John W Kennedy wrote:
I used SORT (and DSORT before it) thousands of times, and they both worked just fine. My best guess, from the use of the word "properly", is that these were idiots who expected the sort algorithm to be stable (i.e., to retain the original input sequence of records collating equal), which no optimized external sort ever has been.
We normally didn’t need the sort of stable sort you describe, and when we did, we’d key the records uniquely ourselves. So it’s hard for me to see that as the reason. I don’t know the details of the particular problem, other than at least three different shops here I worked they didn’t use the IBM sort and the claim in each one was that it didn’t work, whatever that meant. Sorry I don’t have any more information than that.
Of course many of the people who made these decisions were idiots, because they could be and still get respect. You pay someone big bugs to some in DP, and because you pay him so much, you end up respecting what he says, even when it is garbage.
Yes, and it worked in other COBOLs, too, but it was not portable, because the original 1960 design swept under the carpet the question of how many decimal places to give to the intermediate results. Shops that were serious about writing portable COBOL generally forbade the use of COMPUTE altogether.
Shops forbidding this or that statement in COBOL were common. Fortunately the shops were I worked were quite lenient. Anything went except the ALTER statement, and in one shop the manager insisted that we use PERFORM THRU rather than PERFORM because of a bug that was apparently genuine, but had also apparently been fixed ages ago.
Jim Allan --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
