Hi - > From: <shivendra.kumar@xxxxxxxxx> > To: <randy_presuhn@xxxxxxxxxxxxxx>; <ietf@xxxxxxxx> > Sent: Saturday, May 15, 2010 9:02 AM > Subject: RE: Query on SNMP Error Fields ... > This is an strict expectation as per SNMP RFC specs , > that the Set/GEt/GetNext requests set the error status > and error-index field to 0. Citation, please? While it is very reasonable and good defensive programming to set those two fields to zero in a request, I'm not aware of any text in the relevant RFCs that would actually *require* it. As long as the values present in the request are within the bounds permitted by the ASN.1 grammar, and implementation has no legitimate reason to reject them, as far as I know. > depending on this, Most of the SNMP agent implementations > copy the request message buffer Not the ones with which I am most familiar. > and modify the error status and error index to non-zero > only if an errror occured. This would be contrary to the elements of procedure in, for example, RFC 3416 clauses 4.2.1 and 4.2.5. The elements of procedure explicitly identify the cases when the error status and error index are set to zero. > if you already set it to non-zero , the response may actually be > meant as a success but, will always show errored value. This is just bad programming, and is not justified by the relevant elements of procedure. > So, IMHO, thism is an unacceptable practice to set the error > index /status to non-zero vaue in request. It's unwise (albeit legal) for the sender of the request, since an over-zealous receiver might (incorrectly) reject the request. What is truly unacceptable is for the responder to ignore the elements of procedure and not fill in the values of error status and error index required in the response by the standard. Randy _______________________________________________ Ietf mailing list Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf