begin quoting Tracy R Reed as of Sat, Feb 09, 2008 at 10:16:35PM -0800: > [EMAIL PROTECTED] wrote: > >I don't know what that means. I'm implementing Scheme tidbits as > >SICP reveals them to me. This is a one-in-a-lifetime opportunity not > >soon to be repeated... ...to actually > >implement language X while you are learning language X. > > Implementing a lisp is easy and fun. Now perhaps we all understand why > there are so many lisps, why that may have hurt the adoption, why there > are no huge sets of libraries that work with any particular lisp, why > there is no documentation, etc. A lot of the people who created their > own lisp actually fleshed it out and promoted its use but not enough > people so far have gotten behind any one lisp and one standard. Not that > it matters for a lot of jobs but it discourages people coming from a > perl/python background. And here lies the conundrum: Perl/python were a > real pain to implement and took massive amounts of time and are harder > to understand conceptually resulting in one main implementation that > everyone uses and they have huge mindshare. So which way should we > really be going? Hard to implement practically guaranteeing a single > implementation yet conceptually muddles or simple to implement and send > a hit squad after anyone who proposes that their own toy implementation > be used by anyone else?
[One thought or concept per paragraph, please. I keep getting lost in the above...mess.] How about have one entity provide the definitive specification, acceptance tests, and core documentation, and allow anyone to implement the language? You could use the hit squad to keep folks who don't implement compliant systems from claiming that it is compliant (or from implying that the non-compliant version implements the language as defined). Implementations could then range from the simple (but presumably slow) to the mind-bogglingly hard (but presumably very fast), without having to lose the benefits of a directed "vision". -- Regression testing and acceptance testing Are good, just beware of specification creep. Stewart Stremler -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
