- From: Michael Kay <mike@saxonica.com>
- Date: Sun, 23 Jan 2011 00:35:16 +0000
- To: "Costello, Roger L." <costello@mitre.org>
- CC: "xmlschema-dev@w3.org" <xmlschema-dev@w3.org>
>> In this paper I argue that XML Schemas should implement tight data constraints. >> >> >> More ... http://www.xfront.com/Tight-or-Loose-Data-Contraints.pdf >> >> The line of reasoning appears to be something like: * In my sample of N names, none of them contains non-ASCII characters * I can't be bothered to test whether the system works with non-ASCII characters * Therefore, I'll stop non-ASCII characters getting into the system. Now, I can understand that there might be projects where cost and timescale considerations force this kind of approach. But turning that into a model for how schemas "should" be designed seems to be turning an implementation short-cut into a virtue. The ideal to which one should aspire, surely, is to discover what the actual set of allowed real-world names is, and to make sure the IT system can handle them all, so that you don't have to turn users away or force them to enter incorrect names into the system. In an ideal world, the constraints in the schema would only reflect external constraints that exist in the real world, and not constraints imposed by the IT system. I used an Australian web site last week that wouldn't allow me in unless I entered a valid Australian postcode. I managed to invent one. Any system with validation rules that force users to enter false invented data is badly designed, in my book. And that includes forcing them to give an ASCII approximation of their name. Michael Kay Saxonica
Received on Sunday, 23 January 2011 00:35:43 UTC