Considering that you have admitted before to not regularly using masked arrays, I seriously doubt that you would be able to judge whether this is a significant detriment or not.  My entire point that I have been making is that Mark's implementation is not the same as the current masked arrays.  Instead, it is a cleaner, more mature implementation that gets rid of extraneous "features".  Instead of fussing around with a mask directly in the array, the user of masked arrays should now consider the use of views as the masks.  It works beautifully because it works off a well-documented and well-understood feature of numpy.

Of course, when you look at the feature in your way, with those expectations, then I would agree that it might be confusing.  But given that this is a completely new feature, then we have the opportunity to properly document and show how to rethink a user's pre-conceptions of masked arrays.

Users can keep the original array as a plain array and have mask1, mask2, mask3, etc as being separate views.  It is a completely different way to think of masked arrays, and considering that masked arrays are not widely used in other toolkits, I think we can be free to change the paradigm.  Further, there is no reason why we can't keep numpy.ma around for backwards compatibility and for those who "just don't get it".

Cheers,
Ben Root