I agree that there needs to be some knowledge of
how POSTing entries to a particular endpoint will
behave. I don't think we should spec this though.
For most cases it is just something to impart on
users. If you have a POSTing endpoint labeled "New
Entries" and one labeled "Import Entries" and some
help text somewhere that explains the behavior,
I think that is usually enough.

In special cases where machines need to understand
what exactly will go on at a particular POST IRI (this or
some other behavior that we cannot even imagine
right now) extensions can be used.

You make a good point that these expectations will
be important to confirm at times, but I do not think
this is core stuff.

- Luke

On 11/7/05, David Powell <[EMAIL PROTECTED]> wrote:
>
> >> Invalid Atom / atom:id handling
> >>     * http://www.intertwingly.net/wiki/pie/PaceMagicId (arno)
> >
> > Rejected.  There is consensus that clients SHOULD submit valid Atom
> > entries, and we do not specify any server behaviors with respect to
> > handling of atom:id values.
> > (see http://www.imc.org/atom-protocol/mail-archive/msg02739.html)
> > (see http://www.imc.org/atom-protocol/mail-archive/msg02829.html)
>
> I understood "servers can choose to accept/reject whatever
> they want" in those posts to mean that servers could accept/reject the
> entry. <http://www.imc.org/atom-protocol/mail-archive/msg02852.html>
>
> This isn't the same as "not specify any server behaviours with respect
> to handling of atom:id values" - which would allow servers to change
> the id on the server side.
>
> James's rewrite of Luke's proposed consensus, quoted in the message at
> <http://www.imc.org/atom-protocol/mail-archive/msg02829.html> seems to
> back that up.
>
>
> It seems unreasonable to ban servers from changing ids, but equally,
> other applications rely on the id not changing.  I'll try and write a
> Pace on using collection metadata to indicate the semantics of POST.
>
> --
> Dave
>
>

Reply via email to