Alexander Clemm (alex) schrieb:
That a valid APP-NAME and PROCID need to be included is a given.

Are NILVALUEs valid?

Currently, the statement is that originators SHOULD use the same
values for those field for every message to be consistent (e.g.
section 4.1 and 5.3.1).  Should this "SHOULD" be changed to "MUST"
and a statement be added that APP-NAME and PROCID are supposed to
uniquely identify a signer on HOSTNAME?

Yes; if we use these values to distingiush different originators on the same host then they MUST be consistent.

I would suggest:
   This specification
   does not mandate particular values for these fields; however, for
   consistency, originators MUST use the same values for APP-NAME,
   PROCID, and MSGID fields for every Certificate Block and Signature
   Block message that is sent for one Signature Group, whichever values
   are chosen. To allow multiple originators per host, these values
   MUST be unique for the duration of the Signature Group.

Intention here:
a) if there is only one originator per hostname possible (use case: on a printer) then NILVALUEs are ok, because (together with the HOSTNAME) they still identify one originator. b) I think some time duration is necessary and chosen to impose the least restrictions. So the values have to stay the same for one Signature Group. That makes sure they can be used as identifiers for that Signature Group but leaves enough room for implementations. It is then possible to
- restart the daemon, thus changing RSID and PROCID
- use another program, thus changing RSID, APP-NAME, and PROCID
- use different PROCIDs in parallel (use case: one process per Signature Group).

One more thing: Except for the APP-NAME alone basically all selections from (APP-NAME, PROCID, MSGID) could be used. APP-NAME and PROCID would identify one orginator just as well. So if it seems like there could be other uses for a MSGID then that does not have to be fixed.

--
Martin
_______________________________________________
Syslog mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/syslog

Reply via email to