9 May 2011
One of the benefits of using web services is that it helps you to decouple various parts of a system. People can work on separate code-bases with some degree of separation. Although you get some decoupling, you cannot eliminate the coupling completely because the services still have to communicate to each other through their interfaces. The sad thing is that many teams make this coupling much worse than it should be.
The governing law for decoupled collaboration should be Postel's Law:
be conservative in what you do, be liberal in what you accept from others.
-- Jon Postel
In the case of collaborating services, one of the stickiest points is evolution. Although there are some people who believe that you should just get your service definitions right first time so you never need to change them, my regular readers will not be surprised to find me absent from their parties. In order to be able to evolve services you need to ensure that a provider can make changes to support new demands while causing minimal breakage to their existing clients.
A common way to stuff this up is to use some kind of schema-driven binding of your service endpoints. An example of this is code-generating C# classes from an XSD definition. This is touted as a time-saving feature - the service provider publishes an XSD definition of their service, the consumer takes a copy and generates a class. Look ma, no programming. It works well until the provider needs to make any change to the interface, such as adding a field. Adding field to an interface like this shouldn't be a breaking change for anyone - but often does break these schemes.
My recommendation is to be as tolerant as possible when
reading data from a service. If you're consuming an XML file,
then only take the elements you need, ignore anything you
don't. Furthermore make the minimum assumptions about the structure
of the XML you're consuming. Rather than use an XPath search like
//order. Your aim should be to allow the provider to
make any change that ought not to break your code. A group of XPath
queries are an excellent way to do this for XML payloads, but you
can use the same principle for other things too.
On top of this, make sure there's only one bit of code that
reads data payloads like this. One of the purposes of a Data Transfer
Object is to wrap the data payload behind a convenient object so the
rest of the system can just go
and be impervious to changes that would even break a tolerant
It's worth bearing this principle in mind even if your data transfer protocol is binary. Imagine you have java programs at both ends of the connection, and want to use a binary transfer to keep your message sizes down. Most people in this situation would use the built-in serialization mechanism of java to serialize objects directly, but then if one side adds a field the transfer breaks. You can avoid this pretty easily by first putting the data into generic collections (lists and maps) and then serializing those collections. If you add an extra field to a map, it will still deserialize on the other side and it's easy for a tolerant reader to ignore it.
To help the service provider evolve their service, you can then communicate which bits of the communication you are reading. A good way to do this is to send them the reader and its tests, so they can use them in their build process to detect potential breakages. Some of you may recognize this as the next step to Consumer-Driven Contracts
There is a full description of this pattern in Service Design Patterns.
My colleague Ian Cartwright posted a set of useful blog posts about this a few years ago. He points out that schema validation offers a false sense of security, and that there are dangers in serialization, both in general and particularly for domain objects.
Saleem Siddiqui describes how a Tolerant Reader works well with a Magnanimous Writer.