"Don't release this to the world" strikes me as a bit hyperbolic, because parsing is rarely the bottleneck. Considering editors, individual source files are typically (relatively) small, even in a large project. Considering compilers, parsing is already the smallest portion of the run-time anyway.
Maybe it's my bias, but in a lot of applications I touched parsing or unparsing (and/or the associated inefficient I/O) would often be the single largest consumer of CPU time.
