RE: draft of a core protocol spec

Albert,

Yes, it would be a big layer violation to allow L3 clients to optionally set the RPR TTL. L3 should view the ring (almost) no differently than it views a broadcast LAN.

In any case, there is no mechanism to allow a (L3) client to set the (RPR) TTL.

jl

-----Original Message-----
From: A.Herrera [mailto:[email protected]]
Sent: Saturday, November 09, 2002 7:47 PM
To: Vinay Bannai; Anoop Ghanwani; 'Frank Kastenholz'; John Lemon;
[email protected]
Subject: Re: [IPORPR] draft of a core protocol spec


Will it be a big layer violation to allow L3 clients
to optionally set RPR TTL. Given that we're dealing
with 10gig, near-WAN, 100+node ring topologies that
can easily carry megabit streams gone astray.

I agree L3 TTL checks network loops and RPR TTL
checks ring loops but a large ring is a large network
by itself :-)

Albert

----- Original Message -----
From: "Vinay Bannai" <[email protected]>
To: "A.Herrera" <[email protected]>; "Anoop Ghanwani" <[email protected]>;
"'Frank Kastenholz'" <[email protected]>; <[email protected]>;
<[email protected]>
Sent: Saturday, November 09, 2002 11:58 AM
Subject: RE: [IPORPR] draft of a core protocol spec


> Hi Albert,
>
> Your later observation is correct. The setting of the RPR TTL is directly a
> funtion of the local ring topology and it does not and should not have any
> bearing on the L3 TTL. It would correct to keep the TTL from percolating up
> and down the layers. The L3 TTL and the RPR TTL serve two entirely different
> purposes and should not have linkage. The L3 TTL avoids infinite looping in
> a network whereas, the RPR TTL serves the function of keeping the packet
> looping infinitely within a RPR ring.
>
> If one were to choose to use a MPLS label between the layer 2 and layer 3,
> then the issue of setting of the MPLS label becomes slightly interesting but
> that has nothing to do with RPR.
>
> Vinay
>
> -----Original Message-----
> From: [email protected] [mailto:[email protected]]On Behalf Of
> A.Herrera
> Sent: Saturday, November 09, 2002 8:31 AM
> To: Anoop Ghanwani; 'Frank Kastenholz'; [email protected];
> [email protected]
> Subject: Re: [IPORPR] draft of a core protocol spec
>
>
>
> ----- Original Message -----
> From: "Anoop Ghanwani" <[email protected]>
> To: "'Frank Kastenholz'" <[email protected]>; <[email protected]>;
> <[email protected]>
> Sent: Friday, November 08, 2002 11:55 AM
> Subject: RE: [IPORPR] draft of a core protocol spec
> >
> > There used to be a TTL many draft versions ago, so that's
> > probably the reason for your recollection :).  It was subsequently
> > removed because it was decided that anything above the MAC
> > shouldn't need to know about TTL.
> >
> > -Anoop
>
> Just a quick question for the WG. It could be
> outside the scope of the core spec or it might be
> a non-issue altogether.
>
> How is L3 TTL reconciled with RPR TTL. Given that
> we no longer deal with LANs but rather large
> MANs.
>
> Say, a packet from a traditional mesh hits,
> traverses, then exits 3 large metro rings (of
> over 100 nodes each), does the L3 TTL decrement by
> 3. And should this packet somehow find an entry back
> to the first ring, shouldn't there be a mechanism for
> an earlier discard. Or do we treat these rings the
> way PWE3 treats their tunnels where the TTLs don't
> percolate up nor percolate down.
>
> Albert
>
>
> _______________________________________________
> IPORPR mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/iporpr
>
> _______________________________________________
> IPORPR mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/iporpr