Did you undo all of your "temp fix with magic numbers everywhere" from >>52?
They're not related to posting and the parser. They only prevented abusive requests like http://textboard.org/prog/1/1-300,1-300,.... (I'm actually running two instances of SchemeBBS, one for POST one for GET: rl Cheapo multithreading and a fast way to make the board read-only.
>>73
It also worked fine for me in the REPL with calling irregex-match on all the tests but however it wasn't working with calls to irregex compilation in a running instance of SchemeBBS. Just look at all the failures in this thread. Anyway I believe our two regexes are equivalent. I'll commit that and will now take a closer look at the most interesting optimization you made, in the posts-range function. And I do keep in mind your previous rant about the filter-func. I can't remember the reason why I implemented it like that, but it does feel hacky.
What I really need now is a working local instance, I can't go on like that, live testing on the server. And you also need it, if you'd like to test along. I'll publish the necessary nginx.conf which does a lot of caching for improving speed.
Btw, I took some time to read about the compatibility of MIT/BSD and GPL licenses and found that it made more sense to include MIT licensed code in a GPL product than the other way round. After all there's patches I had to made in MIT/GNU Scheme, httpio.scm and most importantly server.scm which was totally buggy. I might redistribute the whole bundle at some point, including MIT Scheme's customized runtine (besides the only BSD -licensed dependency is irregex, SRFIs being public domain)
I had to gave up on many planned features at the time, ambitious things like hot code swapping, so there is no real need to run SchemeBBS in an interpreter with its verbose console output. It could use some logger and compiled the scheme files. Errors handling and a supervisor would be a nice improvement too.