clewis@ecicrl.UUCP (Chris Lewis) (04/05/90)
In article <15308@bfmny0.UU.NET> tneff@bfmny0.UU.NET (Tom Neff) writes: >I understand Larry's reasons for not wanting to implement compile-time >do's. In my defense I wish to point out that despite all the clever >counterexamples of MAYBE indeterminate do's involving goto's and newly >created temps and so forth, this is NOT what happens in the majority of >cases, which was my point. A conservative approach to 'do tainting' >would be fine. Most programs would still pass it. The difficulty that I see is that even if you can tell that a do isn't indeterminate, it still doesn't necessarily work the same way - name scoping in perl is dynamic and you might have your library code work differently depending on the state of your program. Especially if you "do"'d something twice... >Also, the preprocessor flag is a necessary thing for certain situations, >but I don't really like using it everyday. It's certainly no optimization! Perhaps. But run-time compilation versus compile-time compilation of do's ain't going to make that much difference whatsover unless you've got undump working... -- Chris Lewis, Phone: (416)-294-9253 UUCP: {uunet!mnetor, utcsri!utzoo, uunet!attcan!lsuc}!ecicrl!clewis Moderator of the Ferret Mailing List (ferret-request@eci386) Psroff mailing list (psroff-request@eci386)