On 10/31/05, Luke Arno <[EMAIL PROTECTED]> wrote:
> That is a tall order and I think it requires some
> discussion as to how. :-)
>
> I would like to start with atom:id.
>
> Should it be generated client side?

I say no. I am currently writing a server and in general
don't trust the client. The easiest way for me to guaranteee
uniqueness of atom:id's is to over-write the atom:id with my own
server-side generated id.

My code assumes that the client has put an atom:id element
in the entry to make it valid and my server code replaces
the content of atom:id with the new value.
This does beg the question of what the client puts in the
atom:id if it knows it is just going to get over-written.

> Can our use of id at POST time address our
> idempotency concerns?

An interesting idea and a justification for requiring the client
to create a unique atom:id even though the server may
overwrite it but, in my opinion, this functionality is out of scope
for this discussion.

   -joe

--
Joe Gregorio        http://bitworking.org

Reply via email to