Date: Thu, 24 Apr 1997 19:47:27 -0400 (EDT)
From: Pierre Phaneuf <pp AT 55-174 DOT hy DOT cgocable DOT ca>
To: opendos-developer AT delorie DOT com
Subject: Re: Usage of directory entries
In-Reply-To: <199704241124.NAA18259@grendel.sylaba.poznan.pl>
Message-ID: <Pine.LNX.3.95.970424194431.9418C-100000@55-174.hy.cgocable.ca>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk

On Thu, 24 Apr 1997, Mark Habersack wrote:

> > Things like file system checking and defragmentation are things better
> > done while the file system is either unmounted or mounted read-only.

> Exactly - in such a situation the system is in a known state and thus a 
> possibility to damage anything vital is almost eliminated. But the idea with 
> error detection by the file system code is excellent!

Yes, something like a simple 16 or 32 bit CRC on each inode... Without
actually doing any surface checking or "sanity" tests (wrong allocations
or whatever can go wrong with inode-based fs), you can produce those and
check them quite quickly... Which would give a warning about the data
being corrupted and to check the media before any "real" trouble comes
up... :-)

Pierre Phaneuf