On 10/26/07, Tracy R Reed <[EMAIL PROTECTED]> wrote: > Christian Seberino wrote: > > But why is it so important to have the new code be at parity with old > > spaghetti code at all times? What's wrong with it being deficient for a > > while until it gets banged into shape? > > Nothing necessarily wrong with that. You can let the regression tests > fail for a while if you want. But eventually you will want to come back > and fix that stuff and how will you ever remember what it was that > needed fixing? It is important that the final product be at parity with > the old code in a case where you are releasing a new version of software > which is just a minor revision number different and therefore should be > completely compatible with the previous version so nobody's stuff > breaks, just for one example. >
Maybe the key question is this. Does anyone ever write a new system these days? If all we ever do is morph old systems then perhaps all we should do for a few years is write enough unit tests to bring those old systems up to snuff then start from there on the endless enterprise of extension. I have not bothered to look at the source code for say Python or Ruby but I would hope that each is mature enough to have accrued a large number of unit tests. Is anyone familiar with the facts of this situation? Gotta run now, maybe I will look into it later. BobLQ -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
