1 Feb
2007
1 Feb
'07
11:30 p.m.
On 2 Feb 2007 at 1:29, Roger Merchberger wrote:
Mid-drive directories/FATs/GATs a "kludge?"
Ah, no... There are specific
(and very logical) reasons for that, and as long as you don't store files
larger than half of the available storage (else you'd have a guaranteed
fragmented file) what does it matter *where* you stick the directory info???
I've seen varioius interpretations of this scheme (even one or two
CP/M systems, believe it or not) and while it's an interesting twist,
I'm not sure of its benefit. If one allocates alternately from the
free tracks either side nearest the directory track it might have
some performance advantages, but I've seen a lot of implementations
that put the directory splat in mid-disk and then proceed to allocate
from the outer tracks as normal. It's pretty
obvious that something
other than performance was the thought there.
In fact, the outer cylinder used for root directory and allocation
tables does make sense from a certain standpoint--the bit packing
density is the lowest on the outer tracks, so they at least in theory
should be the most reliable.
But there's no reason to have directories or FATs or any such thing.
Displaywriter data diskettes, for example have nothing that one would
recognize as a regular "directory". There is a master header block
located on cylinder 37, but it merely contains a freespace bitmap and
a list of first blocks of other lists. It's necessary to walk
through the lists to find the file names--and the lists they point to
to actually access the data. I've never used a Displaywriter, but
operation when opening documents must have been on the slow side.
Cheers,
Chuck