I came across a post from Claudia Imhoff.  I totally agree that in some cases cleaning data also destroys some of information that can be derived from dirty data, and can destroy the audit trail.

I usually insist on the option of minimal duplicates, this is always useful when converting things like dates and numbers from flat files, especially as the flat file data type is usually character based and that conversion process is rarely 100% reliable either due to dirty data such as 31 Feb 2007, or poor meta-data, is 05/04/2007 the 5th March 2007 or 4th May 2007.  Depending on the localisation of the tool generating the data it could be either.

I really like the sound of ABC tables and would like to hear from anyone with experience of implementing this, or if anyone knows of any tools for achieving such a solution.

If like me you could not make it along to the SQLBits conference, you do not have to completely miss out as the slide are all available here

Having started my life in AI and programming using PROLOG and LISP, I am please to see this announcement that not only F# going to continue to evolve and feed into mainstream programming, such as the new LINQ features of .NET 3.5, but is going to become a fully supported visual studio based language.