Time: Thu May 15 23:53:35 1997
	by primenet.com (8.8.5/8.8.5) with SMTP id TAA17246
	for [address in tool bar]; Wed, 14 May 1997 19:53:57 -0700 (MST)
Delivered-To: liberty-and-justice-outgoing@majordomo.pobox.com
Date: Thu, 15 May 1997 23:51:49 -0700
To: liberty-and-justice@pobox.com
From: Paul Andrew Mitchell [address in tool bar]
Subject: L&J: SLF: FAT16 and Computer Monopolies
References: <274338.500.uupcb@factory.com>

>FAT32 always uses 4Kb clusters, 
>whereas FAT16 clusters vary with the size of your drive.
>For example:
>   if your drive is <= 256Mb, clusters are 4Kb
>                 256Mb-512Mb, clusters are 8Kb
>                 512Mb-1Gb  , clusters are 16Kb
>                 1Gb  -2Gb  , clusters are 32Kb


So, check my logic here, please:

On large (1 - 2 GB) drives, 
cluster is:

32KB under FAT16
 4KB under FAT32

Therefore,

FAT16: average waste is 0.5 x 32KB = 16KB x N
FAT32: average waste is 0.5 x  4KB =  2KB x N
                         delta waste 14KB

where,  N = total number of discrete filenames
            in a given partition

      0.5 = average space used in last cluster
            (1 last cluster per filename)

Thus, FAT16 wastes space 8 times faster than
FAT32, i.e. 16KB / 2KB = 8

Another way to compute this is:

   delta waste attributable to FAT16 =

        16KB - 2KB = 14KB per discrete filename

Therefore, if you have 1,024 discrete filenames,
you are wasting 16MB by storing these files under FAT16:

     16KB x 1KB = 16MB

Do you agree?

/s/ Paul Mitchell
http://www.supremelaw.com




At 09:44 PM 5/14/97 -0400, you wrote:
>I don't know exactly how this topic fits in to the charter
>of L&J (though, of course, its not the only such topic), but
>some clarification is needed.
>
>Gene Kwiecinski wrote:
>> 
>> >N varies in direct proportion to the physical size
>> >of the drive, since the size of FAT16 is fixed at
>> >65,536 entries (each arrow "--->" is an entry).
>> >These arrows are called "pointers" in computer science.
>> 
>> Not quite.  The FAT is simply a *table* showing what space on the drive
>> is free or in-use.  It's bitwise, so that one byte can mark up to 8
>> logical clusters on the disk.  All it is, is a glorified collection of
>> "Occupado" signs for each cluster in that partition.
>
>Not quite.  A FAT is *not* a bitmap.  If it were, a files could 
>not become fragmented (since there wouldn't be a way to "chain" 
>its clusters together), though your *disk* still could.  A FAT
>*is* a table.  FAT16 uses a table whose entries are 16 bits;
>FAT32 table entries are 32 bits.  In either, a table entry
>represents two things for each cluster:
>   1) the "occupied" sign mentioned above
>   2) a "pointer" (table index) of the next cluster in a chain
>
>BTW, FAT16 is *not* fixed at 65536 entries.  It has a maximum
>size of 65526. 10 (it's been awhile - I may be off by one) are
>reserved to indicate end-of-file or bad cluster.  The size of
>a FAT (16 or 32) is proportional to the size of the *logical*
>drive (since you may have more than one logical drive per 
>physical drive).
>
>And, even if you use FAT32, you are still wasting (on average)
>half of a cluster per file (and directory) on your drive.  The
>advantage of FAT32 is that clusters are smaller, especially
>on large logical drives.  FAT32 always uses 4Kb clusters, 
>whereas FAT16 clusters vary with the size of your drive.
>For example:
>   if your drive is <= 256Mb, clusters are 4Kb
>                 256Mb-512Mb, clusters are 8Kb
>                 512Mb-1Gb  , clusters are 16Kb
>                 1Gb  -2Gb  , clusters are 32Kb
>
>Windoze 95 will not allow clusters greater than 32Kb, though 
>NTv4 will allow up to 64Kb (why anyone would *want* that, 
>I don't know).
>
>I always keep my FAT16 (yes, I still use FAT16, until there
>are FAT32 drivers for DOS, NT, and Linux) partitions to 512Mb 
>or below, so that my clusters will be 8Kb.
>
>FAT32 does *not* require that your BIOS be upgraded, since the 
>BIOS doesn't know anything about file systems at all, just 
>physical drives with their attendant cylinders, heads, and 
>sectors.  Now, if you want physical drives bigger than about 
>540Mb, you need a recent BIOS, which can translate read/write 
>requests using <cylinder, head, sector> to some form of logical 
>block addressing.
>
>The only new apps required by FAT32 are apps that physically
>read/write a drive, such as defragging utilities, etc.  And,
>the only entries that "point" to a cluster (not a sector) are
>in directory entries (which have a few bytes to spare anyway)
>and the innards of the file system driver.
>
>BTW, I *love* the idea of FAT32.  I'm sick and tired of having
>9 partitions on my system, just to keep dead space down to an 
>acceptable level.  I just wish someone would come up with drivers 
>for other OS's, and I don't have the time.  Even though I too
>use System Commander (a *great* program) and Partition Magic
>(a pretty good program), the rule on my main machine is that
>all partitions must be accessible to all OS's (I *do* admit
>to using NTFS on my servers, but it doesn't matter across
>the network, and I don't have to like it, do I? :-).
>
>grier
>
>=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>Unsub info - send e-mail to majordomo@majordomo.pobox.com, with
>"unsubscribe liberty-and-justice" in the body (not the subject)
>Liberty-and-Justice list-owner is Mike Goldman <whig@pobox.com>
>
>

========================================================================
Paul Andrew, Mitchell, B.A., M.S.    : Counselor at Law, federal witness
email:       [address in tool bar]   : Eudora Pro 3.0.1 on Intel 586 CPU
web site:  http://www.supremelaw.com : library & law school registration
ship to: c/o 2509 N. Campbell, #1776 : this is free speech,  at its best
             Tucson, Arizona state   : state zone,  not the federal zone
             Postal Zone 85719/tdc   : USPS delays first class  w/o this
========================================================================

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Unsub info - send e-mail to majordomo@majordomo.pobox.com, with
"unsubscribe liberty-and-justice" in the body (not the subject)
Liberty-and-Justice list-owner is Mike Goldman <whig@pobox.com>

      


Return to Table of Contents for

Supreme Law School:   E-mail