Re: WLAN (was Re: IPv6 Host Configuration of Recursive DNS Server)

[Mohacsi Janos <mohacsi@niif.hu>]

Hi Masataka,


On Tue, 8 Jun 2004, Masataka Ohta wrote:

> Dear all;
>
> The problem is rather generic than DNS configuration. But...
>
> I know ND is wrong. That is, I know it is wrong to have generic
> link protocols ignoring link specific properties and has been
> pondering on how such protocols suffer.

I think ND is not wrong. There is a clear need for better support to
discover the link parameters, but it is necessary for mostly to the
routing protocols.

>
> I just recently noticed that WLAN (802.11*) is not very good at
> supporting ND (nor DHCP nor any broadcast/multicast protocol).
>
> Because WLAN use CSMA/CA, it needs ACK for reliable transmission
> by confirming that each packet reaches its destination without
> collision.
>
> However, with broadcast/multicast packets, there is no ACK for
> an obvious reason.

The IPv4 ARP is not better either, as you found. I think the WLAN should
be improved some way to support multicast in some form.

>
> If WLAN is lightly loaded, it is not a serious problem.
>
> However, if WLAN is heavily loaded, ND (and DHCP and any broadcast/
> multicast protocol) works poorly. Note that MIPv6 works poorly
> too.
>
> Still, management beacon frames are transmitted frequently
> and expected to carry information in a long run.
>
> So, it is possible to piggyback broadcast/multicast part
> of address resolution and autoconfiguration (maybe and
> routing) in reserved fields of beacon frames.

I think autoconfiguration can be treated differently, than address
resolution. ND as stand for "Neighbour" Discovery is a very good
improvement to ARP and transport specific things...
Of course ND can get into trouble if no reply for ND requests (with
soliticited node multicast address)...

>
> Though  it is possible to use some data frame frequently
> transmitted like beacon frams, it make the already congested
> WLAN worse.
>
> Anyway, the fundamental mistake is to try not to have link
> specific ways to perform address resolution and autoconfiguration.
>
> DHCP simply needs link specific ways of DHCP discover.
>
> ND needs, IMHO, a lot more.

What do you think?

>
> Note that modifying WLAN protocol to use PIFS is not enough
> unless combined with frequent transmission.

I agree.

Janos Mohacsi
Network Engineer, Research Associate
NIIF/HUNGARNET, HUNGARY
Key 00F9AF98: 8645 1312 D249 471B DBAE  21A2 9F52 0D1F 00F9 AF98