RE: QSD

At 03:44 PM 1/25/02 -0800, Lisa Dusseault wrote:
>It could always be me...
>As I recall, the DASL spec for QSD says that every searchable property must
>be listed in the QSD info. 

I think you are mistaken.  Could you please say where the spec says (or
even implies) such a thing?  Perhaps the spec is poorly worded, such that
it allows one to mistakenly interpret it in that way.  if so the spec
should be re-written.

The author's intent was that QSD would be useful for cases where there were
a limited number of custom searchable properties, as is typically possible
with a document management system.  (The examples in the draft are framed
in terms of documents that are e.g. cooking recipes or nuclear weapons
plans).   But we also expected that DASL might be hosted on servers that
are allowing arbitrary sets of dead properties.  In that case, nobody
expected QSD to provide a complete list of all these dead properties.  (Not
that we would forbid it, but it's clearly too much work to be reasonable.)

That said, the use case you propose is, I think, the ability for a server
to disclose how it handles properties (you call them "custom  properties",
but I am not sure if you mean live or dead, not that it matters much) *as a
whole*, as opposed to on an individual basis.

You are right to observe that QSD for basicsearch does  not do this.
I think that's okay.  basic search is, well, basic, and so is QSD.  I think
it would be okay to release DASL with QSD as it is, that is, without making
it any more powerful.

Indeed, as you have noticed, there are some people who argue for removing
QSD altogether, on the grounds that DASL is useful even without QSD, and
QSD adds complexity (and controversy).

Received on Sunday, 27 January 2002 01:50:32 UTC