Paul Gilmartin's original post remade a point a,lready made by the HLASM Language Reference in the words:
The type attribute can change during an assembly. The lookahead search might assign one value, whereas the symbol table at the end of the assembly might display another. I could wish for the use of [much] more felicitous, even literate language; but the point it makes is unambiguous, and the behavior it illustrates is appropriate. "A foolish consistency is the hobgoblin of little minds". Data dependence, different behavior in different circumstances, is not at all akin to taking decisions by Bernoulli trial using a pseudo-random number generator. My own heaviest use of the type attribute is in such constructs as |&in setb (t'&kwparm ne 'O') --value supplied? | aif (&in).kwparm_in --if so, examine it |&abort setb 1 --no, set quit switch | mnote &el,.'&mnpfx.023i. No value of the kwparm= . . .' | . . . | ago .after_KeywordParameter |.kwparm_in anop | . . . |.after_kwparm anop | . . . and such constructs are of course innocuous. The notion that a language facility that can be abused should be amended to make it into a notionally idiot-proof one is just not a fruitful one. It is the enemy of expressive power. John Gilmore, Ashland, MA 01721 - USA
