The question was raised, "Can a PDF document be a requirement?" During the same conversation, it was implied that a requirement must be some form of text. I think that while most requirement management tools do emphasize text, we should be looking more abstractly for this interface. Here are some thoughts, and please add your own:

The key I believe we're getting at is that a requirement must have a precise description--it has to be able to be communicated through some persistable artifact to be usefully managed by tools. (See Mike Cohn's book on Applied User Stories for a different point of view). But there is no inherent reason that precise description has to be text. In some cases, requirements can be precisely communicated through a modeling language in a diagram, or through an annotated simulation, or other ways. Sometimes text will be the most precise, but we're all too familiar with the shortcomings of text for precise descriptions. Implication: When defining the XML representation of a requirement, there should be a tag for <PreciseDescription>, but the value of that tag could be a complex type other than text or even a link to some complex artifact. Maybe it needs an attribute for the type of description???

Next, even if it is text, we should abstract it from the format in which it's "written down". This gets to the question raised about "content-type" header vs. different URLs. Consider this:

I have a requirement written in a word document.

I convert that word document to a PDF--is this a new requirement?

I read the requirement out loud and record it as a sound file, and attach that to an e-mail. Is this a new requirement?

I translate that requirement into Japanese---is this a new requirement? It shouldn't be, but what if I translated it wrong, or the translation has nuances that make it slightly different.

I automatically convert the text into a UML Activity diagram so it's easier to understand the flow of the steps. Is this a new requirement?

Well, I would say a headline and a basic description containing only text should be an anchor. Additionally it should be possible to attach any number of wished diffrenet representations. If I translate the requirement in some way, why should I ommit the original from which I did translate it? Wether its a technical translation into UML or a literal translation into a different spoken language.

So to your question, is it a new requirement, no, is it a new version of that requirement? Yes.

-- TorgeKummerow - 22 Aug 2009