I was honestly hoping this was a joke, but I don't think it was. By the time someone writes an accurate spec, the business need has already changed. Spec's don't work.
Maybe you are talking about running your own start-up where it might not be successful yet? And you're trying to find your market fit? Then sure, maybe just writing new code and changing the product on a whim is what is best without writing up a spec.
Now consider a software consultancy with enterprise customers. The enterprise usually wants time and cost estimates, which usually means agreeing on precisely what you will be delivering in advance. You WANT to write a spec, to prevent scope creep and limit your liability, and they want you to tell them how long and how much (and a spec helps estimate those things). It's win-win.
Most enterprise customers can't wrap their CFO's head around true agile with iterations and open-ended times and costs. They want as much of a guarantee as you can give them for what/how long/how much.
Not saying one way is better or worse - but maybe it's a case of the right tool for the job?
We use specs because our customers tend to be somewhat staid and want certainty and predictability. By keeping chunks of functionality relatively compact (say 3 person months typically, then developed over a period of maybe 6 weeks) writing an accurate and useful spec is absolutely doable.
Doing that we'll usually be 95% (or more) right and any adjustments can be made in a small release a couple of weeks later.
The approach has elements of agile and iteration in it (small releases, the UAT and fix process is basically a short agile sprint) but specs are absolutely a working part of it. Could we do it another way if our clients would accept it? Sure? But do the specs work in the current process? Yes they do.
If you get your feedback on iterations in writing, you will probably get fewer self-contradictory or illegal, immoral, unethical, or unimplementable demands.
What?
Since when did specs become immutable? Needs have changed? Patch the spec. It's faster than patching the software.
From http://joelonsoftware.com/articles/fog0000000036.html:
"The most important function of a spec is to design the program. Even if you are working on code all by yourself, and you write a spec solely for your own benefit, the act of writing the spec -- describing how the program works in minute detail -- will force you to actually design the program."
Sure, you'll still find out the design was wrong in some areas, and you'll have to change it. But at least you'll have a design to start from, as opposed to a hodgepodge of features carelessly tossed into the interface wherever the programmers found it most convenient to stick them, implemented in whatever way made the most sense to the programmers at the time.
I don't understand how this can be faster. You patch the spec, then patch the software to match. I can't see how the spec could be patched faster. That is unless you patch the spec in isolation, prior to writing any code, which is a bigger mess. I've never, ever seen any group that can write a spec in isolation prior to writing code to demonstrate it, and have it come out well, its usually a disaster.
The idea is by having a spec, you're more likely to see where it needs to be patched upfront, before delivering poorly designed software to angry customers, so you spend less time re-working it later on.
> I've never, ever seen any group that can write a spec in isolation prior to writing code to demonstrate it, and have it come out well, its usually a disaster.
Can you elaborate? I'd be interested in hearing what they were building, who (in general terms) they were building it for, how they went about writing the specs, and what went wrong.
My experience has been completely different, and I'm curious as to why that is.
Edit: Of course you can't expect the customer to write the spec. Designing the software is your job. But that doesn't mean specs are broken; it just means asking your customers to write your specs for you is broken.
