Again, Mr. Dissen is riding his hobby horse: 'ivory tower' approaches are bad, and 'real-world' approaches are good; and he has constructed a straw man in order to do so.
I nowhere recommended that "the entire inventory [of PL/I programs] be recompiled at the same time". I did strongly deprecate the use of PL/I for MVS uinder z/OS 2.1. The IBM language that Timothy Siupples elected to quote included the text: "If you decide to mix old PL/I modules with Enterprise PL/I modules, there are limited circumstances in which this mix will work". These circumstances are in fact very limited and increasingly so. I have advised on several such [significant] conversions, 21 or more, over the years. Unsurprisingly, the difficulties associatied with mixing old and new have grown. It is true that they can be managed; they cannot be managed economically; and, worse, they are fragile: small, apparently trivial source-language changes in a single subroutine can make them unusable. Even this retrograde shop apparently realizes that it must eventually "go to Enterprise PL/I". It is now time for it to do so; it was indeed time for it to do so many years ago. The reflexive identification of what is "real-world" with slothful, reactionary, risk-averse, "if it ain't broke, don't fix it" practices is killing this platform. In this particular case it is almost certainly uneconomic too. Compiled and optimized PL/I code is different for different ARCH levels. That for, say, ARCH(10) is very much better adapted to z/Architecture than the default. John Gilmore, Ashland, MA 01721 - USA ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
