About primary keys:

Are they still a good idea no matter how complex they become or how irrelevant they are?

Example:  I have this calculation table (when used with joins and group by it does wonders )

ComputedID as int,

SrcID as int,

scale as float,

offset as float

I could make a PK field, but why except maybe for making it slightly easier to make an editor)?

I also have a table that is something like

CustID int,

BudID int,

AcctID int,

BldgID int,

SrcID int,

Dt DateTime,

Reading float,

ReliabilityID int

The PK would be (CustID, BudID, AcctID,  BldgID, SrcID, DT)... but does it make sense to make a PK that complex, esp since I don't plan to pull things out in that order (mostly, select Dt,Reading,Reliability order by DT where xxxx)?

Thanks,

Thor