[Ex-plain] Attrset 1=1
mike at tecc.co.uk
Mon Apr 22 15:15:09 CEST 2002
> Date: Mon, 22 Apr 2002 13:12:05 +0100 (BST)
> From: Robert Sanderson <azaroth at liverpool.ac.uk>
> Currently if a server gets a request for scan and there's no default
> value for scan, then there's no definable behaviour for records
> which do not have the scan attribute.
What is a "request for scan"?
> The server cannot assume that missing==false unless the DTD/Schema
> explicitly records this. Missing could just as easily be true, the
> server or server author has no way of knowing which one we meant.
> In the case of search/scan/sort, true false and missing all have
> different semantics. missing is not false, missing is 'I don't
> know'. You -cannot- assume that if it's not present then it's
> false. If you want to search for servers that do /not/ support
> scan, then you can't do a !true search, you need to explicitly do a
> search for false as there is a not true and not false possibility.
I could not possibly be in more emphatic agreement.
> If you need to be able to distinguish between true, false and
> missing then you can't have a single access point which is used for
> more than one XML attribute, ala the current 1=1
OK. (Well, you could, but the details would be unpleasant. So let's
> > > If we're just going to default to false, we should do this
> > > explicitly in the Schema/DTD.
> > Agreed; but we're not going to, right?
> On the contrary, unless we want lack of interoperability and
> functionality or to change the attribute set, we need to do this.
What?! You think we should make the DTD default to "false" instead of
"I don't know"? Why? How does it help interoperability to make
servers lie about how much they know about the servers they're proxing
/o ) \/ Mike Taylor <mike at miketaylor.org.uk> www.miketaylor.org.uk
)_v__/\ The IBM 360 had no stack, and that was stupid, short-sighted
design. The Cray 2 has no stack either, but that's elegant
More information about the Ex-plain