[techtalk] XP vs UML

BiOFH biofh at biofh.org
Fri Mar 30 19:16:01 EST 2001

Couldn't help but respond with an example of how the methodologies 
look from where I sit. This is, of course, just the way things happen 
in my group... but I'd be surprised if it's not a fairly common 

During time period A a project comes across my desk. Help identify 
and validate an external vendor package for a new knowledge base 
system for internal use. We spec out what we need down to every field 
in the database.

Time period A+x  the project is killed by the powers that be as too 
expensive. I decide that I can prove my feeling that "this can be 
built internally for next to nothing". I write the basic framework, 
build the database, build a front-end, back-end and search system. I 
add in many of the features that were included in the initial spec to 
show it can be done. This takes about a week working in my spare 
time. <insert sound of crickets chirping as no one bothers to look at 

Time period A+x+y a formal team is put together to "make it happen" 
internally. I beg out of the project since I really don't have the 
time now (and I know how this works....). I do, however, mention that 
"you know... I've already done this once...".

Today - the team is still hashing out spec (despite the original 
specs). They've moved up to doing some framework layout designs now, 
but no coding has begun, of course. Yadda yadda yadda. We'll probably 
see something tangible by year's end. Maybe.

So, I guess my point is pretty much the same as Jenn's, don't 
over-design or re-invent the wheel. When it's time to "just do it", 
don't let your methodologies run your life. If none of this pre-work 
had been done on this project, I'd heartily recommend using a formal 
design model to the team, but... "A man who runs in place only makes 
deep holes".

While I love the freedom of coding on the fly, but only if it makes 
sense. You can waste just as much time and resource with either 
methodology, you just have to be willing to admit that either one 
'can' work and neither is the be all and end all. Some of the 
greatest things we've produced have been done by one person over a 
weekend and some of them have been meticulously worked out by 
seemingly endless spec'ing.

I should stop rambling, too. ;)

More information about the Techtalk mailing list