RE: Don't copy live properties

Larry Masinter writes:
> I don't think it is possible to specify server behavior on COPY
> when applied to live properties. So I think the standard should
> say that it is unspecified. It's not required, it's not outlawed,
> it's just unspecified. Clients shouldn't depend on it.

This suggests to me that the behavior of a live property on COPY is a
quality of the property, rather than a quality of the server.  Hence, the
COPY behavior of a live property should be specified on a per-property
basis.  The discussion so far has effectively shown the COPY behavior for
all live properties is not a quality of the server.

Larry Masinter writes:
> Similarly, for all of the reasons given for inventing live
> properties in the first place, don't think it is a good idea
> to "outlaw" live properties. However, interoperable
> clients shouldn't depend on them.

The discussion has assumed there are two kinds of live properties,
"standard" and "non-standard".  There is a third class: those which are
registered (via an, as-yet unspecified, registration process) with a
concrete description of their syntax and semantics.  Lisa Lippert's
presentation in Orlando on a class of new live properties suggests to me
that such a registration process is necessary.

Registered properties will allow for interoperation, since client/server
pairs can implement the property according to the same semantics, yet not
require the effort required to create community consensus around a set of
properties.

> However, interoperable clients shouldn't depend on them.

Agreed.

Jim Davis writes:
> Finally, I doubt the distinction between descriptive and operational can
be
> formallized well enough that we could standardize it.

I agree.  Is the "DAV:getetag" property descriptive or operational?
Furthermore, many of the properties that Lisa Lippert described are
qualities of the collection's contents (the list of its children) such as
hassubs and childcount.

Larry Masinter writes:
> I'm suggesting that we declare client dependence on non-standard
operational
> live properties as "bad design" and declare such dependence as
> non-conformant.

This seems too restrictive to me.  HTTP had a similar danger -- you could be
100% compliant, yet dependent on non-standard HTTP methods.  This was not a
major operational problem in practice.

- Jim

Received on Sunday, 3 January 1999 16:53:00 UTC