Deciding between two choices

[Natale, Robert C (Bob)]

Hi,

Concerning what Margaret Wasserman wrote on Thu, 19 Dec 2002 19:34:57 -0500:
mw> Which one of these factors was missing from the great SNMPv2 wars?

Possibly none...it was more a case of extraneous factors not on your list.

mw> We somehow ended up with two useful, timely, competent proposals,
mw> and neither side was willing to compromise to reach consensus
mw> on the other approach. For the bleachers, I couldn't even tell
mw> the two approaches apart...

The SNMPv2 process was not all bad...and, indeed, may illustrate a
positive course of action that WGs clearly in deadlock/impasse might
want to consider:

We had several competing proposals for adding security to SNMP...
roughly, SNMPv2-Party (SNMPv2p), SNMPv2-User (SNMPv2u), and SNMPv2*.
Agreement could not be reached on which of these to go with (after
what I recall as at least a couple of years of debate...my memory
might be exaggerating that a bit, however).

Nonetheless, in the end, the WG made a very sensible decision not
to waste all other work that had been done during that process to
improve SNMP re features other than security.  That result of that
decision was "community-based SNMPv2", aka "SNMPv2c".  SNMPv2c
achieved fairly wide deployment among management application
providers, less so among agent application providers.  It has now
been declared Historic with the advancement of SNMPv3 to Full Std
status.

The important thing to note is that every improvement that SNMPv2c
made over SNMPv1 -- GetBulk, Inform, Trap-PDU normalization, better
error codes, varbind exceptions, needed data types, etc. -- have
been carried forward as an *integral* part of SNMPv3 (which added
the security components).

All I am saying is that, despite the tremendous turmoil that the
SNMPv2 WG experienced, the process did not generate completely,
we did capture what we could of value, and we did carry that
forward in the SNMP evolution process in a consistent and,
ultimately, efficient way.  This provided a sense of continuity
to SNMP developers -- not to mention practical coding, testing,
and deployment experience -- through final completion of SNMPv3.

Not that it really matters, but I've seen many commercial exercises
perform much less rationally under analogous circumstances.

Cheers,

BobN