Sat Nov 21 14:26:50 CET 2009

to thread or not to thread

+ C-based parsers are simpler

- state machines don't need it
- can't use VM code
- gdb is more difficult
- not all targets support it
- requires special memory management

The main question is however: isn't it better to have an event loop
based mechanism in the interpreter?  Original PF at some point had to
do this too.

In this light, isn't it better to switch to non-buffered IO?  Direct