Re: DAMN the reply button
email@example.com (H. Peter Anvin) wrote on 13.11.98 in <199811140011.QAA20157@cesium.transmeta.com>:
> Using the Ethernet address makes some amount of sense: although you
> can set the Ethernet address if you want, it will really screw up your
> network if you have the same address for more than one host, which is
> usually what the licenser people care about.
Well, in a certain product I'm involved with, that's not what we care
about, since there's usually very little reason to run more than one
server (usually, that doesn't even work, because the hardware (not from
us) interfaced to the server doesn't have more than one port to connect a
server to), and the server controls the number of clients.
We still use the MAC address (mostly because it's the only numbered
component we *know* will be present), but it's definitely forgeable. It's
just that our usual customer wouldn't know how to do it. (Plus, I'm not
sure they'd even want to, in significant numbers. They might want more
client licenses, but that would need breaking MD5, or else tracing our
server through a pretty long code path, and I think we can safely assume
that the number of customers willing to go that far is very small - the
product doesn't cost enough to warrant this.)
Anyway, it's strictly a DOS/Win based product right now.
And, more importantly, most OSes already give you a way to find those
hardware addresses anyway, no need to implement anything extra.
Oh, and note that if someone breaks our number-of-clients system, they
will have broken ours, but not that of anybody else - so only our
customers have an interest in it. If it's a kernel service lots of
applications rely on, there's a lot more incentive for a lot more people
to actually break it.