Hi Alex,

Thanks for the explanation - it did indeed clarify things, and seems to provide 
a simple way to fix the situation!

The word "originator" comes from RFC 5424, and the current version of 
syslog-sign seems to assume that originator both originates normal syslog 
messages *and* signs them (originates Signature/Certificate Block messages). 
But your explanation -- a single originator (of normal syslog messages) could 
even have multiple signers (with different APP-NAME,PROCID) that sign the 
*same* normal syslog messages (with different algorithms) -- would seem to 
clarify things.

However, this does require some changes to the draft, right? (introducing the 
term "signer", and replacing some instances of "originator" with "signer")

Best regards,
Pasi


From: ext Alexander Clemm (alex) [mailto:[email protected]]
Sent: 12 June, 2009 09:28
To: Eronen Pasi (Nokia-NRC/Helsinki)
Cc: [email protected]
Subject: Re: Syslog-sign-26

Hello Pasi,

I guess any confusion stems from the use of the word "originator".  Therefore, 
let me use the term "signer" for the purposes of this discussion.  A signer 
signs syslog-messages using a specific algorithm; it is an "originator" of 
syslog-sign messages.  A single host can host multiple signers, which then each 
use their own Signature Groups and algorithms.  The syslog-sign messages can be 
attributed to a specific signer using (HOSTNAME, APP-NAME, PROCID).  Section 7 
does say that you can separate syslog-sign messages according to signer, using 
this triple.  (It is the syslog-sign messages you are concerned about; you 
separate the syslog-sign messages by signers.  You can separate the "normal" 
messages by virtue of who signed them.)  So, in summary, the ability to be able 
to use different algorithms to sign messages is supported, but the 
corresponding syslog-sign messages need to use different 
(HOSTNAME,APP-NAME,PROCID) to be able to distinguish which is used where.

Now, the question is whether to equate "signer" with "originator".  If you 
equate them, then each signer would be considered its own originator of its own 
syslog messages.  However, you can also simply regard it from the perspective 
that the same originator can in effect incorporate multiple signers, if wanting 
to use multiple algorithms concurrently.   It doesn't really matter - just like 
with "normal" syslog messages without syslog sign you don't really distinguish 
if there are multiple originators on the same host or only one - the syslog 
message does not contain an "originator-ID" but (HOSTNAME/APP-NAME/PROCID. ) In 
the end, the effect is the same: you support the ability to sign messages using 
different algorithms from the same host.

Does this clarify?
--- Alex


Pasi Eronen wrote:
"Hmmm... the major challenge in -25 was that although Payload/Signature
Block identify the originator (HOSTNAME,APP-NAME,PROCID), normal
syslog messages do not. So it seems you cannot separate the stored
log files by originator, and process the parts one by one.

If I understand you right, you're saying Section 7 does *not*
in fact assume that you can separate the normal syslog messages
by originator?

BTW, version -26 is still silent about whether a single originator
can sign the same set of messages using different algorithms (VER),
and if it can, whether these are same Signature Groups (with same
message number space) or different. What's your proposal for
addressing this -- or do you think signing using multiple algorithm
doesn't have to be supported?"

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

Reply via email to