37 Signal's Getting Real convinced me that wireframing and writing functional specification documents are a middleman steps unnecessary for building web applications and dynamic websites.
Is the overhead for these steps worth its weight? Is prototyping in HTML/CSS or even PhotoShop documents (so designers can work on them directly) a better option than using software like Visio? Personally, I am swaying towards the latter but am not sure.
37Signals advocates skipping even Photoshop and going right to HTML. See http://www.37signals.com/svn/posts/1061-why-we-skip-photoshop. I agree with their assessment of pre-planning. I don't think its worth the time in the long run when you could be spending time building a working prototype in HTML/CSS/JS.
It probably depends on who you're working with. If it's you and a designer, then a functional spec might be too much trouble. But, in my job, the executives want to know precisely what they're going to get at the end of a project and so we've had a real hard time implementing iterative development. Usually the iterations are defined with wireframes, functional specs and mock ups..:)
"Fail to plan is plan to fail" - or something like that.
Wireframing is not limited to web-apps; it is pervasively used wherever a high-level overview of any system is needed (it's just called something else).
Functional specs, when you know what is to be done & how to do it would indeed be overkill. A high-level diagram of your intentions would suffice. And it will never be unnecessary. It primarily helps you focus on the scope and intent/target of what you want to do.
The focus should be on preventing wasted effort - finding out half-way through that something essential, that impacts on all other objects, are missing is not what you want to discover. Wireframing in this case would assist in detecting most major functional needs. You would only need to elaborate on functional spec where absolutely needed. Using Photoshop to design your will also be 'wasted effort" - far better to use evolutionary prototyping (RAD technique) with CSS/HTML - but still do pen & paper mock-up of your intentions.
I believe it depends on how well you understand what you are trying to do. If you are working for a client and they haven't expressed much in the way of requirements you may want an approach with extremely quick iterations. If you already have a good understanding and can produce something more substantial without worrying about throwing it away because it was the wrong direction then more time can be spent. Either way, a clickable prototype can go a long way in determining what the real site needs to be in the end. If you can get agreement on the prototype, then when your application matches the prototype you know it is complete.