If I could get say 15 minutes or so at the beginning of tonight's meeting, I would like to get feedback on an idea, and perhaps solicit volunteers.
It looks like there might be a way to bootstrap a perl 6 implementation rapidly. Rapidly as in a few of weeks. The core idea is to transliterate the existing bootstrap effort from p6 into ruby. With the ruby written in a very p6-like style (well, kp6-like - kp6 being the current restricted dialect of p6 being used for bootstrap). From the ruby one can then mechanically rederive p6, and p6 ast. So compared to current state, you would get code which runs fast, and has the system components decoupled. Development would no longer be dependent on front-end, compiler, and backends, all working in synchrony. Development of the three could proceed in parallel, without the current incremental bootstrap hell of everything having to always work together. Having the ast separates the front-end from the compiler and backend. Running directly means the compiler and backends can be allowed to break, permitting more aggressive development, and that the current backend slowness isn't a constant burden. And having the derived p6 means there's little overhead or critical commit to the bootstrap - development can simply, and reversibly, shift from the ruby to the derived p6, incrementally, whenever it seems desirable. Most of the pieces exist to create a p6 implementation, working at least as well as pugs, and written in p6. We've largely been stuck in the bootstrap maze (eg, to step across the street, you first have to shave a yak in Cincinnati). The above approach seems to offer the hope of tunneling through the mess. Maybe. Feedback encouraged. Mitchell _______________________________________________ Boston-pm mailing list [email protected] http://mail.pm.org/mailman/listinfo/boston-pm
