Re: bsdinstall-amd64-20110313 remarks

On 03/21/11 03:25, Michael Reifenberger wrote:
> Hi,
> yesterday I tested the images listed in the subject and have the 
> following remarks:
>
> - At least the memstick image contains an empty fstab

The memstick stuff is new and (mostly) untested, so I'll check that out.

> - Does the usage of a "dangerously dedikated disklabel" give any 
> advantage?

Not that I can think of -- I'm not sure about maximum disk sizes for 
pure BSD-label disks. It's a legitimate option, though, for people doing 
manual configuration.

> - The usage of an UFS-Label for root mounting should be more flexible

Yes. It is somewhat difficult however, to cross-correlate gpart labels 
for GPT, APM, and PC98, with the labeled provider names (the label is 
not UFS labels, but gpart ones).

> - The first dialog step should set the keyboard layout

That *is* the first step.

> - The /etc is not writable which would greatly reduce the usefulness 
> for the ISO
>   image (no modified resolv.conf, sshd_config, ...)

This is only partly true. /etc/resolv.conf is a symlink into /tmp, which 
allows DHCP and network configuration to work.

> The usage of a nanobsd based base-installation would give a sufficient 
> advanced Live-OS installation.
>
> You could take a look into src/tools/tools/nanobsd/rescue where I 
> tried to address most of the issues above primary for rescuing GPT/ZFS 
> installations (with still hardcoded keyboard though).
>
> With two nanobsd slices on one memstick you can actually produce 
> combined i386/and64 Live-OS memsticks...
> I get both on a 2GiB memstick (Without packages).
>
> What do you think?

That's interesting, and I'll take a look around what you've done there 
when I get some spare time.
-Nathan