Re: multiple plumbers for interface type
Re: multiple plumbers for interface type
- Subject: Re: multiple plumbers for interface type
- From: Michael Cashwell <email@hidden>
- Date: Wed, 7 Feb 2007 15:49:38 -0500
On Feb 7, 2007, at 2:53 PM, Josh Graessley wrote:
Another example would be ATM. The system does not ship with ATM
support. Someone could write ATM support when they create an ATM
driver. This driver would have to register a plumber. If two ATM
drivers from different vendors were installed, only one of them
would get to register the IP over ATM plumber. In theory, this
could work just fine. The plumber is responsible for attaching the
protocol to the interface and handling the packets. As long as both
drivers interpreted the attach protocol parameters the same and
passed packets in the same state to the plumber, it shouldn't
matter which plumber wins.
I'm not disagreeing here with your point, but regarding ATM, there
are several possibilities. One I read about is indeed an IP-level
protocol over ATM that would reasonably map as you describe.
But there is another ATM protocol called LANE (LAN Emulation) that
basically synthesizes a 802.3-style multicast LAN using ATM circuits
(which can only be point-to-point) and some infrastructure devices
that look like a cross between a one-to-many reflector and a DHCP
server.
I wrote a Mac OS X LANE driver for an ATM card and it looked to most
of the OS like an Ethernet interface and was thus well below the
level of the plumbers.
(Fun, let me tell you. ATM comes from telephony. I've never before
seen so many layers in one onion.)
-Mike
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden