Tom Marchant wrote: <begin extract> I don't follow the logic of this. Can you provide a concrete example of how a customer is deprived of a new facility because my code doesn't use that facility? </end extract>
The customer is deprived of the benefits of its use in your code. Parametric, tailored OCO delivery is not in my experience so problematic as it is being represented as being. Even if---wrongly in my view---SMP/E is not used to install a product, some formalized installation procedure is presumably used; and such a procedure can make use of multiple libraries of executables from which selections are made parametrically. Or again, a single executable can be zapped differentially. The difference between the storage required for, say, an iterated used of TRT and a much simpler but still quasi-iterative use of TRTE is substantial enough so that an unconditional branch around unused bytes is preferable to a string of NOPs; but the reverse may sometimes be the case and installation-time code that makes this decision is easy to write. Or again, I have sometimes found it convenient to use the Binder to replace some of the RSECTs in an executable at installation time. Mark Boonie's point is well taken. There are many ways to solve problems of this sort if the right people address them. Some ISVs view such 'packaging' tasks as low-level, almost clerical ones; and this assumption is a formidable obstacle to the adoption of new approaches. John Gilmore, Ashland, MA 01721 - USA
