The MPLS WG Archive

[Eric Rosen <erosen@cisco.com>]

Ina> Once the Host  FEC was defined, there is nothing to  preclude use of it
Ina> in many creative ways. 

Protocol elements have both syntax and semantics; reusing an existing syntax
for a new semantics is a change in the protocol.

The purpose  of the Host  FEC was the  following.  Some folks  insisted that
they needed the labels to distinguish between the following two cases: 

- a packet destined for some software module in a given router;
- a packet which is being tunneled to a given router for further forwarding.

If  you  don't need  to  make  this distinction  based  on  the label,  then
obviously there's no reason to ever use the Host Address FEC. 

Why did  people think they needed  this distinction?  Because  they had some
hardware which, after the initial  label lookup, needed to decide whether to
send the packet into the CPU-accessible memory, or whether to run it through
the  forwarding logic.  The  forwarding logic  in question  could look  up a
second label  or an IP address,  but could then  only send the packet  to an
output card, not to the CPU. 

AS it turned out, the need for this distinction never showed up in practice,
and  no one actually  originates Host  Address FECs.   To me,  this suggests
pretty strongly  that it needs to be  removed as part of  the progression to
draft standard. 

> If for example  one needed to carry some kind of  host information in LDP,
> would one define a new TLV, or just reuse the host address FEC? 

It depends on  whether one is trying  to avoid the IETF process  or not.  If
you are  trying to carry  some new kind  of host information,  shouldn't you
present an  internet-draft explaining that,  and requesting a  codepoint for
that information? 

If we keep the Host FEC in as  it is described in RFC 3036, it has an impact
on  all LDP implementations,  whether or  not those  implementations provide
features  which  have been  specified  in  some  other forum.   Since  those
features,  and the  related  use of  the  Host Address  FEC,  have not  been
presented to or approved by this WG, I don't understand how this can be kept
as part of the IETF standard. 

_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls