Ok, so you have this cool sounding technology called SPOT. People are impressed that you, Microsoft, have come up with a cool way to send information to your watch. So what do you do? Hide the technology under a crappy name like MSN Direct. So does that mean I have to have MSN? What if I already have AOL? Every part of Microsoft that I have come into contact with has impressed me, except for marketing… marketing causes Microsoft more headaches that any other department. (Just think about .NET) [.Avery Blog]

As someone who has spent most of the past three years explaining .NET to customers, I’ve given up on the marketing department. I tell customers point blank: “Microsoft is the company with the great technology and bad names”.

MSN Direct and .NET aren’t even the worst cases of this:

• Argue/flame if you like, but MTS was the first of the modern application servers. (The original version of EJB was essentially a clone of MTS.) But did we make a big deal about it? No, we shipped it as an “additional technology” included in the NT 4 Option Pack. As if it was optional.
• Windows 2000? Is that the new version of Windows 98? 😦
• MTS at least stood for something. What the heck does COM+ stand for?

Brad Abrams blogged on the importance of measuring performance before making design changes. “Have you measured it?” is a question I ask customers all the time. Recently, I was meeting with an ISV that had been directed to me regarding improving the performance of their .NET app. They were using .NET Remoting between their web and app tiers. They were using the HTTP channel with the binary data formatter and felt that if they moved to the TCP channel they would improve their performance dramatically. However, making this switch would entail writing a non-trivial amount of “plumbing”. They would need their own service to host the objects instead of just using IIS. They would also need to build custom code to support encryption and authentication on the TCP channel (I know there is democode for doing this floating about – but as an ISV, they would need to support the code as if they had written it which is still non-trival).

It turns out they had not, in fact, measured the impact. And in digging deeper in their architecture, it turned out that they were making multiple calls between their web front end and application tier on each web page request, with the average being about five. While TCP is in fact faster than HTTP, it’s not going to provide the same level of improvement that collapsing those five remote calls into one would be. I explained to them the Remote Facade and Data Transfer Object patterns as a way to dramatically improve the performance of their application while avoiding the need to build/maintain a bunch of plumbing. They liked it and as far as I know, they decided to head in that direction. I’m waiting to hear exactly how much of a performance gain they get.

If you don’t measure, you don’t know where the bottle necks are and you don’t know where to get the greatest impact for the least amount of effort.