() [EMAIL PROTECTED] (Ludovic Courtès) () Tue, 10 Jun 2008 09:37:39 +0200 Instead, applications should be written against an encoding-oblivious API. This is what core R5RS constructs and SRFI-1[34] provide at the Scheme level.
Unfortunately, that position doesn't provide a clean path towards an encoding-knowledgable API. Better would be to design an encoding-knowledgable API now (or as soon as possible), with the encoding value `unencoded' (or `8-bit-raw' or what-have-you) as default. This way, things are explicit and the (future) changes to application code will be less invasive, or even unnecessary. In short, the current API is not actually "future proof" (yet). Requiring obliviousness in the API makes for ugly work long term. thi
