ML> I don't really remember when extended partitions were added to
 ML> MS/PC-DOS, perhaps because I skipped the 80286 generation, moving
 ML> from a PC/XT to an 80386-powered PS/2 model 80.  But, IIRC, there
 ML> weren't any hard drives for desktop machines having more than 32 MB
 ML> capacity before the PS/2 came along.

I, in my turn, (perhaps because I have never used one) don't remember offhand
when the PS/2 came along, but I do happen to know that the whole idea of an
"extended partition" was new to MS-DOS version 3.3.  MS-DOS 3.2 and earlier
didn't know about "primary" partitions.  It just had "partitions", and there
could be a maximum of four of them.  If they were formatted as FAT, they could 
be a maximum of 32MeB in size, meaning that the largest possible "all FAT"
hard disc size was approximately 128MeB .

As I said before, the concept of having multiple visible primary partitions,
far from being something new that Microsoft has introduced in order to move
the goalposts again for other operating systems, is in fact an *old* one in
the PC world.  Whilst it's *certainly* not the way that one would design a
partition table if one were writing it from scratch, it is the way that the
partition table on PCs has always worked.

So there is a strong argument that this is a bug in OS2DASD.DMD that needs to
be fixed.

 ML> With today's large-capacity cheap hard drives, losing a track's
 ML> capacity for each extended partition is most certainly not an
 ML> important criterion.  With the old, expensive, low-capacity drives,
 ML> losing a track didn't lose much capacity!  That "wasted space"
 ML> argument is the same meaningless red herring as is the "cluster size"
 ML> argument when discussing the choice between FAT and HPFS formats!

That's not quite true.  With the fake geometries created by CHS translation
the size of a "cylinder" can become quite large.  As Mike Ruskai mentioned in
a recent post talking about his own disc drive, the space occupied by a single 
cylinder can be almost 8MeB .  (Indeed, once the disc size approaches 8GiB the 
size of a cylinder is, simply by definition, always going to be 1/1024th of
the entire size of the drive.)

I find it hard to believe that you, who only three messages back in this echo
was praising the small size of a few-hundred-odd byte program written by David 
Noon, and who always makes such a fuss about the couple of MeB consumed by the 
system structures on an HPFS volume, would discount as trivial a *whole* *8*
*MeB*.  (-:

 ¯ JdeBP ®

--- FleetStreet 1.22 NR
114/477
143/1