Abbie:
I see 3 relatively minor sub-issues:
1. When profiles, interop tests are written, the authors of those
probably need to acknowledge that there should be no requirement for an
implementer to be forced to handle the currently largest maximum number,
since none of us would.
2. What is a realistic number that *should* be supported to meet
conformance/compliance requirements or interop scenarios? I presume a
rollover test should be in the interop tests.
3. If a specific implementation handles less than the unsigned or
signed long, how to they declare such or do they simply forward a
rollover? I think you and I are agreeing that there needs to be a
declaration capability.
Duane
Abbie Barbir wrote:
>I really do not see the problem here.
>
>the size of long allows 9 quintillion, that is fine, agree with Chris,
>most of us will die before the number is exahusted in a session.
>
>However, an RM can choke on other factors, such as avilable buffers etc.
>as opposed to the size limits.
>
>At the end of the day, it should be able to state its limitations.
>
>Abbie
>
>>
>>
