On Sun, Apr 24, 2011 at 1:54 PM, Scott Long <scottl_at_samsco.org> wrote: > Indeed, there's nothing wrong with preserving access to the system details > for the use of administration, troubleshooting, and even mere geeky > knowledge. This isn't about taking power away from the superusers, it's > about making the system smart enough to handle common situations reliably. > I'm sure that there some among us who pine for the good old days of manually > configuring and linking a kernel, but it's hard to argue that an > auto-configured kernel isn't pretty darn convenient most of the time. What > I'm proposing is just the next step in that process. > For me, your proposal would make life more difficult as it is on Linux. I've had to do more deployment/autoconfig setups recently and FreeBSD's method of device naming makes it much easier for me to deal with. I like the fact I easily know what disk is attached to what controller and what NIC driver is in use. The NIC specific naming is more useful than disk controller, but both have their places for me. It makes tweaking/troubleshooting quicker in some situations. In fact, I would like even more of it it, eg /dev/usbda0. What a disk is called is already different(for me) than what is in fstab since some(maybe many?) people are already using some of the abstraction methods currently available. If a sys-admin makes an effort, a consistent fstab is pretty easily achieved but it's better done by pre-deployment planning rather than after. If one of the new installer proposals handled this automatically, even better. My point is that device names are still an important hint to functionality particularly for auto-deployment/configure settings where specific hardware isn't always known ahead of time. -- Adam Vande MoreReceived on Sun Apr 24 2011 - 18:12:56 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:13 UTC