First, thanks to all for their answers. I think using the lowest MAC address on the
system might be a good enough solution, as it's reasonable to require that the
customers use the "real" MAC address assigned to their network interface.
> Unfortunately, I don't know enough about exactly what he's
> trying to do.
> All I know is he wants a unique identifier on a per system basis.
Sorry if I wasn't clear enough in my description. I work on a product which is an
example of a distributed computing application. It's purpose is to provide reliable
transaction communications between clients (possibly thousands) and servers (usually
between 2 to 20). It has no 'mother ship' (I wish it did as it would make things much
easier) because that provides for a single point of failure.
Customers use it in a very spread out fashion. The servers will always be very fast
(32 processor machines for example), highly firewalled, machines. But the clients can
be spread throughout an entire country (as they are with some customers). That's why I
said that the nodes may be on separate LANS where the operators will have no
communication with each other.
It runs on Linux which is one reason why I work on it :)
One problem we are facing now is that it is not always possible for nodes to agree
that they are talking to the same node. For example:
_________
| nodeA |
---------
IF1 IF2
| |
| |
| |
_________ _________
| nodeB | | nodeC |
--------- ----------
IF1 represents network interface 1, and IF2 represents network interface 2. The
problem is that nodeB and nodeC both have different addresses for the same node,
nodeA. So they cannot agree they are talking to the same node using just the
addresses. And the picture gets more complicated when nodeB and nodeC also have
multiple network interfaces, and also talk to other nodes.
One major thing our software has to do on each machine is to be able to do is sort all
the nodes, AND agree on the sorted list. For example, nodeB and nodeC would both sort
all the nodes it can talk to, and they HAVE to agree on which is the 'smallest' or
'largest' node. The reason for the sorting is because of the quorum concept which is
something I even really don't understand so I won't try to explain it here :) But
quorum, and hence sorting, is a critical element in distributed, transactional
communications products.
Creating a one time GUID would be fine when installing the software. Security is not
much an issue at this time.
Anyway, having said all this, I think taking the lowest MAC address of the network
card is a feasible solution. Perhaps I can stick a timestamp, FQDN, and heck, even
random numbers in there to make the possibility for it being generated twice basically
nil.
Thanks!
Warren
*****************************************************************
To unsubscribe from this list, send mail to [EMAIL PROTECTED]
with the text 'unsubscribe gnhlug' in the message body.
*****************************************************************
