RE: DASL draft issue: identification of query grammars

At 08:04 PM 12/15/01 +0100, Julian Reschke wrote:

>The problem is in the draft saying things like:
>
>The DASL response header indicates server support for a query grammar in the
>OPTIONS method. The value is a URI that indicates the type of grammar. This
>header MAY be repeated.
>
>For example:
>
>DASL: <http://foo.bar.com/syntax1>
>DASL: <http://akuma.com/syntax2>
>DASL: <FOO:natural-language-query>
></quote>
>
>It doesn't say *how* these URIs identify a query grammar.

True.  The intent and understanding of the original authors was that the
meaning was to be specified out of band.  Our understanding (based on
experience with other extensible query languages, e.g. Z39.50, SQL, and
DMA) was that the semantics of query languages is always specified out of
band, and that all that is required is for a client and server to agree
which language they use.  Opaque tokens are sufficient for this, and, at
the time, a URI was the best way to do it.  (Another approach, which we
rejected, was to use GUIDs).

So, your client contacts server X, and discovers it supports three
languages.  One of them is basic search, which is always required.  If you
discover to your delight that there is a better language in common between
you, you use it.  But otherwise basic search is your fallback.

But if your client discover the server supports some query language you
have never heard of, there is no concept of finding the meaning of this
language *at run time*.

The only operation on the query grammar is to compare it, as a unique
token.  To obtain the URI, you just concatentate the ns and the element
name.  (We did not specify this in the draft, but we should have).  So 

<basicsearch xmlns="DAV:" /> is the same URI as  <search xmlns="DAV:basic" />

Whether this is wise, or even legal, I can not say.  but that is what the
authors intended.

Received on Monday, 17 December 2001 00:16:08 UTC