Again, I did not say that, not in the blog post, not in this thread. I said, that I did not conclude that the results in the blog post should "all but put the nail in the coffin on integers vs naturals keys." (your quote, not mine).
That was a terrible sentence from Gail. I wish she'd stop that mess, but I know she won't. It implied she didn't have a coffin getting nails in it, but I found the text.
While this may not the final nail in the coffin for natural keys, it is worth keeping in mind when choosing between natural and artificial keys for a system, especially one likely to process large numbers of rows, such as a decision support/datawarehouse system. Test carefully with expected data volumes and expected loads if you’re considering natural keys and decide based on the result of those tests.
I'm going on record here as objecting to this coffin, the announcement of the death of Mr. Natural Keys is premature.
Here's the deal, there are some folks who claim value to the use of natural keys, I'm one of them, but I think the value behind the use of natural keys might not be as easily flushed out by the scientific method as presented by Gail, in fact, many of the challenges programmers face don't fit the easily measurable experiments demonstrated by the article.
Additionally Gail makes another mistake for conflating "data type" with the natural / surrogate key classification. I'm not saying that the set of natural keys that can be held in the integer datatype is the same set that can be represented by varchars because that would be just as wrong as Gails article, however, its clear that "datatype" describes one concept, the natural / surrogate key duality describes another.
Look at some of the data warehouse texts out there, clearly folks like the integer data type for some columns, and with data warehouses, moving lots of data fast means we should really care about data types. But even with some of these "date tables", its clear to see that while we can represent the year as an integer (gaining some advantages), that integer still reads like a year so it shares some characteristics of being a natural key. Even an integer representation of the date (YYYYMMDD) can be viewed as a natural key despite the datatype. Am I to now be left with the conclusion that Gail doesn't even KNOW what a natural key is?
IN MY OPINION the article was hurt by the coffin paragraph, Gail fumbles on what being a natural key actually means, and she additionally doesn't expend much effort in enumerating which subdomains of our challenges can easily benefit from the scientific method and which can't. Opportunities lost, in my opinion. There are PLENTY OF SCIENTIFIC METHOD demonstrations out there that DON'T have these problems. There are also awesome discourses on the complexities of software development that AREN'T solved by counting milliseconds or for that matter, anything easily measurable. This particular article, well I regret reading and discussing it. Yet again, SSC disappoints.
THERE. I HAVE POSTED MY CAREFULLY CONSIDERED OPINION. YOU MAY NOW COMMENCE WITH THE PERSONAL INSULTS.