Jim, I think we may be stretching the analogy here.
I am OK with you proposal: "I am ok if we just state that this scenario can happen, and that the client is responsible for determining if the update was successful (from its view) by GETing the representation immediately after and checking the content and comparing the ETags." Pls draft some text and say where it should be included so we have a concrete proposal on the table. Regards, ___________________________________________________________________________ Arthur Ryman DE, Chief Architect, Reporting & Portfolio Strategy and Management IBM Software, Rational Toronto Lab | +1-905-413-3077 (office) | +1-416-939-5063 (mobile) From: James Conallen/Philadelphia/IBM@IBMUS To: Arthur Ryman/Toronto/IBM@IBMCA Cc: Adam Neal <[email protected]>, [email protected], [email protected] Date: 09/13/2012 04:44 PM Subject: Re: [oslc-core] Unrecognized content Arthur, While I am not advocating that we respond with a 4xx when the server does not update a resource at a client's request. I point it out as an example of how even RFC2616 requires some interpretation in context. What I do want to do is address this very real problem (that DM and RRC are experiencing) that clients have when attempting to update resources (in general). Using your own Java analogy, a program that essentially is x = -42; out.println( x ); where the output is +42, because this particular type of variable only understands positive integers, and the programmer doesn't know this. Now he is forced to check every time he makes an assignment before proceeding with the program x = -42; if( x != -42 ) throw exception I think because the OSLC explicitly says that the server can ignore unrecognized content, and there is no guaranteed way for a client to know what content the server recognizes, the onus is on the client to be responsible for checking the results. I think we should provide some guidance and raise awareness of this scenario in our specs (non-normatively), so client developers can be prepared for this situation. I am ok if we just state that this scenario can happen, and that the client is responsible for determining if the update was successful (from its view) by GETing the representation immediately after and checking the content and comparing the ETags. Thanks, jim conallen Rational Design Management (DM) Integration Architect, OSLC AM Lead [email protected] Rational Software, IBM Software Group From: Arthur Ryman/Toronto/IBM@IBMCA To: James Conallen/Philadelphia/IBM@IBMUS, Cc: Adam Neal <[email protected]>, [email protected], [email protected] Date: 09/13/2012 04:03 PM Subject: Re: [oslc-core] Unrecognized content Jim, I disagree. The server ignored the content it didn't understand and did the update, but the before and after state was the same. According to your proposal, if I did a GET and then immediately PUT the resource, that should also result in an error because nothing changed. That would not be reasonable. Consider the following Java program: int x = 42; x = 42; That shouldn't result in a compiler error. In fact, the behavior of the server is somewhat undefined if a PUT would result in no change to the resource. The server could try to be clever and only update the resource if some property changed. Or it could take the request literally and replace the resource with an identical copy, but as a side affect, the modification date of the resource might change. Regards, ___________________________________________________________________________ Arthur Ryman DE, Chief Architect, Reporting & Portfolio Strategy and Management IBM Software, Rational Toronto Lab | +1-905-413-3077 (office) | +1-416-939-5063 (mobile) From: James Conallen/Philadelphia/IBM@IBMUS To: Arthur Ryman <[email protected]> Cc: Adam Neal <[email protected]>, [email protected], [email protected] Date: 09/07/2012 11:16 AM Subject: Re: [oslc-core] Unrecognized content Hey Arthur, The spec for the PUT method says: If an existing resource is modified, either the 200 (OK) or 204 (No Content) response codes SHOULD be sent to indicate successful completion of the request. If the resource could not be created or modified with the Request-URI, an appropriate error response SHOULD be given that reflects the nature of the problem. In this scenario the server did not modify the resource, because it didn't recognize the content. So according to RFC 2616 we should be returning an error response. Thanks, jim conallen Rational Design Management (DM) Integration Architect, OSLC AM Lead [email protected] Rational Software, IBM Software Group From: Arthur Ryman <[email protected]> To: James Conallen/Philadelphia/IBM@IBMUS, Cc: Adam Neal <[email protected]>, [email protected], [email protected] Date: 09/07/2012 10:15 AM Subject: Re: [oslc-core] Unrecognized content -1 for the 400 response code Jim, I don't understand what you are asking for. The spec already makes it clear that the server will discard unrecognized content. The client should expect that. What aspect of behavior is unclear? Regards, ___________________________________________________________________________ Arthur Ryman DE, Chief Architect, Reporting & Portfolio Strategy and Management IBM Software, Rational Toronto Lab | +1-905-413-3077 (office) | +1-416-939-5063 (mobile) From: James Conallen <[email protected]> To: [email protected] Cc: Adam Neal/Ottawa/IBM@IBMCA Date: 09/07/2012 09:03 AM Subject: [oslc-core] Unrecognized content Sent by: [email protected] In the current specification we have the statement: For OSLC Defined Resources, clients SHOULD assume that an OSLC Service will discard unknown property values. An OSLC Service MAY discard property values that are not part of the resource definition or Resource Shape known by the server. We are running into a problem. When a client (in this case another application server) PUTs an update to a resource that includes a 'link' to another OSLC resource, and the server, at the time does not recognize the link type, the link is not accepted, but a 200 OK is returned. The server returns a 200 OK, because it feels like it can ignore the unrecognized link. The client gets that 200 OK, and thinks that the link was successfully added. This doesn't feel right. The only way a client can be sure that the PUT worked as expected is to re-GET the resource and compare it to what it expected to see (with the new link included), and maybe do a little looking at ETags to make sure things haven't changed in between. I guess the server could instead return a 400 Bad Request, and include in the response the reason for not accepting the PUT. But if the content that was submitted really should just be ignored (i.e. is part of a future version of the resource), then we don't want to abort the update. The OSLC verbage does not provide any guidance as to what to do. It would be helpful if we had more detailed explanation of this statement in the spec. Thanks, jim conallen Rational Design Management (DM) Integration Architect, OSLC AM Lead [email protected] Rational Software, IBM Software Group _______________________________________________ Oslc-Core mailing list [email protected] http://open-services.net/mailman/listinfo/oslc-core_open-services.net
