Iterative Package Development

Posted on 2017-04-18 by Chris Dornan

Here is the thing. Between the release of regex- on 2017-02-18 and the release of regex-1.0.0 on 2017-04-15 we released 22 revisions of the package, excluding regex- itself, each of these revisions being driven by community input.

And I have no doubt that regex- has a somewhat better API than regex-

You could take the view (and I suspect that you do) that regex- was a shoddy release and that I should have taken longer to get the API right, and avoid the confusion and loss of confidence by making all of those releases, and there is a part of me that agrees with you.

But is this right? I mean I know that some people can get it right first time but I am clearly not one of those developers. When it comes to finding the right design for a modestly complicated API like regex it takes a while and a number of attempts.

(One small digression. Do I believe that regex should try to do less? No! I have reviewed the scope of the package many times and for sure I could have saved my self a lot of work by trying to do less, but I am convinced that it would be at the expense of the utility of the package. I might of course be still missing some tricks to making the API still better.)

So, is iterative development to be encouraged or discouraged for package developers? Should we announce that a package is WIP and go to the community for input and use an iterative development model to bring it in?