Tina Norris Fields
tinanfields at torchlake.com
Thu May 8 09:15:05 CDT 2014
Thank you, Arthur, for showing me (once again) something I had not thought of. I will incorporate that concept of the Point in Time into my work from now on. TNF Tina Norris Fields tinanfields-at-torchlake-dot-com 231-322-2787 On 5/6/2014 5:41 PM, Arthur Fuller wrote: > And none of this addresses the PITA issue (Point In Time Architecture), > which basically says that the traditional relational concepts of Insert and > Update and Delete are destructive of what data existed a moment ago. That > is a very significant problem and requires a whole lot of planning, but to > oversimplify, it means that Updates and Deletes are forbidden, and that > every table in the DB be able to handle this. That means that every table > has a pair of date fields (call them EffectiveDate and EndDate), and that > all queries respect these fields. For example, take your average > 50-year-old medical patient: she has probably had a few family physicians, > optometrists and so on, due to moving to Seattle for a new job and a year > or two later it's Denver. > > What this sort of database requires is the ability to wind the clock back > to 2002, say. That is flatly impossible with the traditional Insert, > Update, Delete model. The only way this can be handled is the PITA model: > every table has EffectiveDate and EndDate columns and every query and > stored procedure respects these columns. Who was your dentist in 2004? Who > sold you that Ford that you gave to your daughter in 2012? Where did you > live in 1999 and who was your physician at that time? > > > > Arthur > > >