Especially if you want to permit mistakes/erroneous decisions?
This is not forbidden by the SRFI process to amend with PFN a SRFI, or supersede with another SRFI.
Yes, SRFIs can be rewritten but it's somewhat costly. The thing is that for many problems we have started writing SRFIs without a clear idea of the problem domain, which is the main reason we haven't kept to the 90-day deadline in the SRFI guidelines.
The solution is to do most design work using a lightweight process, and only submit the result as a SRFI once it is stable. But we don't have a standard lightweight process; everyone has been doing their own thing, often by themselves. https://github.com/pre-srfi goes a step toward more collaboration, and has been a success, but each draft is still siloed up in its own repo and there's a sense that the person who first wrote the draft owns it, with others commenting in issues.
I expect that we'd get things done faster with better morale if we went a step further to use a monorepo with collective ownership.
The problem I see with R7RS is there is no standard test suite outside Chibi, which requires an advanced implementation to make it run. The SRFI process has no dedicated Scheme implementation officially, Chibi will not be the best choice for the long term.
Agreed. https://github.com/srfi-explorations/srfi-test was and is an attempt to fix that.
