<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Fri, 13 Apr 2007 15:53:38 +0100 (BST)
From   : chris@... (Chris Johns)
Subject: Possibly available: SJ FDFS

On Fri, 13 Apr 2007, Mike wrote:

> Different here: Syquest is random access device, tape is a stream type
> device.

The MDFS used tape devices that were (slow) random access devices. I guess 
this made the backup code easier (sector copy everything) and allows the 
%TAPE stuff to do on-line single-file retrieval.

> If anyone is thinking of hacking the MDFS code, my request
> list is:

> 1.  Bigger disc partitions than 60MB

This would be nice, but it would I think be a re-write of the MDFS 
filesystem code, as it uses units of 1K, and there are around 64MB worth 
of these in a 16 bit number.

If you change the format of the filesystem, you will have problem reading 
$.Fs to boot the thing.

This was different on the HDFS, which had a seperate System and User area 
of the disc. You couldn't attach floppies to the server, so writing of the 
system area had to be done in "utility mode" (only i think it was calle d 
something else) with a program on a client station to get the code across. 
With that done, you could boot that code and use FAST to do anything else, 
similar to the MDFS.

> 2.  More than 8 partitions per disc

Different sized partitions would also be handy - with the current MDFS, 
they're all 35 or 60MB (except the last one wich might be less).

A re-write would be an interesting project, but it's way too much for me 
to even consider. Without docs, it would be a lot of reverse-engineering 
of the current code.

Supporting of larger discs (ie using >480MB) also places a straing on RAM, 
as the allocated block map is kept in RAM. There's not a huge amount of 
RAM in the MDFS.
-- 
  Chris Johns <chris@...             >
<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>