[oslc-core] Search via POST vs. AtomPub-style collections

[Dave]

Based on the feedback in this thread, I'm going to remove Query by
POST from the OSLC Core spec. I will make these two changes today:

1) In Core spec Query Capabilities section: remove sub-section HTTP
POST Queries.
2) In Core spec, Service Provider section: remove the sentences that
reference Query by POST in the description of the oslc:queryBase
property.


I think the consensus is that Query by POST is important but not a
priority right now. We should return to the topic later. When we do,
we should pick up the conversation with the options we've discussed so
far:

a) Query by POST to Query base URL (as we had originally spec'd)
    Form-encoded post means query anything else means create

b) Query by POST, but to special URL
    Give each Query capability a query-by-post URL

c) Query by POST to Query base URL, but only with override header
    Query by POST indicated with header of X-Method-Override with magic value

Thanks,
Dave


-- 
David M. Johnson
OSLC Core Workgroup Lead
IBM Rational Software


On Mon, Mar 21, 2011 at 8:03 AM, Scott Bosworth <bosworth at us.ibm.com> wrote:
> Dave, the pattern you are trying to promote/allow for here makes a lot of
> sense to me as the simplest (and perhaps most common) way that
> ServiceProviders are defined. I can see this especially in cases where Query
> syntax is not fully employed but the querybase is just used to GET the full
> list of resources. So, I add to the list by POST and retrieve the list with
> a GET request to the same URI.
>
> I worry that the POST query will be overlooked by many service provider
> implementations. The fact that we have a spec topic with few or any
> implementations indicates to me that we got a bit ahead of ourselves and our
> principle of specifying only the minimum to support key use cases. That
> said, it seems reasonable that (as Andy and Nick have pointed out) long
> query strings are a realistic issue to deal with. Are there other
> alternatives to the POST query approach? If we can see a way to another
> solution in the future, then I agree that stepping back from the current
> approach is, in balance, a good thing to do....Scott
>
> oslc-core-bounces at open-services.net wrote on 03/21/2011 01:45:15 AM:
>
>> From: Nick Crossley/Irvine/IBM at IBMUS
>> To: oslc-core at open-services.net
>> Date: 03/21/2011 01:45 AM
>
>> Subject: Re: [oslc-core] Search via POST vs. AtomPub-style collections
>> Sent by: oslc-core-bounces at open-services.net
>>
>> I agree that the POST query is important for long queries. While
>> there may not be any implementations
>> that support such queries today, that may well change soon as people
>> start to add more complex clients
>> and integrations using OSLC. Queries from a central index using the
>> Change Log service currently
>> under discussion may need to be long and complex, to query against
>> differently named or valued
>> properties in the different service providers contributing to an index.
>>
>> Nick.
>>
>
>
>
> Scott Bosworth | IBM Rational CTO Team | bosworth at us.ibm.com |
> 919.486.2197(w) | 919.244.3387(m) | 919.254.5271(f)
>
> _______________________________________________
> Oslc-Core mailing list
> Oslc-Core at open-services.net
> http://open-services.net/mailman/listinfo/oslc-core_open-services.net
>
>