Комментарии (3)

1 Oldtimer68 оставил комментарий Постоянная ссылка

Automobile manufacturers only deliver a car to the showroom if they know that it can be driven around the block with a car salesman in the backseat and get back to the car lot to entice a sale. Software should be the same :)

Agile is OK if driven by sound Software Engineering Principles not MBA Manager Hubris. Back up from the external financial deadline to a time point where there is a solid statistical probability that you can reach the above objective and start then if you can afford it, otherwise do something else and complete the original concept on a realistic affordable deadline at a later time. If you follow the first sentence in this paragraph that time may be sooner than you think.
If testing isn't augmented by a solid one entry one exit modular design with exception whenever the logic predicate is violated and proper use of structured program representations of your directed logic graphs then you will have tested something but what it is is up for conjecture? Designing the logic predicates is the hard work.
All external data needs to be validated for correct domain and correct associative reference prior to entry into your system.
When, what,where,coded to a (collected reason data set ) logging needs to be appropriately included or maintaining is an order of magnitude harder.
Testing alone is never a complete paradigm:

2 KeithCollyer оставил комментарий Постоянная ссылка

Bruce, I couldn't agree more. It always annoys me when agile proponents claim that "traditional" approaches mean you have to know everything up front and then follow a rigid plan. To me, that just says that they have never actually worked inside a real development organization. As you say, the plan is wrong. To quote George Box "all models are wrong, some models are useful". He was talking about statistical models, but the generalization to other forms of model is true, and a plan is a model.

Добавить комментарий Добавить комментарий