re:Modeling Address using Relational Theory

Coming in after 5 pages of posts is a little difficult but a few points from my experience are:

1. Line based addressing is a time bomb as a business user will come back later for some sort of geographical based reporting that will require parsing every line.
2. Address formats are determined nationally but all the elements can be analysed and defined. Formatting would have to be handled programatically using rules based on target country.
3. Input of address data is normally by very junior clerical staff (or direct by customers) without any understanding of the technicalities. Lots of scope for data quality issues
4. In the UK at least issues of local pride may mean that customers can get upset when mail is not addressed in the format that they specify.

The practicalities of these requirements mean that implementation ends up as a hybrid between structured and free format addressing.

Is it worth modelling addresses "properly"? I think it is as the designer should always understand what compromises are being made and what the risks and issues of those compromises are. Just don't expect to implement the model into a database exactly as it stands. Received on Sat Dec 03 2005 - 09:10:54 CET