Justin du Coeur (jducoeur) wrote,
Justin du Coeur
jducoeur

Death to Specs


All right, I've had it. If I hear one more comment along the lines of, "Oh, but of course that feature is in the spec! It's implied by these three words here on page 12!", I shall lose my cool entirely and explode into little ranty programmer bits all over the place.

It isn't even that the people here are particularly bad at writing specs. The problem is that it isn't possible to accurately, completely and clearly describe a massively complex, user-interactive system with about eight hundred features, that doesn't exist yet. Both the design and implementation sides of the spec wind up making assumptions, and those assumptions frequently don't match. So we are finding ourselves at the end of the process, with the management-level designers finally trying the system out, and pointing out all the ways in which What We Built isn't exactly What They Thought They Were Asking For.

As soon as the current release cycle ends, we are going to have a long chat about processes. If we're going to be developing something this complex, there's just no substitute for an iterative process, with the designers working with the whole thing as it's being developed and finding the assumption mismatches (and just plain bad design ideas) quickly...

Somewhere, I have a button that says (more or less): "Building a system that conforms to the spec is easy. Writing a spec that clearly describes what you want is impossible." Truer words have ne'er been spake...
Subscribe
  • Post a new comment

    Error

    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 2 comments