Once again, folks, we are looking for an approach to a syslog replacement
that is practical in embedded devices which can NOT support IPSEC tunneling
due to limited resources. Authentication and privacy wrappers on
individual elements within a log record during transmission are highly
desirable (IMO) because they provide security without requiring
authentication/encryption of the whole message, connection or host-host IP
tunnel. The final log file or GUI might retain these wrappers, or present
plaintext to an authorized user, or recode the data with different
authentication information, e.g. to defend the log files within a host even
when the host is compromised (a la Schneier and Kelsey). This is a clear
benefit of separating the problem of encoding for log data transport on the
network, from representation in a file or GUI for human use.
XML provides a good way to structure messages so that these
security-critical elements can be wrapped, but it also has many other
advantages for structuring message data, e.g. the data element array Chris
has proposed. This could be used to add structure to the ULM draft's log
data model as indicated in the list of tags -- that model is basically flat
now. It's worth pointing out that XML provides that flexibility, but not
necessary to elaborate.
[EMAIL PROTECTED] wrote:
> I'm in a quandary. What do you all think?
I think this is more detail than is needed now. I think all that's need is
a description in one or two pages of a hypothetical XML encoding of some
examples from the ULM draft to demonstrate that the coding method has
sufficient flexibility. Since this is a security BOF, I think it should
mention the potential for authentication and privacy wrappers on elements
within the message, and mention the xmldsig work, but again not elaborate.
Such wrappers would be needed in any encoding, but giving it the XML
context relates it to current IETF work, which is good.
Alex Brown <[EMAIL PROTECTED]> +1 508 323 2283
