Hi Gary,

Gary Mills wrote:
>> To tell the truth, I am not quite convinced we should
>> allow TI implicitly
>> label disk, I think it should be controlled by the
>> consumer instead.
>> I remember there was some discussion about this that
>> we might inadvertently
>> destroy user's data. CCing Sundar, since if I
>> remember correctly,
>> he was raising those concerns.
>
> You have to make sure that it works on new disks.
> On SPARC, a new disk will not have a Solaris label.

I agree with this statement, but the reverse implication seems to be
the issue - even if disk doesn't contain Solaris label, it still might
have data on it. And in this case we don't want to let TI wipe them out.

> On x86, a new disk will not have a Solaris fdisk partition.
> Those cases should be easy to recognize.

This is not the problem we are trying to address by this bug,
since the installer is already able to install on a new disk
on x86.

I think that the scope of this bug is to

* enhance Target Instantiation (TI) with mechanism for creating
  label on Sparc disk
* implement appropriate TI API in order to allow consumers control
  this behavior

Best regards,
Jan


Reply via email to