Re: Andy's review of the the protocol doc (part I)

On Thu, Aug 04, 2005 at 04:35:17PM +0100, Seaborne, Andy wrote:

>     <xs:element name="malformed-query"></xs:element>
>     <xs:element name="query-request-refused"></xs:element>
>     <xs:element name="fault-details" type="xs:string"></xs:element>
> 
> which is 3 one of which is called "fault-details"

I don't think that makes that element a fault message, though.

I really meant a schema for this kind of structure (and I realize the schema
above doesn't describe this):

<malformed-query>
<fault-details>blah, blah, blah</fault-details>
</malformed-query>

<query-request-refused>
<fault-details>This query would kill our service. Go away.</fault-details>
</query-request-refused>

or

<malformed-query/>
<query-request-refused/>

Or, serialized in HTTP (some details wrong...):

HTTP/1.1 500 Internal Server Error
Date: Fri, 06 May 2005 20:55:12 GMT
Server: Apache/1.3.29 (Unix) PHP/4.3.4 DAV/1.0.3
Connection: close
Content-Type: text/plain

This query would kill our service. Go away.

I.e., I think we need a way to communicate fault messages optionally. I'll
get this sorted out.

> >Hmmm, not sure I understand yr point.
> 
> Just that <_:abcd> is illegal as an IRI.  This MalformedQuery does not 
> allow much wiggle room for (non-standard) language extensions using the 
> protocol.  A SPARQL Protocol service MUST bounce illegal queries so can't 
> use the protocol with non-absolutely-correct-SPARQL queries which is what I 
> though UMD wanted.

Ah, good point. I guess that must can be a may or a should, then.

> "legal sequence of characters" can be read as "legal by the grammar" or 
> "legal by the grammar and any other rules that apply".

Yes, it can. I see at least two options:

1. redefine malformed query
2. leave the def'n as-is but change to should or may

Cheers, 
Kendall

Received on Thursday, 4 August 2005 15:51:20 UTC