| William M. Klein 2007-08-29, 6:55 pm |
| "Rick Smith" <ricksmith@mfi.net> wrote in message
news:13db2kuoft2919a@corp.supernews.com...
>
> "Pete Dashwood" <dashwood@removethis.enternet.co.nz> wrote in message
> news:5jla0pF4srfU1@mid.individual.net...
> [snip]
<more snip>
>
> In other words, it is possible to accommodate indirect
> forward references; but doing so adds an unprecedented
> level of complexity to COBOL compiler development
> with, in my opinion, no significant benefit.
>
Now, when we get away from these references in the "text manipulation stage" -
which I think everyone agrees SHOULD be prohibited, we then get to the "more
resources required to figuour out what is required" issue, that Rick mentions
here. I think it is clear that the '02 Standard *does* place such a requirement
on the implementer - and as far as I can tell everyone in J4 and Karl also agree
this requirement is placed on the implementer. This *CAN* be done and a
conforming compiler must do it.
As someone who both helped develop the '02 Standard and eventually expressed my
opposition to its final approval, I can safely state that this standard places
MANY "very expensive" requirements on the implementer (many of which would
provide minimal pay-back to the eventual user). This is why the current
direction is to make some new '02 fetures OPTIONAL in the next revision (if
any). Consider the Exception Handling RESUME statement and STANDARD ARITHMETIC
(using an arbitrary intermediate data type that doesn't exist in the language)
as prime examples.
There is (to me) a major difference between (obvious?) defect in the
specification (like the COPY constant-name one) and "difficult but not
impossible to implement" requirements like using TYPE statements before the
corresponding TYPEDEF.
--
Bill Klein
wmklein <at> ix.netcom.com
|