drop sortby from simplesearch

I propose that we drop the sortby clause from simplesearch for the
following reasons:

1) Although SQL supports sorting (ORDER BY), other widely deployed search
systems (e.g. Verity) do not.  (More precisely, Verity sort results in a
fixed order, relevance, and there is no way to change this.)

If sorting is mandatory, it raises the bar of implementation.  If sorting
is optional, then conservative clients will never use it, since it might
not be implemented by all servers.  (My understanding is that it is
currently optional, in the sense that the set of sortable properties maybe
empty.)

2) The client is just as competent to sort the results as the server,
assuming that the properties on which it desires to sort are also
selectable.  Removing sorting from the protocol simply shifts a constant
amount of work from the server to the client, but there are more clients
then servers, so this distributes the work better.

3) In at least applications, clients will prefer to sort client-side
anyway.  For example, my email reader can sort messages on date, author, or
subject.  It does this sort in the UI, not on the server, because it's
faster to sort client-side.

4) In order to describe sort in query schema discovery, we were forced to
introduce an artificial property (dav:rank) (see 7.19.1)  this is ugly, and
we can get rid of it by abolishing sort.

I don't feel strongly about this.  I don't think it does much harm.

Received on Friday, 24 July 1998 14:14:26 UTC