[mdnsext] The Scope Swamp: Wireless Considerations

[Alf Watt <[e-mail]>]

Following up from the BoF discussion in Berlin today on scoping of service
advertisements.

The discussion of scope needs to acknowledge the difference between wired
and wireless networks.

In all IP networks the size of the broadcast domain is a fundamental
limitation, as the number of devices on the link increases the broadcast
traffic begins to consume a larger and larger percentage of the available
bandwidth. Traditionally large wired networks were segmented along
physical lines, by floor or wing, for example, with each area being
assigned a subnet and routed to provide end-to-end connectivity.

Wi-Fi users expect (and the 802.11 protocol requires) that users on a
single SSID are connected to the same IP network and broadcast domain so
that they can roam seamlessly anywhere within a particular ESS (which can
be a building, campus or metro area). Limiting the size of these broadcast
domains means separating users onto different Wi-Fi networks depending on
their role, instead of their location.

So for wired networks we have a natural, physical scope depending on the
physical connection of hosts, in wireless networks we have scopes which
depend largely on the users access rights (since Wi-Fi is almost always
the security demarkation in networks).

It¹s also worth nothing that the BCP for broadcast domain size is only 255
hosts (a /24) which makes the 802.11 ESS requirement for a single
broadcast domain behind a given ESS difficult to achieve for large
networks, which is why many Wi-Fi vendors support multicast filtering,
client isolation and ARP proxies all of which can impact mDNS usability.

Best,
Alf

FYA http://www.youtube.com/watch?v=-FizhXOW2VM&noredirect=1