Surrogate or natural keys…
A very interesting topic, exspecially as many frameworks (f.e. ruby on rails) tend to produce automatically surrogate keys. Lee Richardson tries to advocate surrogate keys in his blog-post. I hate surrogate keys, often calling people using them “M$-Access-Users”. But, to be honest, sometimes, exspecially when mapping OOP-Classes into tables, there is no other way. Best answer to Lee’s post seems to be Ian Batty’s comment. Let me cite one paragraph:
“Personally I am a natural key guy. Saying that, and this is where your example loses credibility, practical experience has shown that for a person there is no natural key, so I create a surrogate. However, if I am creating an entity that has a true unique identifier, that doesn’t change and meets all the other requirements then I will use it.”
Seems to be a very good general advice.