Subject: RE: Topic: Just Works and 1.0
From: Paul Rohr (paul@abisource.com)
Date: Fri May 04 2001 - 15:40:40 CDT
At 10:19 PM 5/3/01 -0700, WJCarpenter wrote:
>Really, I don't think it makes that much sense to talk about
>complicated things like tables or notes and citations as single
>features.
For us as developers, no. The perspective below is much more useful.
However, I'll assert that most users who have any experience with similar
products don't care.
For them, any of the features you mentioned above *are* atomic. Either
{ tables | notes | citations } Just Work, or they're gonna be somewhere
between disappointed and annoyed.
The fact that it's a ton of complicated work for us to reach that point is
*our* problem, not theirs. As far as they're concerned, it either Just
Works, or it doesn't.
>They're really a composite of interwoven featurettes. They
>can quite often be implemented a few featurettes at a time with a plan
>for the whole thing. In fact, in my experience, the ability to
>decompose the implementation is one of the signposts that point to a
>good design and architecture.
Yes, yes, yes. Absolutely. Getting that composite interwoven properly is
the real artform here.
One possible approach when facing a complicated problem is to just whittle
away at it from the edges. Do what you can, then do some more, then when
you hit a wall, stop and rethink. Sometimes you have to throw out
everything and start over.
I share your preference for a different approach. Figure out what it'll
take to do the *whole* job, break that down into reasonable chunks, and then
start phasing them in. Do some of the hardest ones first, so you know the
approach will work, then ask for help to parallelize all the easy bits.
It's a lot easier to convince me to ship partial implementations when I know
that we're following the latter approach. Doing it the other way makes me
nervous.
Paul
This archive was generated by hypermail 2b25 : Sat May 26 2001 - 03:51:02 CDT