Re: ftp://ftp.cisco.com/fred/rreq-03.txt

[[e-mail]]

> I think that the transition issue is just so much smoke and mirrors.
> There is no reason that an old, non-conformant device can't continue
> to be managed with SNMPv1 or clear text TELNET sessions with reusable
> passwords.  There is no "simultanous" conversion that has to occur.

I agree with your last two sentences.  "Simultaneous" is the wrong
word.  I thought the real transition issue was of allowing manager x to
see into domain y, so to speak.  If all managers x are bilingual for the
transition period (the period during which non SNMPv2 routers exist in
significant quantity/places in the global Internet), or if the routers
must be bilingual while the managers transition, then there can be no
issue.  So that gives 4 options, of which I think one ought to be
chosen explicitly:

   1) bilingual routers during manager transition
   2) bilingual managers during router transition
   3) bilingual everything for some long period of time, then reconsider
   4) maybe some managers can't see some routers during transition

I thought #4 was deemed a bad thing but if people think it's OK, then
you're completely right.

> I suspect that the new RREQ RFC will achive "standard" status after
> the SNMPv2 RFC does, right?

Hadn't thought about that.  I can't figure out whether RFC 1602 mandates
6 month and 4 month delays for advancement of RREQ (in the event that it
only references full standards, that is).  Good point.

>                              This would be interesting information to
> get from the SNMP folks.

If they wanted to promulgate an applicability statement for transition
(a document that says whether the manager and agent transitions are
supposed to be disjoint or overlapping, etc.) that would be good too.
I have a feeling that we're starting to write that here, and this doesn't
seem like the right WG to do that(?) considering that router and host
and manager issues are intertwined.

>                           I think that not mandating some sort of
> secure management mechanism is doing a big disservice to folks that
> build networks.  Real, operational experience on the internet supports
> this requirement.

I support your goal - just questioning whether the experience supports
the SNMPv2 mechanism and spec yet, or soon enough.  If we're at that
point then SNMPv2 should go full standard and we have our answer.
If not, then I think we have to wait, much as we might wish otherwise...

/Bill