I can see how your system works well, but consider a site where we have
a number of types of disks and several instances of those types of disks.
It is extremely useful to have a database (disktab) of disk types with
various configurations which we can use.  Not only do we not have to
remember what we called the file containing the description of the disk,
but after the initial time it takes to calculate the relevant parameters
for a drive and to come up with some basic configurations, we can disklabel
a bunch of disks quickly.  I conceed that the method I use requires some
initial set up time with paper and calculator, but once the task is done, I
know I have an entry that I can use again and again with confidence.
	Contrast this with a bunch of Sun machines we have around here
running SunOS 4.1.x.  We don't bother to write format.dat entries for these
disks, and, thanks to a bug in format which prevents you from using the
alternate (backup) label if it doesn't match the primary label, we have
been burned by this policy on more than one occasion.
	In short, I like the disktab system because of the history mechanism
it provides by allowing me to accumulate disktab entries.  As I label more
disks, I generate what amounts to an expert system on drive geometries, and
useful partition information.  Great for restoring trashed disk labels and
lost partition information in a pinch.
-Brian
On Dec 13, 10:41am, Dave Burgess wrote:
} Subject: Re: Why have disktab entries for SCSI at all?
} > 
} > 
} >    How about a tool that would generate a prototype disktab entry from
} >    MODE SELECT data?
} > 
} > That *still* doesn't deal with the partition info, which is the
} > primary use of disktab to begin with.  You can't divine that
} > information.  (Hell; you don't even need a label if all you want to do
} > is access the raw disk.)
} > 
} > I, for one, have never used disktab under NetBSD.  But that's another
} > subject.
} > 
} > 
} 
} Not really.  The way that the disklabel information is used is precisely
} the point.
} 
} 
} For example, when I use a new disk, I almost always use the 'there isn't 
} a label here yet' label generated by disklabel, labelling the disk with 
} a whole disk partition.  Once I get the partition information on the 
} drive in the first place, I edit this (now existing) table, with
} 'disklabel -e' and set up my partitions.  I then save the partition 
} information to a floppy and the root directory of wd0 so that I have it 
} for later.  I have found that this system of three disklabels works much 
} more effectively for me than the older disktab entry method.
} 
} Since the disklabel of an unlabeled disk generates a disktab prototype
} with one partition (the whole drive) that can then be put onto the disk,
} it sounds as though the select() call that you are looking for is
} already ready to go.
>-- End of excerpt from Dave Burgess