[c-nsp] TTL not decrementing (MPLS, SXI, Sup720)

[Peter Rathlev]

On Mon, 2011-02-07 at 18:19 +0100, Gert Doering wrote:
> On Mon, Feb 07, 2011 at 03:16:30PM +0100, Peter Rathlev wrote:
> > But we're also getting tired of people not able to
> > interpret traceroutes, so we've thought about disabling it anyway.
> 
> Part of our next bid for upstream connectivity will be "working
> traceroute".
> 
> We're getting tired of people blaiming our upstream link for
> transatlantic delays (hidden by traceroute hiding in upstream
> network), packet loss in upstream network, problems with the next-hop
> ISP border routers, etc.
> 
> Yes, we have our share of "customer not able to read traceroute", but 
> the pain of "being a customer, and not being able to *do* traceroute"
> is much worse.

I would tend to agree if "the customer" is someone versed in networking.
But for us (an enterprise-ish installation) the users are everything but
experts in networking. We receive countless requests from e.g. medical
equipments vendors, telling us that somethings's "wrong" with the
network. We simply use too much time trying explaining how traceroute
works.

One might wish for selective TTL propagation, maybe using an access list
of some kind. But since it has to be symmetric (AFAICT) it might not be
easy at all. (I'm aware of the "local" exception when disable TTL
propagation but it's not easy to script traceroutes through a PE. And
scripted traceroutes is something we use heavily.)

One of our last-mile providers recently disabled TTL propagation, and
that has somewhat limited us when trouble shooting, so I do understand
you point.

-- 
Peter