I've had an ongoing discussion with several programmers regarding the merits (or lack of merit) of using data aware widgets. That is to say, third party developed UI controls, which have a lot of functionality to them and often times are responsible for updating critical data based on user actions.
The long and short of my position is that "yes it is quick and easy and of course I will always have a use for such things but they do fly in the face of the principles of a model view architecture." By definition having a ui control (especially one with code that you don't have access to) control data, increases the likelihood of data errors whether or not you send the data through a middleman. The more of these controls you use the less reliable you data will be regardless of whether the control offers methods to validate data before sending it along or not (a "beforeUpdate" thingy etc..) because it makes no requirement that these methods be implemented.
By my experience, the use of these controls seems to be a matter of gaining a gradual understanding of exactly how they react to user actions and "plugging all the holes" where unwanted results can leak through. I think this is bound to be fraught with peril.
Previous Topic
