[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [cti] Re: [EXT] [cti] Embedded Relationships
I wanted to send out a ping to see if anyone can answer my questions below. Maybe someone has a little downtime this Friday afternoon to help me. I’m really curious what the return value should be.
Since I cannot gauge everyone’s body language over email, and I have not met everyone I’m not sure how to read most of these emails. I hope my intent to learn and improve the spec is coming through my emails. Writing a spec is challenging, and I understand that.
> The key is that for a product to be called say a "TAXII Collection Server", it MUST implement all of the methods that are defined.
Thanks for that statement. I was not sure if that was the case, and that is how I interpreted the spec. Then can someone provide an example return code and JSON object for a POST request with regards to Jason’s comments on personas/profiles that do not act as a repository?
> Just because you POST TAXII information to me, does not mean you can then later GET that same information from me, because I may not be acting as any kind of repository. I may not even have a TAXII "read" facility at all, and only accept POSTs to a channel or collection, and all GETs against the channel or collection return empty all the time. A use case for this may be a device that wants to expose a TAXII collection or channel to allow people to submit CTI to it to trigger some action, such as adding something to a watch list or launching a remediation. As such, I have no need to store this information at all, anywhere.
I’m not sure what the return code should be. The POST documentation says, “This Endpoint allows authorized TAXII Clients to add objects to a specified collection.” I assume the return code is 202 because the other codes are error codes, even though it is not really being added to a collection. What is the expected return code?
I also do not know what the return JSON should be. The ‘status’ JSON object is listed as the expected return JSON object. What would be the status code inside of the ‘status’ object? I assume complete since no further processing is needed but would like clarification.
Then for each object that was in the POST request they need to be put into one of three bins, successes, failures, or pendings. Into which bin in the status return object would each of the objects go? Pendings does not seem like a good bin because you already processed it and getting the status later returns the same type of status return object. It seems like an infinite loop. Failures does not seem appropriate because that seems like an error case to me. If I received a failure I would probably try again, but maybe that is the expected bin? Successes is an option, but it is a List<status-success>. A status-success object contains a required attribute of a url that is “the URL location of the created object.” I take the POST description statement of “add objects to a specified collection” and a returned “URL location of the created object” to mean I could do a GET request to try and download that object that I added to the collection, but maybe that is an incorrect assumption? Which bin would contain the objects?
Per the conformance clauses in section 8, if a product wants to be called a "TAXII Server" or a "TAXII Collection Server" then it MUST satisfy all of the requirements that are associated with that clause.
I think the problem is that you are assuming that because someone implements the GET methods, that the server has to answer them with data. As Jason has illustrated so well, the server can do what ever it wants with the data. Further, and implementation of the said server can do anything they want with the data.
The key is that for a product to be called say a "TAXII Collection Server", it MUST implement all of the methods that are defined.
Long-term, we may identify some "profiles" for TAXII, like some of the great ones that Jason has called out. But those will be profiles for what you do with the data, not which features you get to implement.
From: Reller, Nathan S. <Nathan.Reller@jhuapl.edu>
> Again, you are assuming that people are always occupying the "repository" persona. There are many personas for TAXII, not all of which store data. I highly suggest you take a look at the personas defined in the Interoperability Subcommittee's use case specification - https://docs.google.com/document/d/1l54RhjxwuXrZUQ19zIHUiZ7_c6otbLbVVfluKJogU7s/edit#heading=h.4do73o99e2l7
OK, now I understand. In the specification for TAXII I was reading Section 8 as meaning that all functions must be implemented because it says, “It MUST support all requirements as defined in section 3, section 4 and section 5.” It was unclear to me what that really meant, but all of the collections API stuff is in Section 5. With such a small spec it seemed to me that I must implement all of the methods. Here you are saying that I do not.
I would suggest making it clearer that you do not need to implement all of the methods and functionality. I would compare this to the OASIS KMIP spec that does clearly indicate there are different profiles and what methods and objects need to be implemented for each profile. They even have a separate document to list all of the different profiles, which I think you are calling personas although I don’t quite get that feeling after skimming through the document you linked. In the KMIP world if you don’t implement an operation then you MUST return an error action of “Operation Failed” with reason “Operation Not Supported.” Perhaps that is the intent of the 404 error, but I did not read it that way. The KMIP spec also provides a Query method that returns all of the supported operations. This way a client can verify what type of KMIP server it is talking to.
KMIP v1.3 spec:
KMIP v1.3 profiles:
In addition to making it clear that not all operations needed to be supported the profiles document provides a nice checklist of what objects and functions need to be implemented. As we implemented the KMIP spec for our open source library PyKMIP it was nice to have a checklist of what needed to be implemented. For instance, we first wanted to support symmetric keys. There is a profile called “Symmetric Key Lifecycle Profiles” and it tells me which algorithms and key types I need to implement. It seems like there is some of that in the personas document, but in my opinion I think the KMIP format is much cleaner to read. The nice part about the personas document is that it provides more context, so maybe we could combine the approaches? I think that would be very helpful to readers.
Symmetric Key Lifecycle Profiles:
For what it’s worth, KMIP Specification 1.3 and KMIP Profiles 1.3 are finalists for the “Outstanding Approved Standard” at the 2017 Open Standards Cup.
> Could you further explain this? I don’t understand how something can be a TAXII server but not have a historical database.
Description: S/MIME cryptographic signature