Implementing Standard Protocols

I am writing this while waiting for the new version of XCode to download and install on my MacBook. This 3.5G install gives you time to run errands, grab coffee, or in my case - write a blog entry.

I often spend time talking about new APIs with executives and engineering teams. Most of the time executives and engineering managers understand the value of a well-designed interface, but when it comes time to fully supporting some standard you get some resistance. Today I'd like to dissect my experience in trying to convince people to sign up for the work it takes to fully support a standard.

90% of the time an enhancement to the API happens because a customer comes in and has an issue that can't be solved with the current set of interfaces. Features creep in one by one and in general the default behavior is to add just enough to solve the problem at hand. From a cost perspective it also makes sense to avoid building things you don't need right away.

In addition to the internal inertia there is also a school of thought of iterative design, except that iterative design is not such a good thing with the API. Companies that have integrated with you aren't so keen on going back and refactoring that piece of code on your schedule.

Signing up for the work to fully support a standard needs to be broken down into dollars and sense. First supporting a standard rather than adding a couple of features has its costs:
1) Standards have elements you don’t immediately need.
2) Standards take away some flexibility
3) It takes longer
4) Someone needs to read and understand the standard in order to implement it.
What are the benefits? Here are some benefits that go beyond engineering purism:
1) the design time put into a standard doesn’t need to be replicated internally
2) a lot of times there are test kits you can use to do automated testing
3) standard protocol generally have a good deal of documentation which eases the burden on the internal doc team.
4) Helping people implement a solution based on a standard will require less support and professional services involvement.
5) Implementing a good standard is good for developer marketing.

While the time it takes to understand and develop full support of a standard are very obvious costs. The benefits need to be explained. The things that worked best for me are #2 and #5. If you find a test kit and can shave time off the release in the testing department it’s a slam-dunk. If there are some apps or companies that already know how to work with a protocol and because of that are going to integrate with your service – that’s another easy win. Still depending on the cost of supporting a standard you might need to consider all five benefits and put a dollar or time value on those.

Good luck defending standards implementation. Remember – it pays off in the end!