By and large, I agree with the practice of closing questions which refer to specific datasets (Matt Parker very nicely summarizes the issue in his answer here).
Nevertheless, it's undeniable that some datasets hold a unique place in the realm of Data Analysis, transcending their particular (specialized) domain (e.g. Fisher's Irises, Old Faithful data). Shouldn't these be treated differently?
I realize that it's hard to draw a line between these exceptional cases--perhaps an easy rule-of-thumb would be: "If it's built into R then it's ok."
Anyway, this is just something I was thinking about earlier and wanted to hear what you guys thought about it--as I think about it more I'll update this post accordingly.
[Note: I also think that Cross Validated could benefit overall if it even outright encouraged the use of these types of built-in datasets in the questions themselves. But that may be a whole different meta-post entirely...]