[EMAIL PROTECTED] wrote:

 > Which is why I suggested propogating the originator's timestamp in the log
 > message. This solves the problem without exceeding our scope. Whether you
 > want to include intermediate hop timestamps as well is a topic for discussion.
 >
 > Cross-machine time synchronization is NTP's job, not ours.

IMO, it's very important to include intermediate-hop (or at least final destination)
timestamps because it's very important to be able to serialize events (this took
place before/after this), even if the events were generated on different machines
and you can't rely on NTP for this.  Why?

    * NTP is not part of the logging system and not everyone will set it up
      correctly.  You can argue that they should, but why design a system that relies
      on it when you don't have to.
    * It's hard enough getting trustworthy logs out of a system that's under attack.
      The assumption that the NTP itself isn't under attack seems like a bad one
      (it's pretty easy to inject forged NTP packets on the wire, etc.).

--
Chris Calabrese
Internet Infrastructure and Security
Merck-Medco Managed Care, L.L.C.
[EMAIL PROTECTED]
.


  • RE: syslo... by way of "Chris M. Lonvick" <[EMAIL PROTECTED]>
    • RE: ... Rob Cermak
    • RE: ... by way of "Chris M. Lonvick" <[EMAIL PROTECTED]>
    • RE: ... by way of "Chris M. Lonvick" <[EMAIL PROTECTED]>
    • Chris Calabrese

Reply via email to