Hi there and welcome to the 32nd interview in my series of chats with thought leaders in the “connected technology” space. This month, we are talking with Pablo Cibraro who is the Regional CTO for innovative tech company Tellago, Microsoft MVP, blogger, and regular Twitter user.
Pablo has some unique perspectives due to his work across the entire Microsoft application platform stack. Let’s hear what he has to say.
Q: In a recent blog post you talk about not using web services unless you need to. What do you think are the most obvious cases when building a distributed service makes sense? When should you avoid it?
A: Some architects tend to move application logic to web services for the simple reason of distributing load on a separate layer or because they think these services might be reused in the future for other systems. However, these facts are not always true. You typically use web services for providing certain integration points in your system but not as a way to expose every single piece of functionality in a distributed fashion. Otherwise, you will end up with a great number of services that don’t really make sense and a very complicated architecture to maintain. There is, however, some exceptions to this rule when you are building distributed applications with a thin UI layer and all the application logic running on the server side. Smart client applications, Silverlight applications or any application running in a device are typical samples of applications with this kind of architecture.
In a nutshell, I think these are some of obvious cases where web services make sense,
- You need to provide an integration point in your system in a loosely coupled manner.
- There is explicit requirements for running a piece of functionality remotely in an specific machine.
If you don’t have any of these requirements in the application or system you are building, you should really avoid them. Otherwise, Web services will add an extra level of complexity to the system as you will have more components to maintain or configure. In addition, calling a service represents a cross boundary call so you might introduce another point of failure in the system.
Q: There has been some good discussion (here, here) in the tech community about REST in the enterprise. Do you think that REST will soon make significant inroads within enterprises or do you think SOAP is currently better suited for enterprise integration?
A: REST is having a great adoption for implementing services with massive consumption in the web. If you want to reach a great number of clients running on a variety of platforms, you will want to use something everybody understand, and that where Http and REST services come in. All the public APIs for the cloud infrastructure and services are based on REST services as well. I do believe REST will start getting some adoption in the enterprise, but not as something happening in the short term. For internal developments in the enterprise, I think developers are still very comfortable working with SOAP services and all the tooling they have. Even integration is much simpler with REST services, designing REST services well requires a completely different mindset, and many developers are still not prepared to make that switch. All the things you can do with SOAP today, can also be done with REST. I don’t buy some of the excuses that developers have for not using REST services like REST services don’t support distributed transactions or workflows for example, because most them are not necessarily true. I’ve never seen an WS-Transaction implementation in my life.
Q: Are we (and by “we” I mean technology enthusiasts) way ahead of the market when it comes to using cloud platforms (e.g. Azure AppFabric, Amazon SQS, PubNub) for integration or do you think companies are ready to send certain data through off-site integration brokers?
A: Yes, I still see some resilience in organizations to move their development efforts to the cloud. I think Microsoft, Amazon and others cloud vendors are pushing hard today to break that barrier. However, I do see a lot of potential in this kind of cloud infrastructure for integration applications running in different organizations. All the infrastructure you had to build yourself in the past for doing the same is now available for you in the cloud, so why not use it ?
Q [stupid question]: Sometimes substituting one thing for another is ok. But “accidental substitutions” are the worst. For instance, if you want to wash your hands and mistakenly use hand lotion instead of soap, that’s bad news. For me, the absolute worst is thinking I got Ranch dressing on a salad, realizing it’s Blue Cheese dressing instead and trying to temper my gag reflex. What accidental substitutions in technology or life really ruin your day?
A: I don’t usually let simple things ruin my day. Bad decisions that will affect me in the long run are the ones that concern me most. The fact that I will have to fix something or pay the consequences of that mistake is something that usually piss me off.
Clearly Pablo is a mellow guy and makes me look like a psychopath. Well done!
Richard Seroter's Architecture Musings 
One thought