But this is just a modem, in bridge mode. Not a router. So it doesn't have "routes" to manage. Any thoughts of routing IP/TCP don't come into the discussion at all.

The only thing the modem knows about are ethernet packets, which are strictly routed by MAC. So unless something on my LAN feeds it a packet with the MAC of the DSLAM at the far end, then no ethernet packets will travel across the phone wires. EDIT: or perhaps they may pass that far, then get dropped at the far end due to MAC mismatch?

Bitt brought up the one point I'm only 99% certain about: what happens with ethernet broadcasts, that are not MAC address specific?

The regular PPPoE negotiation from the router, through the modem, to the DSLAM, and back.. uses PADI/PADO MAC discovery broadcasts, which are specific to the PPPoE protocol.

I suppose the Great Unknown here is whether or not the modem ever passes any other broadcast packets through the phone wires, or if it identifies and passes on only the PADI/PADO ?

For the curious, the "modems" in this case are by TP-Link. One is a TD-8841, the other is a TD-8901G at a buddy's house. Each has a built-in 4-port LAN switch, built-in router (disabled), and ADSL2+ modem.

We're each using them with multi-line MLPPP, managed by a WRT54GS/L router with Tomato/MLPPP on it. Since I have two modems here, one connects to the router over the (totally secure) WAN port, the other uses a LAN port. Since the latter is already on the LAN, its three extra ports become available for LAN use too.

In my buddy's case, his modem is in the basement, but the router is in his second floor office, with a single ethernet link between them. He needs more ethernet ports in the basement, so we just use the spare ports on his modem. This avoids the need to open up walls for running additional ethernet cables.

(yes I know/use the trick of two ethernets per cat-5 cable, too).

Cheers