How many times have you come across an interface design that leaves you wondering if the designer has ever used the interface themselves? I have one for you and I'll omit the vendor name and product so that I can communicate these facts without worrying about someone getting upset and missing the point altogether. Ultimately, we want to learn from mistakes and fix things.
There is an IP phone on the market that is very full featured. However, there is one feature that is highly dysfunctional. It would not be obvious if you just read through the documentation but very obvious when you use the phone.
The unit has a set of soft-buttons that are dynamic; their functionality is based on the context of your operational state (on call, directory lookups, settings, etc) . For example, If you reach over and adjust the volume up or down, a soft button changes to 'SAVE' so that you can make this adjustment permanent to your preferences. If in 3 to 4 seconds, you do not hit the soft-button, it changes back from 'SAVE' to its original assignment which is........END CALL.
Yup, you guessed it, if you hesitate and hit this button too late instead of saving this newly adjusted parameter, you will issue the command to end the call. Where is the interface police when you need them!
When dealing with soft-buttons, I think there should be an analysis performed so that functional classes are grouped such that the chances of you accidentally hitting the same button for wildly different commands are reduced.
What if there was a mouseover type of button on your browser that sensed that if you were in spell correct mode it would bind itself to 'learn word' and if you did not act on it in a few seconds changed back to 'erase all'. Bad design, no donut.
I have to just put this in the category of Interface No-Nos because I have not read any research to allow me to place it in any formal category of interface design. I'm sure you have a bucket of interface no-nos too. The trick is to always learn from our mistakes.
--tk