On Mon, 16 Jul 2018 01:21:06 +0000, Jesse 1 Robinson wrote: >My very first oh-dark-thirty support call as a sysprog was for a program I had >just updated. I had cluelessly assembled it using JCL that had RENT,REFR in >the link edit step. After testing, I moved it from the non-APF test library to >the APF production library. Big S0C4. > >All inadvertent. A set of lessons I have never forgotten. > If something like the function of REFRPROT had operated on non-APF libraries the problem would have been discovered during testing rather than production. I consider this uniformity a powerful argument for universal REFRPROT.
>-----Original Message----- >From: Behalf Of Peter Relson >Sent: Sunday, July 15, 2018 5:55 PM > ><snip> >There's also REFRPROT nowadays. But that should have never been needed as an >option; it should have been the universal behaior ab ovo. How much extra >would it have cost to load user programs as well as system programs into >write-protected storage? ></snip> > >Page protection did not exist at the time that this logic was introduced. >So "ab ovo" was not possible if used the way I interpret it. > How, then, was it possible to protect code loaded from APF authorized libraries but not from other libraries? >Changing the behavior when page protection was added would be viewed, >properly, as incompatible and thus unacceptable. > It is always proper to report errors, even those that had never been detected before. >And the extra storage utilization (page multiple on page boundary) might have >been felt to be unacceptable too (perhaps no longer). > Understood. I don't know how great an impact it would have been. Apparently it was deemed acceptable for APF libraries. >Users are creative and might have chosen to rely on the existing documented >behavior. It is also well known that the system does not "protect" against >reentrant programs writing into themselves, just makes it harder. And it is >also well known that a program could write into itself and still be reentrant >(although that is surely frowned upon these days). > Are you conflating RENT and REFR? I understand the long-standing exception for RENT, but was it ever documented that a program could write into itself and still be REFR? -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN