Agreed, it seems what this article (along with most of the comments) is dancing around is the idea of a stronger type system. String != string, e.g. UserText vs

Agreed, it seems what this article (along with most of the comments) is dancing around is the idea of a stronger type system. String != string, e.g. UserText vs SqlStatement. Using explicit conversion methods between those types helps clarify the actual “boundaries” of your system, or rather the independent parts and their individual boundaries. Joel Spolsky’s article illustrates the problem well. The problem stems from our simplistic type systems, which reflect a value’s storage class (“array of bytes”) rather than its semantic type (“raw input from user”, “sql safe TEXT literal”). Once your type system can differentiate between those two, it can help you identify where conversion between the two (aka “escaping” or “encoding”) is necessary. Then the problem of “dangerous string” disappears, because there is no more string: if it’s a UserText, it can’t be concatenated with a SqlStatement without conversion. Just like an “int” for example, or an array. Anyway I’m just rehashing Spolsky’s article, poorly. Don’t let my inaccurate summary reflect negatively on his point :p