question on survey question

["Jon Paris" <Jon.Paris@xxxxxxxxxxxxxx>]

> My question to you all is this:
> Is this a viable question, and is it clear enough?

Having spent a _lot_ of time as a member of the team investigating the
options when the notion of *PRV support was first introduced, I have to say
that I don't believe it is a viable option.  There was an enormous amount of
investigation done and budget was not a huge factor.  At the very least the
request as stated would have to change.  I'll come back to that in a minute.

I suspect that you could guarantee two things if this were to happen.  First
the number of enhancements would shrink radically (the testing component,
which is already a major factor. would balloon).  Second the compiler would
become less reliable and very prone to accidentally introduced
incompatibilities (oopppsss sorry I mean features!).  Any code that that has
to add additional conditioning all over the place is going to become fragile
and harder and harder to change.

Two suggestions.  First of all make the request a bit more realistic by
limiting the backward compatibility to the releases supported by OS/400 *PRV
support.  Second, it seems to me to more practical to ask that the compiler
permit the object to be saved back to an earlier release, and in those cases
where this cannot be done without a change in run-time support that such
support be PTF'd back to the earlier release.  Even then there will be cases
where the compiler would have to say "sorry - no can do" (e.g. where the
underlying OS/400 support does not exist on an earlier release).

Frankly, at this time I'm a lot more concerned by the news that Rochester
have dropped plans to re-do the embedded SQL support so we'll be left with
the half-assed pre-compiler support that we have today instead of something
that works properly.

Jon Paris
Partner400