Returning an AutoResult from an AutoRequest creation (was: Proposal: Link header in responses to AutomationRequest creation)

[Martin P Pain <martinpain@uk.ibm.com>]

See replies inline below


Tim Friessinger <TFRIESS@de.ibm.com> wrote on
03/12/2014 16:34:46:
> when I read the Auto V2 spec some time ago, I thought the idea is,
that an
> OSLC Automation Service Provider can return an Auto Result as response
to
> the Automation Request POST action by a client.
> With that returned Auto Result comes of course the URI of this Auto
Result
> (rdf:about), so the client knows how he can refresh/ update this Auto
> Result without the need for doing a query.
> I think this wasn't mentioned explicitly in the spec, but it was at
least
> how I thought it makes sense. ;-)

The resource that is returned by either a POST to
a creation factory, or from a creation dialog, can have both rdf:type oslc:AutomationRequest
and oslc:AutomationResult (i.e. it can be multi-typed), but I do not believe
that it can be solely a Result. Although it probably doesn't say that explicitly...
If the provider decides to create a single resource
that is a multi-types Request/Result, it probably ought to keep it that
way throughout the lifetime of that resource. For many implementations
I expect this would be fine, but not for some.

> So basically it's the same idea as the Link option, proposed by Samuel,
but
> maybe a little bit more "OSLC style":
> In Auto V3 one could specify that a POST action for an Automation
Request
> returns at least an Auto Result stub, so even if execution is delayed,
the
> client has at least the URI where to fetch the Auto Result again,
> to see if meanwhile something happend.

I believe (although Steve might be able to give more
information on this) that it was an explicit decision in v2 to allow providers
to not create the Result when the Request was created. For example, they
may have the Result created by their "agents"/"workers",
rather than by their central server. Or the Result may even be hosted on
a separate server (although I'm not aware of any best practice guidance
suggesting how to make sure the clients can find them, so that might be
problematic).

Unless we want to remove this sort of flexibility,
I think we will always need the clients to be able to deal with both cases:
where the server provides the Result (or its URL) at the time of Request
creation, and the case where the Result is not created until later.

So I believe that means the clients, to be fully 'interoperable',
would need to be able to:
- find the Result for a given Request using whatever
query facility is available in OSLC 3
- find the Link header on the Request's representation
(whether a response from the POST creation , or from a GET on its URL)
which points to the Result
- detect if the Request object is multi-typed with
oslc:AutomationResult as well.

And for v2 compatibility (if the server chooses to
be compatible):
- be able to find the Result for a given Request using
a v2 oslc:queryCapability.





> 
> ----- Forwarded by Martin P Pain/UK/IBM on 02/12/2014 10:09 -----
> 
> From:        Samuel Padgett <spadgett@us.ibm.com>
> To:        oslc-automation-comment@lists.oasis-open.org
> Date:        20/10/2014 18:23
> Subject:        [oslc-automation-comment] Proposal:
Link header in
> responses to AutomationRequest creation
> Sent by:        <oslc-automation-comment@lists.oasis-open.org>
> 
> 
> 
> Currently, as I understand it, Automation clients must query to find
the
> AutomationResult for an AutomationRequest they created. Something
like
> 
> oslc.where=oslc_auto:producedByAutomationRequest=<request-uri>
> 
> This requires at least two requests, one POST for the AutomationRequest
and
> one GET to find the AutomationResult. It also forces servers to support
> OSLC query. As someone who has written several OSLC providers, I can
say
> query is a lot of work to implement and difficult to get right.
> 
> I propose we add an optional Link response header [1] in responses
to OSLC
> AutomationRequest creation. This is an optimization to remove the
query
> step. For example,
> HTTP/1.1 201 Created
> Content-Length: 0
> Link: <http://example.com/auto/result/82>;
rel="
> http://open-services.net/ns/auto#produces";
anchor="
> http://example.com/auto/request/82"
> Location: http://example.com/auto/request/82
> 
> 
> 
> Here a client can discover the AutomationResult just by looking at
the Link
> headers on the POST response.
> 
> We could relax the MUST requirement for queries if a provider supports
the
> Link header. Providers that don't immediately create an AutomationResult
> for an AutomationRequest could leave off the Link header and continue
to
> use query.
> 
> Link is used in several places in LDP 1.0 [2], and we've used it in
OSLC
> 3.0 resource preview drafts [3].
> 
> [1] http://www.ietf.org/rfc/rfc5988.txt
> [2] http://www.w3.org/TR/ldp/
> [3]
> https://tools.oasis-open.org/version-control/browse/wsvn/oslc-core/
> specs/resource-preview.html?rev=42&sc=1#linkHeader
> 
> --
> Samuel Padgett | IBM Rational | spadgett@us.ibm.com
> Eclipse Lyo: Enabling tool integration with OSLC
> Unless stated otherwise above:
> IBM United Kingdom Limited - Registered in England and Wales with
number
> 741598.
> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire
PO6 3AU
> 
> [attachment "pic04575.gif" deleted by Martin P Pain/UK/IBM]
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number
741598. 
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6
3AU