Alexander Clemm wrote:

> I don't see anything that would prohibit sliding windows?  Then
> again, I don't see it mentioned particularly. Would probably not
> hurt to mention it explicitly.  Perhaps some text in section 4.2.5,
> in conjunction with discussion of First Message Number?  Similarly,
> in section 7.1, when describing the algorithm, can include something
> along the line that we need to maintain a cursor, indicating the
> highest message number that was processed so far, and if the
> subsequent Signature Block message has a smaller FMN, we "skip
> forward" accordingly.

The step "Skip all other Signature Blocks with the same First Message
Number" in the algorithm doesn't work well with overlap.

For example, a sliding-window implementation with rule "send signature
block after every 5 messages, and include at most 15 hashes in one
block" would, after rebooting, send signature blocks

FMN="1" CNT="5"
FMN="1" CNT="10"
FMN="1" CNT="15"
FMN="6" CNT="15"
FMN="11" CNT="15"

...where skipping the second and third signature block isn't the 
correct behavior.

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

Reply via email to