Re: implicit - explicit label values

My understanding ..
At label imposition the ingress LSR will change the Layer2 frame headers PID
(Protocol Identifier) or Ethertype value to indicate that the packet is a
MPLS labeled packet.
The LSR which pops the last label off the stack must be able to identify if
the packet's network layer protocol is IPv4 or IPv6. However, the label
stack does not contain any field which explicitly identifies the network
layer protocol. This means that the identity of the network layer protocol
must be inferable from the value of the label which is popped from the
bottom of the stack, possibly along with the contents of the network layer
header itself.

So when the last-hop LSR advetises an Explicit Null label upstream it
instructs the previously PHP router whether it is a IPv4 or a IPv6
destination. For its own good.
With Implicit Null this is obviously not required.

Feel free to correct me if my understanding is wrong :)

-- 
<ruhann>
blog.ru.co.za
On Thu, Feb 11, 2010 at 3:47 PM, Roger Pfaeffli <rpf23543_at_gmail.com> wrote:
> Hi Ruhann,
>
> Thanks for that, I see. Also while reading related RFC's (3031, 3032) I can
> see that it is not directly specified to ipv4.
>
> But can someone confirm (or not) my interpretation why there is different
> label value for ipv4 and ipv6 explicit null value?
>
> I post again my assumption:
>
> My assumption is that since with the explicit null label, the egress LSR
> (PE) receives a labeled packet and if he already knows while "checking the
> label" whether it is an ipv4 or an ipv6 packet after popping the label, he
> already knows in which table he has to do the lookup, the operation goes
> much quicker instead of just popping the label and then checking if it is
> ipv4 or ipv6. A bit similar like the Ethertype in the Ethernet Frame.
>
> regards
>
> Roger #23543
>
> Von: Ruhann [mailto:groupstudy_at_ru.co.za]
> Gesendet: Dienstag, 9. Februar 2010 15:46
> An: Roger Pfaeffli
> Cc: groupstudy
> Betreff: Re: implicit - explicit label values
>
> The implict null specification is not specfic to ipv4.
>
> The corrected descriptions:
>
>
> -a label value of 0 stands for the ipv4 explicit null label
> -a label value of 2 stands for the ipv6 explicit null label
> -a label value of 3 stands for the implicit null label
>
>
> --
> <ruhann>
> CCIE (R&S)
> blog.ru.co.za
> On Tue, Feb 9, 2010 at 2:38 PM, Roger Pfaeffli <rpf23543_at_gmail.com> wrote:
> Hi group,
>
> I have a question regarding the implicit and explicit label values. We all
> know that
>
> -a label value of 0 stands for the ipv4 explicit null label
> -a label value of 2 stands for the ipv6 explicit null label
> -a label value of 3 stands for the ipv4 implicit null label
>
> O.k., fine until now.
>
> My question is why is there no label value for the ipv6 implicit null
> label?
> Or with other words, why is there a special label value for the ipv6
> explicit null label?
>
> My assumption is that since with the explicit null label, the egress LSR
> (PE) receives a labeled packet and if he already knows while "checking the
> label" whether it is an ipv4 or an ipv6 packet after popping the label, he
> already knows in which table he has to do the lookup, the operation goes
> much quicker instead of just popping the label and then checking if it is
> ipv4 or ipv6. A bit similar like the Ethertype in the Ethernet Frame.
>
> On the other way, with the implicit null label, due to the PHP at the
> egress
> LSR (PE) gets the packet unlabeled and can already do the decision between
> ipv4 / ipv6. So no need to create a special label value for ipv6 implicit
> null label.
>
>
> Can someone bring light into this????
>
>
> Roger #23543
>
> BTW: In the example above, I was thinking about mpls labeled ip traffic,
> not
> vpn traffic since this would require the additional vpn label.
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
>
>
>
>
>
>
>
>
>
>
-- 
<ruhann>
blog.ru.co.za
Blogs and organic groups at http://www.ccie.net