Pablo Castro wrote:
> Once we have something like that in place, if we allow
> developers to customize the names of the links (rels), they
> can align them with common vocabularies if they exist for a
> given domain.
<snip>
> Unused elements: I understand the concern about the unused
> Atom elements. We are thinking of adding more support to map
> elements of your data model that are a good match for Atom
> constructs to minimize that effect. That said, it would be
> good to explore is whether it is interesting to push Atom to
> somewhat new spaces where it makes a great format because of
> the fact that is broadly known, even thought for a particular
> case no all the elements of the Atom format are relevant
> (this is clearly not new, there are many folks that already
> explored this space).
If you can easily allow your users to map their custom schemas into
(pseudo-)standard formats based on Atom, such that the resulting
documents can be usefully processed by tools that don't know anything
about the database schemas (e.g. importing the events into Outlook),
then that is a useful application of Atom. But, I don't see any
technical advantage of using Atom if you aren't using its core features.
What is the advantage of using Atom over plain-old-XML or SOAP? In
particular, what is the advantage of replacing <Venue>, <Event>, and
<Attendee> with <entry m:type='EventSample.Venue'>, <entry
m:type='EventSample.Event'>, and <entry m:type='EventSample.Attendee'>?
If you need to represent multiple entities in a single document, why not
use <soap:Envelope>?
> One unfortunate effect of having $expand result a feed is
> that it changes the nature of the entity that is sent back to
> the client. If you say /People('abc') you get an entry,
> wouldn't be at least weird that if you add $expand as in
> /People('abc')?$expand=Addresses then the response is now a feed?
It wouldn't be more weird than having entries and feeds nested within
atom:link elements. Besides, I don't know how I can tell that
"/People('abc')" is going to return one entry or multiple entries just
by looking at it. Plus, RESTful software will not attempt to extract
information out of the IRIs; instead, it will base its expectation on
the format of the result based on the context in which it saw the
hyperlink.
- Brian
