Re: [DNA] L2 link and L3 link

[Greg Daley <greg.daley@eng.monash.edu.au>]

Hi JinHyeock,

JinHyeock Choi wrote:
> Dear Bernard
> 
> Kindly find my in line comments
> 
> 
>>>Let me be more precise. A host will detect the distinct difference between AP1 and AP2.
>>>It will receive a different BSS ID and observe it has attached to a different link-layer
>>>medium (AP).
>>>
>>>Roughly saying, it will notice Current Link/ Current AP Down and New Link/ New AP Up
>>
>>The host can also detect a difference between Ethernet Switch 1 and
>>Ethernet Switch 2, since it will receive packets from a different switch
>>MAC address.  For example, the source address for 802.1X packets will be
>>different, as could the contents of the EAP-Request/Identity packet (e.g.
>>the nas-id field will be different).
>>
>>
>>>What I mean is that they are at the different BSS, which requires new association.
>>
>>The notion of "association" is not unique to 802.11.  This is required for
>>any layer 2 protocol that allows multiple networks to be resident on the
>>same media.  This is because, unlike classic IEEE 802 media,
>>receipt of packets is not sufficient to conclude that the sender wishes to
>>"join" the network.  An explicit association step is required.
>>
>>PPP over Ethernet supports the notion of association, as will IEEE
>>802.1af.
> 
> 
> Thanks for your elaborate and enlightening comments. It seems there is more of 
> link-layer technology than I know. I am glad to learn that, in Ethernet, there also 
> is difference between point of attachment/ L2 link and (L3) link. 
> 
> I presented Ethernet model to illuminate the issue. Because it rather clouds it, 
> we'd better keep away from the difference between Ethernet and 802.11. 

As an aside, It's interesting that Bernard mentioned PPPoE.

PPPoE is inherently an L2 mechanism but defined by IETF.

Do L3 DNA procedures apply here?

It may be that there's no L3 router advertised on the 'Ethernet'
link. So existing router discovery procedures don't necessarily
apply.

Should we rely on PADI messages to determine reachability,
and then provide (authoritative) link hints when IP6CP
has gone to open state?

(I Guess so).

>>>But we'd better make it clear that, in 802.11,
>>>1) link and AP are different.
>>>2) An 'AP change' should not be automatically termed as a 'link change'
>>
>>I agree that there is a difference between 'link' and 'point of
>>attachment'.  While the 'point of attachment' may change, this does not
>>imply that the 'link' changes.
> 
> 
> I guess roughly 'point of attachment' is what I meant by L2 link. But I wonder 
> whether it is a widely understood and used term. 

It has been accepted fairly readily for DNAv4.

I don't know of any problems experienced with its use.

> And it may not be convenient to use the expression like 'When a host attaches 
> to a new point of attachment,'  or 'When a host makes new connection to 
> a point of attachment,'. 
 >
> 
>>>We may define a new term corresponding to AP, for example L2 link :-).
>>
>>I'd suggest that we continue to use the term 'link' to refer to L3
>>links.  Otherwise, when VLANs are implemented, we will find ourselves
>>drowning in confusion.
> 
> 
> I agree. 
> 
> And what about Point of Attachment/ L2 link? 

I guess that a (L3) link  is composed of devices
mutually discoverable through the same set of
Attachment Points.

We don't need to use the term L2 Link. if
we can identify the wireless media,
ethernet segments and etc, as Attachment Points.

Greg