something I have had some requests for - it means that changing one 
prefix declaration changes the input graph and the query).

> 
> 3. Define prefix/URI expansion to happen in the order they are given
> in the syntax, with later prefixes overriding earlier ones.

+1 because this is a common programming paradigm.


One alternative is to make it illegal to redefine an explictly defined prefix.

This is acceptable to me but I'd note some systems have builtin definitions of 
things like rdf: and so implicitly defined prefixes will be necessary for such 
systems.

Another alternative is to just leaving it up to implementations as to the 
handling of redefinition is also possible.

Note to protocol - predefined prefixes greatly shortens URL for query requests.

	Andy

> 
> Dave
>

Received on Monday, 1 November 2004 12:21:37 UTC