Susan Harkins
ssharkins at gmail.com
Sat Mar 14 14:24:10 CDT 2009
The first is the implication that any null column cannot be > indexed, or even if the system allows it, the index will never be used. =======What you're suggesting is that even though SQL Server allows the index to be set, it won't use it -- I'd be surprised if that's the case and not surprised at all -- paradox... but we're dealing with MS. > > At any rate, if the conjecture is true, then by extension one should > arrange > the columns in order of frequency of use.(in indexes and queries). I think > that for a small database, a few hundred megs let's say, the performance > difference will be minimal even if the conjecture is true. But medium to > large databases (20 gigs to a terabytem say) the difference, if any, might > be significant. ========Well, let me say this -- I will be surprised if this matters. You'd think, if this were really the case in SQL Server, that it'd be well known already. Susan H.