RE: Move and Delete (was: bind draft issues)

> The protocol that is used to implement something like "mv" should
> be an operation that preserves all characteristics of the objects,
> so the client (not the server) can decide whether to accept the
> information lossage inherent in copy/delete.
> 
That seems unrealistic to me.  Clients have shown that they don't care
about propertybehavior -- none of them implemented that part of RFC2518.
They just want to do a move! 

A server can often do a better job of move than the client's copy/delete
attempt.  Even if the server must transfer resources from one partition
to another, the server can at least set properties like "creationdate"
to the creation date.  A user doing a COPY will cause "creationdate" to
be initialized to the time of the COPY, and many servers don't allow
that property to be written. 

Neither the client nor the server can guarantee a 100% perfect
cross-partition move operation, but because the server can short-circuit
some permissions issues, the server is much more likely to do a much
better job.  

lisa

Received on Saturday, 8 March 2003 10:56:57 UTC