Do manufacturers recycle MAC addresses?

hex_omega

Just a silly question that popped up in my mind.

Say for instance, Intel keeps a database of the MACs they assign to all NICS they make. And if they have entries in there from 15 or 20 years ago, maybe they assume that these devices have been decommissioned or junked/destroyed and then re-use those addresses for new NICS? And because MACs are of local network significance only, if there are duplicates out there it's probably no big deal.

Also, why use 6 hex digits for the OUI. Isn't 3 more than enough? That's over 4k different combinations.

Devilsbane

Well a MAC address is 48 bits. That means that there are 281.5 trillion possible MAC addresses. The first 24 bits usually pertain to the manufacturer. So that means that 16.7 million manufactuers could each produce 16.7 million devices. Even if Intel (or another large company) used up it's 16.7 million MACs, they could just get another 16.7 million I'm sure.

Now the goal has always been to have a unique MAC address, but really all that mattered was that you didn't duplicate on the same network. With the oncoming of IPV6 where the MAC becomes part of the IP, maybe that will become a larger problem. I don't know very much about version 6 yet though.

And that answers your question about the is 3 enough. Is 16.7 million enough for every computer, to have a unique address? Probably not. There are 300 million people in the US alone. Even if only 6% of the population had a computer (which is extremely low, especially since some have multiple) then you would run out of unique numbers.

Asif Dasl

This website says they reuse them in different parts of the world.

Manufacturers re-use MAC addresses and they ship cards with duplicate addresses to different parts of the United States or the World so that there is only a very small chance two computers with network cards with the same MAC address will end up on the same network.

Fugazi1000

In the days of Netware 3.11 the best ISA NIC around was the NE2000 (10Base-2). A company I was doing an install for bought a load of NE2000 'compatible' cards.

The cards were duly installed (in about 100 PCs), and the Netware server commissioned, IPX/SPX config completed for each machine (WSGEN) and then found only 1 machine would work at any one time. Power off that machine, and another would start working shortly afterwards! Power that off and a different machine would start. Very baffling.


Turns out the customers' cheap 'compatible' NE2000 clones ALL had the same MAC address burned in.

At the time, there was no way to change the MAC (in software) so binned the lot.

hex_omega

Devilsbane wrote: »

Well a MAC address is 48 bits. That means that there are 281.5 trillion possible MAC addresses. The first 24 bits usually pertain to the manufacturer. So that means that 16.7 million manufactuers could each produce 16.7 million devices. Even if Intel (or another large company) used up it's 16.7 million MACs, they could just get another 16.7 million I'm sure.

Now the goal has always been to have a unique MAC address, but really all that mattered was that you didn't duplicate on the same network. With the oncoming of IPV6 where the MAC becomes part of the IP, maybe that will become a larger problem. I don't know very much about version 6 yet though.

And that answers your question about the is 3 enough. Is 16.7 million enough for every computer, to have a unique address? Probably not. There are 300 million people in the US alone. Even if only 6% of the population had a computer (which is extremely low, especially since some have multiple) then you would run out of unique numbers.

WHen you have a public v6 address, does the MAC integrate with the v6 address as well? Or is it only for link-local addresses?

hex_omega

Fugazi1000 wrote: »

In the days of Netware 3.11 the best ISA NIC around was the NE2000 (10Base-2). A company I was doing an install for bought a load of NE2000 'compatible' cards.

The cards were duly installed (in about 100 PCs), and the Netware server commissioned, IPX/SPX config completed for each machine (WSGEN) and then found only 1 machine would work at any one time. Power off that machine, and another would start working shortly afterwards! Power that off and a different machine would start. Very baffling.


Turns out the customers' cheap 'compatible' NE2000 clones ALL had the same MAC address burned in.

At the time, there was no way to change the MAC (in software) so binned the lot.

Your story is somewhat humorous. Did these cards come from Uzbekistan??

Trifidw

hex_omega wrote: »

WHen you have a public v6 address, does the MAC integrate with the v6 address as well? Or is it only for link-local addresses?

Yes it does but it is still only relevant on a single network, the same as IPv4.

MentholMoose

Devilsbane wrote: »

Even if Intel (or another large company) used up it's 16.7 million MACs, they could just get another 16.7 million I'm sure.

Yes, they can and do. See the list of MAC address assignments from IEEE... there are currently 15,120 assignments by my count:
http://standards.ieee.org/develop/regauth/oui/oui.txt

Devilsbane wrote: »

Now the goal has always been to have a unique MAC address, but really all that mattered was that you didn't duplicate on the same network. With the oncoming of IPV6 where the MAC becomes part of the IP, maybe that will become a larger problem. I don't know very much about version 6 yet though.

It's no different. EUI-64 address autoconfiguration in IPv6 will use the MAC to generate an IP... the MAC goes in the local identifier (has to be extended from 48 bits to 64 bits), and the subnet prefix is obviously based on the network, so as long as there aren't duplicate MACs on a particular network, it's a non-issue. Even if there ARE duplicate MACs, part of EUI-64 autoconfiguration includes checking for duplicate IPs, so it really shouldn't be a problem in IPv6. I kind of wonder if EUI-64 will be widely used, anyway, due to security and other concerns.