The Routing Over Large Clouds Mailing List Archive by date

[jhalpern@us.Newbridge.com (Joel Halpern)]

M. Ohta asserts that the absence of TTLs in PNNI routing is a bug.
Actually, the PNNI developers looked at this question.  We considered a
TTL.  But we used a different mechanism, that is as strong.
The PNNI call carries a hierarchically complete source route.  The call can
only be forwarded according to that source route.  Thus, the only way to
get a perpetual loop in a call (setup or data) is to have an infinite
source route.  Since that is impossible, the mechanism is as robust as
TTL.

It should be understood that all error detection/recovery mechanisms have
limitations.  The general goal is that one misbehaving switch should not
be able to severly damage the network.  Wtih TTL, one mis-forwarding
router (even if it fails to decrement TTL) can not bury the network.
With PNNI, one switch which mis-understands the soruce route will get
caught (and the call terminated) at the next switch.

However, with CSR, if the call gets forwarded the wrong direction,
it is possible to establish a path which has a loop, and has no
decrement of TTL on it.  Significant care must be taken in the overall
system design to prevent this or to provide an alternate recovery mechanism.

Yours,
Joel M. Halpern				jhalpern@newbridge.com
Newbridge Networks Inc.