Re: draft-irtf-nmrg-snmp-tcp-08.txt

>>>>> Wijnen, Bert (Bert) writes:

Bert> Mmm.. I think I see what bothered me.  In RFC1905 it says
Bert> "request-response transaction" which makes me think about a
Bert> transaction that sends a request and gets back a response.

Bert> In the snmp-tcp doc I see request/response transaction where I
Bert> thought about a request transaction or a response transaction. I
Bert> guess the / confes (confuses) me. If you mean the same as in
Bert> 1905, then I am fine with it. Are there others who would confuse
Bert> the slash (/) to mean "or" ??

The updated ID uses request-response instead of request/response
and so I consider the problem solved.

Bert> - sect 2.2 last para Does this have an impact on FRAMEWORK-MIB?
Bert> Or how does it relate to that?  Might be good to explain.
>>  I have no idea why you think the message size affects the
>> SNMP-FRAMEWORK-MIB.
>> 
Bert> Since we do specify snmpEngineMaxMessageSize in that MIB.  And
Bert> you use MUST language. Now the snmpEngineMaxMessageSize says:

Bert>        DESCRIPTION "The maximum length in octets of an SNMP
Bert> message which this SNMP engine can send or receive and process,
Bert> determined as the minimum of the maximum message size values
Bert> supported among all of the transports available to and supported
Bert> by the engine.  "

Bert> So if I have a UDP socket where I do not want fragmentation,
Bert> then maybe I want to set MaxMessageSize to 1472. But for TCP I
Bert> MUST set if it to 8191 if I read your text correctly, so the
Bert> 1472 is no longer valid according to the description clause.
Bert> Maybe I am just worrying too much here?

The intention is to say that every SNMP over TCP implementation must
be able to accept or generate messages of a size of 8192 bytes. This
does in my view not really affect snmpEngineMaxMessageSize since it is
defined "as the minimum of the maximum message size values supported
among all of the transports available". So in your example, the value
of snmpEngineMaxMessageSize would be 1472.

Still, an application which uses SNMP over TCP can rely on 8192 bytes
since this is what the SNMP over TCP transport mapping guarantees you.

Can you agree with this?

/js

-- 
Juergen Schoenwaelder    <http://www.informatik.uni-osnabrueck.de/schoenw/>