Jeff Moden (2/21/2015)
I guess my point is that while there is the difference in Nullability, an NC-PK isn't really going to have any I/O advantage over a Unique NCI as the user thinks. The confusion between "PK" and "Clustered" may be exactly what is article-worthy here because a lot of folks think that a PK has to be the Clustered Index.
Good point! Would it be worth while touching on the storage difference between CL-PK vs. NC-PK, Unique Index and Unique Constraint or rather that the latter three are all implemented as NC indices? Somehow have the recollection that this has already been covered.
😎