W.r.t.
> Here are my 3 thoughs on this:
> 
> 1) So far, sysUpTime has been used by the existing TCP-MIB objects.
>    Introducing a new discontinuity indicator now raises the question
>    whether doing so breaks existing implementations.
> 
Good point!

> 2) Personally, I doubt that there will be many implementations where
>    TCP can experience discontinuities that do not also affect other
>    parts of the networking stack or the whole system.
> 
OK, that is what we need to know.

> 3) I do see the problem that sysUpTime might be more frequently reset
>    as we like due to some other MIBs which do not use explicit
>    discontinuity indicators and experience more frequent
>    resets. However, the root of the problem is then really in those
>    MIBs, not in the TCP-MIB. (Sure, by introducing a discontinuity
>    indicator for the TCP-MIB we can protect against other misbehaving
>    MIBs - but is this going to set a precedence which at the end
>    requires a CLR that better every MIB module has at least one,
>    perhaps multiple discontinuity indicators?)
> 
You can understand that those "other MIB modules" will also use argument 
1) to not have to change, rigth?

> In the case of the TCP-MIB, I believe 1) is the killing argument. To
> introduce a new discontinuity indicator changes the implicit semantics
> of existing objects and would require to introduce new objects, which
> I think is prohibitive.
> 
> Conclusion: Spell out that sysUpTime is used as a discontinuity
> indicator for the counters.
> 
I can live with that if that is the consensus.

Bert
> /js

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to