Re: sketchy laugh test: device discovery BOF

["Bernard Aboba" <bernard_aboba@hotmail.com>]

If you read IEEE 802.1ab closely, it is not really a general "device 
discovery" protocol.  It is more along the lines of Cisco CDP -- a protocol 
useful for discovering L2 topology and learning some things about device 
configuration on the link.

One reason why I say that IEEE 802.1ab isn't really a general "device 
discovery" solution is that unlike the 802.11 discovery, there is only 
support for Beaconing, and no Probe Request/Response mechanism.  That means 
that 802.1ab is not really suitable for use on wireless links, or even wired 
links with multiple potential networks (such as an Ethernet replacement for 
PPPOE, say).  Among other things, not having  Probe Request/Response 
mechanism means that a host has to wait for the Beacon (every 200 ms with 
LLDP) which can be too time consuming.

In addition, the LLDP attribute list is somewhat limited in that it only 
supports 255 attributes and there is not very good support for extending it.

I'm told that IEEE 802 is contemplating creating a separate "device 
discovery" solution to handle some of these other scenarios.  Such a 
protocol, if implemented would be likely to be used by multiple groups (e.g. 
IEEE 802.16, 802.20, etc.) and therefore needs a fair amount of 
extensibility.

Glancing over the DDP design, it shares some of the limitations of LLDP in 
that it has no Request/Response mechanism. Running over IP is problematic in 
that regard because the querier may not have an IP address yet.  DDP is more 
general in that it supports ASN.1 encoding and so might be quite amenable to 
addition of functionality from different groups. However, most of the 
generality is needed in wider usage scenarios than is contemplated in LLDP 
or DDP, so I'm not clear that there are real benefits to this.

_________________________________________________________________
MSN 8 helps eliminate e-mail viruses. Get 2 months FREE*.  
http://join.msn.com/?page=features/virus