[mdx] what to trust

[Ian Young]

On 5 May 2009, at 16:37, Josh Howlett wrote:

>> I guess we
>> could fix that by (in the new publishing
>> protocol) always wrapping single entity documents in a
>> separate EntitiesDescriptor, but that sounds pretty unpleasant too.
>
> Is it the redundancy that you find unpleasant? IMO, it seems like a
> reasonable price to pay...

I think I expected (and I think it's natural to expect) that an API,  
particularly a RESTian one, would just return the EntityDescriptor  
when you explicitly request an entity by name.

    http://blah.blah/root/url             returns entire aggregate as  
EntitiesDescriptor
    http://blah.blah/root/url/entity/foo  returns one entity, as  
EntityDescriptor

I don't want to add complexity to cover some case that will be hard to  
explain.

On the other hand, maybe the ask-for-an-entity-get-back-an-entity  
model isn't the right one.  It's probably less of a stretch to regard  
all results as potentially multi-valued if we're thinking of adding  
some kind of query-by-tag as well as just the cases we originally  
thought of.  It's more capable of being generalised in the future to  
other kinds of query, too.

[I think it might also allow us to preserve an existing signature on  
the EntityDescriptor, from whatever source.  Given the brittleness of  
XML DSIG that might not be realistic, even if we could think of a  
reason it might be desirable.  That inner signature would in any case  
break if anyone did any transformation on the EntityDescriptor's  
contents whatsoever, which I'd expect would be happening quite  
frequently in practice.]

In other words, perhaps the way to spin this is saying that  
"everything comes back as an EntitiesDescriptor" is analogous to what  
you get back from something like an SQL query... you get back a row- 
set even in cases where you know in advance it will only have one  
thing in it.

One thing this approach doesn't help you with is the "no results"  
case; you still need to be able to say "no answer" in some way other  
than returning an EntitiesDescriptor with no contained  
EntityDescriptors, because that isn't schema-valid.

Scott might like to comment on whether using the @Name in this way (as  
the name of the group of entities of which the contained entities are  
a part, instead of as the name of the group of entities contained) is  
out of line with respect to the relevant spec.  To wording seems to  
allow for things to mean whatever a deployment wants them to, but I  
guess I might be choosing to read that section with more flexibility  
than intended.

	-- Ian



-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2448 bytes
Desc: not available
URL: <http://lists.iay.org.uk/pipermail/mdx-iay.org.uk/attachments/20090505/9393162c/attachment-0002.bin>