Let me try to be more clear; my objection to your original statement
(paraphrased as "we've got lot's of space; why should a couple of bytes
matter?") is grounded NOT in practical matters, but rather issues of
design theory and discipline. I would agree that in terms of
performance you'd be hard pressed to find differences between char and
varchar for 90% of use cases that you can come up with. Char is useful
in other ways besides just performance (e.g., validation, but that too
can be emulated by other means); however, I'll concede (as I did in my
original post) that distinguishing between char and varchar IN MOST
CASES is moot in terms of performance.
However, my point is this; if you disregard the differences here
because of substantial resources, why bother with any sort of datatype
at all? If a varchar(50) will do the trick of a char(10), what's 42
extra bytes? In the days of terabytes sized RAID arrays, why not use a
varchar (8000) for everything? Why use an integer if a varchar will
do? Obviously at some point, there will be a performance impact; why
not be disciplined enough to strike at the lowest level of that curve
and use the least expensive resource that is the best fit for your data?
Received on Thu Aug 18 2005 - 17:31:39 CDT
