The MPLS WG Archive

[Bora Akyol <akyol@pluris.com>]

seenu@samsung.co.kr wrote:

> Hi,
>
> I have the following doubts on mpls-ethernet.
>
> Draft-ietf-mpls-label-encaps-07.txt  says that the Label Stack (label encapsulation on LAN media) must appear AFTER the data link layer headers, but BEFORE any network layer headers.
>
> But the draft-srinivasan-mpls-lans-label-00.txt says that: (ARIS too)
>
>         The main drawback of using the above approach for label-encapsulation is that accessing the labels requires that the frames be processed by software, reducing the benefit offered by label switching. Alternatively, hardware technology specific to label switching may be developed. However, standardized devices incorporating this technology do not exist today and cannot be built until the MPLS standard becomes stable.
>
>         And this draft has proposed the use of destination MAC address field for label encapsulation.
>
> Now, my doubts are:
>
> ? If one follows the first case (label-encaps-07.txt), how the label swapping will be done (s/w or h/w)
> ? Is there any ethernet standard (h/w) that supports the above type of switching (label-encaps-07.txt).
> ? Which is the standard draft that is being followed.
> ? If any of u have implemented the above, pls help me.
>
> Thanks in advance
>
> regards
> seenu

I haven't read this draft, but I see no reason why having the shime header between IP and Ethernet headers would cause packets to be processed in SW.

Bora