>>96
I'm honestly OK with that value just being a string or some error code or something.
R7RS doesn't define any exception hierarchy, just the API. You can raise anything, but you probably want to define some ``
exception?'' datatype.
Who actually puts a large amount of code in catch blocks?
Java does an extraordinarily bad job at dealing with exceptions, forcing you to write bogus try/catch statements. C++ exceptions are just bad.
Exceptions are meant to be used in exceptional (hence the name) cases (i.e. when a TCP connection fails) and to be handled in even more exceptional cases. (i.e. retrying to connect (restartable exception))
Threading some error value by hand is also unSchemey, leave C to C programmers.
I actually prefer coroutines to exceptions
They are two orthogonal concepts.
>>97
the point of Scheme is to be minimal, not practical. Leave things like this to implementers.
R7RS is trying hard to be minimal and useful: it's 67 pages long now and it's not going to get much bigger. It's the spiritual successor of R5RS, and I was/am strongly against R6RS, except when I wasn't. (syntax-case, exceptions, better records)