On 2021-01-30, at 12:23:03, Charles Mills wrote: > > Not an MVS- or IBM-specific comment at all ... > > Hardware documentation tends IMHO to be rigorous. The PoOp -- just as an > example -- tells you exactly what will happen for many "stupid -- don't do > that" cases, such as branching to an odd address. For reserved flag bits in > an instruction it specifies whether they are ignored, cause a specification > exception, or result in unpredictable outcomes. > Or possibly model-dependent.
> Software documentation tends to be much more "here is how to use the > feature" > That largely belongs in a Users Guide with only basic examples in the Ref. > leaving unspecified "what happens if you do something stupid?" The > answer tends to be "don't do that" rather than "here is exactly what will > happen." > And the unpredictable or unspecified cases should result in warnings. (But what if syntactic correctness is an uncomputable problem?) -- gil
