I had judged that there was little more to be said----or at any rate little more for me to say---about this issue, but Mark Zelden has changed my mind.
He wrote: <begin extract> I think the point was about how often they are done. Quarterly? Monthly? Weekly? Daily? I have never been at a production shop that had a need to do them more often than monthly. Maybe some hardware vendors do them daily on a given system, I don't know. <end extract> because this does not seem to me to be the point at all. To use a lawyer's word, we have stipulated/agreed that disaster can/will ensue without serialization. Mark's view is that, this conceded, the important question is not whether it will occur but how often. He and I disagree about this. My view is that if the unfortunate can occur it will occur and at a disagreeably inappropriate time. Moreover, the substantive question at issue is a curiously trivial one. There are z/OS macros available for implementing the serializations required here. Any sysprog competent to chase control-block pointers at all is presumably also competent to use these macros. Why then court the unfortunate by refusing to do so? Russian roulette is a diseased pastime, all but independently of whether the probability that a trial will have an unfortunate outcome is 1/8 or 1/4096. John Gilmore, Ashland, MA 01721 - USA ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
