- To: Bernard Aboba <aboba@internaut.com>
- Subject: Link identification
- Date: Tue, 30 May 2006 10:09:33 -0400
- User-Agent: Mutt/1.5.11+cvs20060403
Hi Bernard,
In draft-iab-link-indications-04.txt, it says:
"
Link A communication facility or physical medium that can sustain data
communications between multiple network nodes, such as an Ethernet
(simple or bridged). A link is the layer immediately below IP. In
a layered network stack model, the Link layer (layer 2) is normally
below the Network (IP) layer (layer 3), and above the Physical
layer (layer 1). Each link is associated with a minimum of two
endpoints. Each link endpoint has a unique link-layer identifier.
"
The text "Each link is associated with a minimum of two endpoints."
makes me think that the definition is something different from the
link definition in RFC 2461:
- For Ethernet media, if there are N nodes connected to the same
Ethernet hub, then it will have N*(N-1)/2 links, because the hub
itself does not have a unique link-layer identifier.
- For 802.11 media, if there are N Non-AP STAs associated to the same
AP of 802.11, then it will have N links.
My personal opinion is that in the shared mediam, there should be only
one "link" in the two cases above in terms of RFC 2461. Or the draft
is trying to introduce a new IETF definition of link?
On the other hand, in 802.21 WG, there is a discussion on how to
identify an L2 link in a media-independent way. 802.21 definition of
link is somewhat similar to that is defined in
draft-iab-link-indications-04.txt:
"
Link: A communication facility or medium over which L2 network nodes
may communicate for exchange of L2 messages. Each link is associated
with a minimum of two endpoints. Each link endpoint has a unique
link-layer identifier.
"
And there is a proposal for link identifier in 802.21:
http://www.ieee802.org/21/doctree/2006-05_meeting_docs/21-06-0608-00-0000-Link-Identifier.ppt
In the proposal, Link ID consists of the following attributes:
- Link Type
- Link-Layer Address of mobile node
- Link-Layer Address of PoA (point of attachment) (optional)
Link Type is 4-octet RADIUS NAS-Port-Type value
Link-Layer Address of MN and PoA is Address type of Diameter protocol
(RFC 3588). Link-Layer Address of PoA is optional because PoA in some
link-layer technology may not have link-layer address (e.g., an
Ethernet hub).
The 802.21 WG (and myself) is not sure whether this proposal is good
or not.
Your opinion is appreciated.
Best regards,
Yoshihiro Ohba
- To: Yoshihiro Ohba <yohba@tari.toshiba.com>
- Subject: Re: Link identification
- From: Bernard Aboba <aboba@internaut.com>
- Date: Tue, 30 May 2006 11:03:45 -0700 (PDT)
- In-Reply-To: <20060530140933.GB25081@steelhead>
- References: <20060530140933.GB25081@steelhead>
> My personal opinion is that in the shared mediam, there should be only
> one "link" in the two cases above in terms of RFC 2461. Or the draft
> is trying to introduce a new IETF definition of link?
Good point. The definition was taken from the IEEE 802 definition of
"link" not the IETF one. I will remove that sentence in the next
revision. In the IETF view, all ports on a switch connected to the same
VLAN are part of the same "link". However, IEEE 802 typically does not
use the term that way.
> On the other hand, in 802.21 WG, there is a discussion on how to
> identify an L2 link in a media-independent way. 802.21 definition of
> link is somewhat similar to that is defined in
> draft-iab-link-indications-04.txt:
Yes. I believe the 802.21 WG definition is similar to that used elsewhere
in IEEE 802.
> And there is a proposal for link identifier in 802.21:
>
> http://www.ieee802.org/21/doctree/2006-05_meeting_docs/21-06-0608-00-0000-Link-Identifier.ppt
>
> In the proposal, Link ID consists of the following attributes:
>
> - Link Type
> - Link-Layer Address of mobile node
> - Link-Layer Address of PoA (point of attachment) (optional)
>
> Link Type is 4-octet RADIUS NAS-Port-Type value
>
> Link-Layer Address of MN and PoA is Address type of Diameter protocol
> (RFC 3588). Link-Layer Address of PoA is optional because PoA in some
> link-layer technology may not have link-layer address (e.g., an
> Ethernet hub).
>
> The 802.21 WG (and myself) is not sure whether this proposal is good
> or not.
>
> Your opinion is appreciated.
I guess it depends on what you're trying to do with the link ID. If
you're trying to identify a connection between the station and switch/AP,
then this definition makes sense. If you're trying to identify a
link in the IP sense (e.g. a connection to a network) then this definition
would not be useful.