Re: couple of questions

On Sep 5, 2011, at 8:03 AM, Carlos Buil Aranda wrote:

> Dear all, 
> 
> in the algebra section of the fed document (http://www.w3.org/2009/sparql/docs/fed/service#algebra_service), in the algorithm, when checking if something is an IRI in the SERVICE address, I just write "if IRI then evaluate the pattern". Is this enough? not sure about it since the address must be an IRI.

Some more context might help here. I pushed Carlos on this issue, because I'm worried about how its specified. I initially thought this was just a syntactic issue. Saying "if IRI then…" without any reference to what "IRI" is being compared against seemed really odd, especially since there's only one case here: the "IRI" thing can't NOT be an IRI because that's the only thing the translation rules to algebra are defined to emit.

Upon digging deeper into this issue, I'm worried about how and where we define that the variable form of SERVICE patterns is an error. Right now it's defined in the evaluation semantics. But, as I said, the algebra translation rules aren't defined to produce algebra patterns with non-IRIs (VARs), so the else block in the evaluation semantics seems needless. I'm concerned that we can't just move the error condition to the algebra translation rules, however, because it feels really strange to define the VAR form of SERVICE patterns as an error, and then go right on to describe how the VAR form would work in the following section. This isn't (I don't think) like expression evaluation rules where implementations can extend error cases.

> also, in the SERVICE VAR section, is enough to specify that this section is informative to make clear that the implementation of SERVICE VAR is optional or should I be more specific?

I think this will depend on how the first issue is addressed (above).

thanks,
.greg

Received on Monday, 5 September 2011 17:02:51 UTC