On 04/05/2012 08:10 PM, Bill Davidsen wrote:
Andrew Haley wrote:
On 04/04/2012 06:47 AM, Zoltan Szecsei wrote:
Well, perhaps someone should also think about impact of name-changing. Quite frankly, what's in a name, so why change it?
It's because in the old scheme when a machine had several ports the mapping between the physical hardware and the dev nodes was not fixed: the kernel would just assign names as it found interfaces. This caused a fair bit of inconvenience.
You may think this is fixed, but in truth it is *broken*! Instead of having a single predictable interface name you could put in a script which would work on all typical user machines which have a single NIC, or a hardware NIC called eth0 and the first wireless NIC called wlan0.
I take your point: that is a special case, albeit a very common one.
But your reply really does beg the question: what does "the first" even mean in the context where there is more than one of anything? All you can do is hope they won't change, or hard-wire some names. This will be better once people get used to it.
On top of that, the names are not in any way "fixed," it appears that adding another eSTAT controller renamed my NIC, possibly because it moved to another slot.
From my side, I have a "no longer supported" SW product that licenses itself against the MAC address of ETH0. Yep, ETH0 and not any other name.
Put a symlink in?
To what? If a program does "ifconfig eth0" where do I put a symlink to make that return p57p3 results?
Sorry, yes. Point taken: not a devnode.
Andrew.