Hi Sergey,

OData already supports this using the HTTP headers odata.version and 
odata.maxversion, see 
http://docs.oasis-open.org/odata/odata/v4.0/cos01/part1-protocol/odata-v4.0-cos01-part1-protocol.html#_The_OData-Version_Header.

A server that only accepts the OData query language but does not support OData 
JSON can respond with Content-Type:application/json and its own JSON dialect as 
long as it doesn't claim to be an OData service by also including an 
odata-version response header.

A server that supports both OData JSON and some other JSON dialect can serve 
OData clients as well as non-OData clients: the OData clients SHOULD include an 
odata-version or odata-maxversion header in their request.

A server that supports only OData JSON will respond with OData JSON, which 
after all is just JSON, so non-OData clients can digest it with JSON.parse or 
something similar.

And OData JSON in its "minimal metadata" or "no metadata" flavor looks pretty 
much like what someone would hand-craft: if a book has an id, the JSON object 
for a book will have an id name-value pair. Not much choice there. If you want 
to return a list of books, that's essentially an array of book objects. Given 
that an array as the root object of a JSON document is vulnerable to attacks 
(http://haacked.com/archive/2009/06/25/json-hijacking.aspx/), format designers 
are well advised to make the array a name-value pair of a wrapping plain 
object. Now the only design choice left is how to name the array. OData opted 
for "value", which seems pretty neutral.

Does that cover your use case?

Thanks!
--Ralf

-----Original Message-----
From: Sergey Beryozkin [mailto:[email protected]] 
Sent: Thursday, 13. February 2014 17:43
To: [email protected]
Subject: [OT] Is OData open enough ?

Hi All,

Now that CXF supporta OData $filter expressions with the help of Olingo, 
I've started wondering, after checking OData 4.0, are we really going 
for supporting 'OData the query language' to let CXF developers enhance 
the search capabilities of their existing applications not specifically 
written with 'OData the protocol' in mind or do we basically add one 
more CXF search extension (fine with me per se) which will let users 
'borrow' the OData $filter syntax ?

For example, suppose we have a service returning a collection of Books - 
this is all formatted in whatever format supported by the application, 
example, "application/xml":

GET /mybooks
Accept: application/xml

<Books>
   <Book id="1"/>
   <Book id="2"/>
</Books>

Now the users liked a $filter syntax and would like support the 
filtering of books by id,

GET /mybooks$filter=id%20eq%201
Accept: application/xml

returns

<Books>
   <Book id="1"/>
</Books>

; the same for application/json

may be I've got confused but it appears to me that it is really 
supported by the formal document ?
Example,

GET /mybooks$filter=id%20eq%201
Accept: application/json

is presumably expected to return JSON formatted per the OData rules *only* ?

I'd really prefer the document:
- making some 'space' between the actual query language and the protocol 
itself so that people can apply the OData the query language to their 
existing applications without having to re-write them to be OData centric
- this might be supported via $format query, but I'm not sure, I;d 
rather see something like

GET /mybooks$filter=id%20eq%201
Accept: application/odata+json

indicating that OData-centric JSON payload is expected back.

Simply
Accept: application/json;format=odata

would probably do too

is it already possible ?

IMHO, the lesser the binding of the language to the protocol is the 
better it is for the adopters

Thanks, Sergey







Reply via email to