Re: Trying to close ISSUE-14

hello ashok.

On 2013-04-02 9:54 , "Ashok Malhotra" <ashok.malhotra@oracle.com> wrote:
>Are you saying that the ordering may be completely determined by the
>server?

sure. just look at collections of resources made available through
services such as relevance engines. the ordering of resources in those
services is the main value those services expose (dump in data, get it
back in an order that is "useful" for consumers), but it is computed in
ways that are highly complex, require a lot of additional data, and are
opaque to clients.

>This is the case when the collection is unordered.  The server can
>present members in
>any order it likes.  But if the collection is ordered it has to be on
>some visible
>facet of the data.

why should order always be a function of some naïve ordering on a visible
facet? we are very interested in big data applications, and one important
scenario is to assume that you create a collection, have data in it, and
based on complex/expensive server-side computation, expose the same data,
but ordered in a way that is more useful than any naïve ordering.

restricting the representation of a container to only support no ordering
or naïve ordering seems very limiting. and afaict, richard's proposal
would not create that kind of limitation.

cheers,

dret.

Received on Tuesday, 2 April 2013 17:48:32 UTC