Your ISP is buying Bandwith from a carrier. This is an ATM circuit
connecting back to the carrier. This is a layer 2 conection (OSI model).
It is connected to his router for the layer 3 connection. You are
connecting to his DSLAM (equipement that provides you your DSL connection)
this is a layer 2 connection (most of the time, this is a PPoE connection)
you are then routed back to the router. So the broadcast domain is from
the router or switch of the ISP and this is why subscriber A and B can
talk to each other.
/32 subnetting is mostly use for point to point connection between 2 sites
using private line or frame relay.
Hope this helps
>> You are correct with your answer. DSL is a shared technology with no
>> garantied bandwidth. It works just like a modem. You connect to an ISP,
>> this ISP brings your traffic back (called backhaul) to its main site and
>> then gets transferred to the Internet. The Backhaul portion is (most of
>> the time) ATM UBR traffic. UBR means unspecified bit rate. This kind of
>> traffic is a lot cheapper for ISP.
> That does Hugo... our carrier tells us that the ADSL does have a minimum
> channel, which is MUCH smaller than the purchased speed of course... a
> has a committed availability of only about 512kb!
> But you are the first person who seemed to know about this stuff, so I'll
> ask a question that I was searching for before incase you have insiration.
> My understanding is the ATM network provisions point to point links form
> subscriber A to the ISP, and for subscriber B to the ISP. Subscriber A and
> Subscriber B are on the same subnet. This means they can't communicate. A
> expects B to be within it's broadcast domain, and vice versa, but ATM
> doesn't work that way - there is no effective broadcast domain - all
> must be routed through ISP. So how do A and B communicate?
> I've heard of /32 subnetting, but never seen it implmented in FreeBSD. I'm
> going to start managing my own ADSL, which will give me the option of
> running private addresses for A and B, which could mean that I can make
> be on separate subnets, but this isn't ideal - it forces me to do 1:1
> natting for all those clients at the ISP, which will complicate things I
> Anyways... a little off topic, please forgive me, but I'm trying to do it
> with monowall in the long run (possibly regular FreeBSD at the ISP though)