Now, that’s cute.
I’ll get right to the point. He writes;
However, in the asynchronous case, imagine your Web Service is just like a letterbox. Letters happily fit into the letterbox and get processed by the back-end (i.e. you), but the 42″ plasma screen TV you ordered doesn’t fit into the letterbox – it doesn’t match the physical dimensions (as per its “schema”). This seems RESTful, no?
Absolutely.
The fact is that regular Web Services have a far more uniform interface than the REST style. I’ve laid down the challenge before, but you can’t get much more uniform than one (imaginary) verb, can you?
That’s really good to hear you say that Web services have uniform semantics. But it’s impossible for them to be more uniform than REST, because REST prescribes uniform interface semantics by definition. That verb you talk about – “processThis” – if it is indeed uniform, then it has semantics that are, for the purposes of this discussion, identical to HTTP POST. And you can build a perfectly RESTful system with just that one verb, and it would be accurately characterized as loosely coupled and based on document exchange. In fact, I just built one a couple of weeks ago like that for pumping three channels of seismographic data from a digitizer to a communications controller. It’s just that in my experience, for most applications, you can build a better RESTful system with additional verbs, in particular GET. For example, if there were any need to have my comms controller initiate the request for data, I’d use GET.
The other approach to requesting data is to, for example, submitting documents which describe query terms or templates to services which process them and return the results. And that also has its place in RESTful systems, including ones I’ve developed. But when you don’t have to query, e.g. when the endpoint identifier for the service/resource to which you POST that document identifies something stateful, then GET is perfect for retrieving that state (such as the aforementioned seismic data). And of course, URIs makes the perfect endpoint identifier because they’re easily identified and contain all the information you need to grab the data (at least the good ones do 8-), without needing to know how to do a query on them.
Anyhow, this discussion is super-duper encouraging from my POV. We’ve come a long way from early Web Services Architecture WG meetings where folks were claiming things like “GET, PUT, and POST are only good for browsers”, or “REST requires humans”. A long way, indeed.
My one year contract with Nanometrics wraps up shortly, so I’m out looking for new and interesting things to work on. I’m particularly interested in any large scale (aka Internet scale) data integration projects. Any vertical would be fine too, though I’m most experienced with telecom and wireless. I’m also happy to work either with the companies that have the integration problems, or with an ISV developing software that targets those companies.
Getting back into standards work would be fun too.
My resume is available; I’ll be updating it later.
Robert sent me a great email that deserves to be on the Web. He doesn’t have a weblog, but gave me permission to post it here.
Might be of interest to you, if you haven't seen it already. Came across this justification for constrained interfaces and unified namespaces...it ties the argument to economics and the "Wealth of Nations" (Adam Smith). Reiser File System v4 ( http://www.namesys.com/v4/v4.html ) "The expressive power of an information system is proportional not to the number of objects that get implemented for it, but instead is proportional to the number of possible effective interactions between objects in it. (Reiser's Law Of Information Economics) This is similar to Adam Smith's observation that the wealth of nations is determined not by the number of their inhabitants, but by how well connected they are to each other. He traced the development of civilization throughout history, and found a consistent correlation between connectivity via roads and waterways, and wealth. He also found a correlation between specialization and wealth, and suggested that greater trade connectivity makes greater specialization economically viable. You can think of namespaces as forming the roads and waterways that connect the components of an operating system. The cost of these connecting namespaces is influenced by the number of interfaces that they must know how to connect to. That cost is, if they are not clever to avoid it, N times N, where N is the number of interfaces, since they must write code that knows how to connect every kind to every kind. One very important way to reduce the cost of fully connective namespaces is to teach all the objects how to use the same interface, so that the namespace can connect them without adding any code to the namespace. Very commonly, objects with different interfaces are segregated into different namespaces. If you have two namespaces, one with N objects, and another with M objects, the expressive power of the objects they connect is proportional to (N times N) plus (M times M), which is less than (N plus M) times (N plus M). Try it on a calculator for some arbitrary N and M. Usually the cost of inventing the namespaces is much less than the cost of the users creating all the objects. This is what makes namespaces so exciting to work with: you can have an enormous impact on the productivity of the whole system just by being a bit fanatical in insisting on simplicity and consistency in a few areas."
I forgot about this argument, held by some, including Stefan;
Mark follows up with asking what a successful response to a message in “contract mode” would look like, and that an HTTP 200 response code would not have semantic meaning with regards to the message. Right, it doesn’t – that’s why the response code doesn’t matter as much as the response’s content itself, and why we have things such as WS-ReliableMessaging.
The issue with not using a response code is that it’s impossible to distinguish between success and failure in some cases. Consider a portType/interface/operation of “getLastFault”, whose intend semantics is, as it sounds, to return the last fault sent by this service; how do you know if the fault is a result of trying to find out what the last fault was (which is a real fault), or if that really was the last fault (which isn’t a real fault)? That’s why message metadata needs to be orthogonal to content.
But anyhow, the 200 example was just illustrative of the point that the contract changes depending upon which style of WSDL you’re using. So unless there’s some bit in the message which says which interpretation is intended, then the semantics of that interaction is ambiguous. And even without that bit, the contract itself is ambiguous.
Savas goes beyond the call of duty, and responds to two of my posts simultaneously. I’m glad he did, because the “uniform” question I asked came about as a result of my initial response to him. He writes;
However, doesn’t the consumer now need to understand the specification of the type that governs the returned document’s structure? Isn’t this similar? In fact, I think it’s worst because the service may return information to a consumer that the consumer did not understand. Hence, we allowed an interaction to take place for nothing.
I’d say that the capacity to receive data you don’t understand is highly underrated. 8-) What you might not understand now, you might understand later.
So, my view on this is pretty simple; stuff that changes shouldn’t be fixed in a design-time element like a service description. IMO, that includes the schema(s) that a service inputs and/or outputs. Which brings us to …
One of SOA’s principles, as described by Don Box, suggests that services will only interact with one another only when they have determined, based on policy assertions and metadata about the document structure of the information, that they can meaningfully exchange information.
Ok. But as I see it, until some advanced form of AI comes along, a service will only ever interact with other services that conform to the preconceptions hardcoded into it. A constrained interface (the uniform interface, in particular) shines when you keep this in mind, because it suggests that any service can exchange information with any other service. If Web services don’t do similarly, then they have necessarily forged islands of interoperability (where each island is comprised of all services that share the same preconceptions).
WS-I heralds breakthrough architectural style
Thursday April 1, 2004 6:00 am ET
San Jose, USA. -- The WS-I today announced the creation of the "NULL
architectural style", the next step in the evolution of large scale
distributed systems. It is a generalization of so-called Service
Oriented Architecture ("SOA") styles, which have so far been used
primarily behind corporate firewalls.
The NULL architectural is characterized as having no architectural
constraints, providing the ultimate in flexibility for today's dynamic
enterprise. "By removing all architectural constraints, we have built
the perfect base from which to develop dynamic, loosely coupled systems",
said Ben Dover, WS-I spokesperson. "It seemed the next obvious step.
First, we improved upon the Web by disregarding its most important
constraints. So clearly, disregarding the constraints of Web services
themselves must be an even better idea!". He added, "As one can do no
better than to have no architectural constraints, this is quite
literally the ultimate in architectural styles!".
About the WS-I
Waste Services Incorporated is a 'Total Waste Management' company
providing a complete, integrated and professional range of waste
management services including waste collection, ... Erm, no, that's not
us, really. We do profiles, not waste. Yah, that's the ticket.