And it depends upon what kind of MBT you are talking about.  I have successfully used grammar based models (GBM) to saturate large databases with synthetic but realistic data; then other forms of testing can be conducted such as performance.  Another form of GBM also generates complex queries with expected results.

I discovered the MS Calculator Scientific Mode bug, (((((((((((((((0=))))))))))))))), with a GBM that generated large, random arithmetic expressions.  This string is a reduction of the rather large string that originally uncovered the defect in Windows 2000 (which still persists through Vista).

On the other hand, any model that requires maintenance when the product undergoes change should be examined carefully for value.

Alan A. Jorgensen, Ph.D.

Paul Rogers wrote:

ive done some mbt, all on web apps. The current app I work on doesnt seem particularly suitable ( its a data collection app, the nearest well known example would be google analytics) although maybe I just havent seen where a model could be useful yet.

The biggest problem in the past was what seemed like the daunting task of building the model. But if you do it just a bit at a time, possibly one screen or small functional area it turns out that it isnt as hard as you thought.

Ive written my own models, which were a layer above watir. I was one of the original authors of watir, so it may have been a bit easier for me.

Paul

On Wed, Apr 15, 2009 at 7:34 PM, lin.yaxiong <lin.yaxiong@...> wrote:

Just want to get a poll on the type of testing on what type of application you've found MBT is good (effective) for? Did you use any tool (comercial or open source) or any framework? What are the challenges you had to overcome to make MBT work?