World of A1200/World_Of_A1200.iso/programs/disk/misc/pfs/plaindoc/readme.doc

Text File | 1995-02-27 | 17KB | 452 lines

            Professional File System (c) 1993/94 by Michiel Pelt




                     Version 9.5 10-10-1994



1. Distribution Conditions

  This program is copyright (c) 1994 by Michiel Pelt.  It may be
distributed freely under the condition that no profit is gained from
its distribution, nor from any other program distributed in the same
package.

  Permission is granted to distribute this package by Bulletin Board
systems or network sites, under the condition that no fee is charged
on downloading it.

  All files that are part of this package have to be distributed
together and none of them may be changed in any way other than
archiving or crunching.  The only exception to this rule are the
icons, including the tooltypes, which may be changed to accommodate
them to the rest of the disk it is distributed on.

  This program is distibuted as shareware.  If you like this product
you should register.



Disclaimer

  These programs are provided "as is" without warranty of any kind.
All risks involved using these programs are entirely yours.  The
author assumes no responsibility or liability whatsoever for any
damage or dataloss caused by using this package.  Although everything
has been done to make sure the program is fully functional, the author
cannot guarantee this for a 100%.



2. Introduction

  The standard AmigaDos filesystem (OFS) is not very fast.  Commodore
has tried to improve the performance of their filesystem a number of
times.  Workbench 2.0 added the fast file system (FFS), and now, with
Workbench 3.0, we have the fast file system with directory caching
(FFS-DC).  This makes directoryscan faster but writing to disk and
directory-operations like delete and rename are slower than ever.

  This made me decide to write a filesystem that does better.  I used
all the knowledge I gained from my computer science study to make it
as good as possible.  The professional file system (PFS) has the
following features:

High performance
- write: 3-5 times faster
- read: 50% faster
- directoryscan: 10-20 times FFS, 3 times FFS-DC
- delete/rename/protect etc: 10-20 times faster

Other features
- full AmigaDos compatibility
- upto 10% more data on a disk
- better faulttolerance
- parralel access almost without performance loss
- disks NEVER get invalidated

  Both double density and high density floppy disks are supported.
This version of the filesystem also supports small harddisk
partitions.  Partitions up to 31MB are supported.  A version for
larger partitions is under development and will be available to
registered users only.



System requirements

  PFS will work on any Amiga with kickstart 2.0 or higher.  It accepts
both double- and highdensity diskdrives and RDB (Ridgit Disk Block)
harddisks.  The amount of memory PFS uses depends on the contents of
the disks present.  It will vary between 30 and 60K per floppy-drive
and up to 300K for a harddisk partition.



3. Installation


Quick start

  If you want to try PFS first, just doubleclick on the "PF0" icon.
"PF0:" now refers to PFS disks in drive 0.  Now proceed to chapter 4
about formatting and using PFS disks.


Installation on Workbench 2.0

  First you have to boot from your standard systemdisk or harddisk.
Now execute the "install2.0" script by calling it from the shell or
doubleclicking its icon.  This script will copy the filesystem to the
l: directory of your systemdisk and add two entries to your
mountlist, PF0:  and PF1:.

To use the filesystem you have to mount it using the mount command:

    MOUNT pf0:
    MOUNT pf1:

"PF0:" and "PF1:" now refer to PFS disks in drive 0 and 1. If you want
to have PFS mounted on startup you should add these commands to your
startup-sequence.


Installation on Workbench 2.1/3.0

  First you have to boot from your standard systemdisk or harddisk.
Now execute the "install3.0" script by calling it from the shell or
doubleclicking its icon.  This script will copy the filesystem to the
l:  directory of your systemdisk and put two files in your
SYS:Storage/DOSDrivers directory called PF0 and PF1.  Doubleclicking
their icons will mount the filesystem on the corresponding drive.
"PF0:" and "PF1:" then refer to PFS disks in drive 0 and 1.  If you
want to have PFS automaticly mounted on startup you can move the PF0
and PF1 files to your Devs:DOSDrivers directory.


Turbo users

  If you are the proud owner of a 68020+ machine you can, after
installing the filesystem normally, replace the 68000 version in the
l:  directory with the 68020+ version (you can find this in the l
drawer). Make sure the name of the filesystem is the same as stated in
the mountlist.  You can do that by renaming the filesystem in
"ProfFileSystem" or by changing the mountlist.


Making and changing mountlistentries

  If you have more than two drives, or you don't like the default
devicenames, you can change the mountlistentries to suit your needs.
Note that the "Buffers" keyword has no meaning with this version of
the filesystem. Its merely added for future compatibility. Make sure
that you set the "Unit" right. Please refer to the DOS manual for more
details.


Installing PFS on a harddisk partition

  Now you can install PFS on a harddisk partition.  Make sure you
backup the partitions you're going to change!  The size of the
partition should not exceed 32MB.  You can make such a partition with
Commodere's HDToolbox, which is supplied with the Workbench software.
With the same tool we can install PFS on the partition.

  Select 'advanced options' in the in the partitioning window.  A
number of extra buttons appear among which 'add/update...' with which
you have to add PFS.  Set the dostype field to '0x50465300' , version
9.5 and select ok.  After you have done that you can install PFS on
the desired partition with the 'change...' button.  Use all the
default options:  automount on, block size 512 (other sizes are not
supported), 2 reservered blocks.  Now select ok and the partioning
menu will appear again.  Go back to the main screen and save the
changes.

  PFS should be installed on the partition now.  After formatting it's
available for use.  If you've done something wrong and the computer
doesn't startup properly you can prevent mounting with the boot menu.

Notes:
-  PFS partitions are not bootable, so don't select 'bootable' in
   HDToolBox.
-  Every PFS device (including floppies) will use the version of PFS you
   have installed in the RDB now.  So if there is an update, changing
   l:proffilesystem is not enough, you have to update the RDB version too.
   This can be done with HDToolBox too.
-  A version for larger partitions is being worked on.



4. Using PFS disks


Formatting disks

  After mounting you can format floppies in PFS format by using the
standard c:format command. To format a disk named "hello" type:

format drive pf0: name hello

If the disk has been formatted before, you can use the quick option,
even if the previous format was FFS or OFS (this works the other way
around too).

format drive pf0: name hello quick


Accessing disks

Shell use:

  PFS disks can be referenced to by diskname (e.g.  "hello:") or by
mountname (e.g.  pf0:  or pf1:).  Almost all cli-commands work on PFS
disks like on any other disk, only faster.

Workbench use:

  When a PFS disk is inserted in drive 0 two icons will appear on your
workbench:  one with label "DF0:PFS" and one with the diskname as
label.  The first comes from the standard filesystem, the second from
the PFS filesystem.  You can access the disk through the latter icon.
It will behave just like df0:, only faster.


Differences with FFS/OFS

  When you use the c:list command the 'blocks used' information will
be to high.  The list command assumes every file consumes a block of
free space to use as headerblock.  This filesystem uses blocks much
more efficient, however.  As a rule of thumb you can say that the real
number of blocks used is the given amount minus the number of files
listed.  You can check this with the "info" command, which does give
the correct amount.  Directory tools like Directory Opus have the same
problem with their spaceleft figure during a copy session.  You'll
notice that after copying is finished the figure will go up, back to
the correct figure.

  You'll probably understand now that there generally goes more on a
PFS disk than on a FFS disk. PFS uses a larger part of the blocks for
actual data storage.

  Programs asking the filesystem information about a pfs disk will be
fooled into believing the disk is a FFS disk.  I do this to prevent
those programs from saying it's not a dos disk without giving the
information requested to the user (e.g.  the "info" command).

  The "c:AddBuffers" command used on pfs devices won't have any
effect on the cache, nor on memoryusage. PFS uses a dynamic directory-
and allocationblock caching algorithm, without any cachelimits. This
won't take too much memory (max about 30K), because the number of
directoryblocks on a floppydisk is limited.



5. AmigaDos compatibility

  All Amiga-Dos features are implemented except hardlinks, filechange
notification and recordlocks which are not used much anyway.  All
these features may be implemented in a future version, though.  A PFS
disk currently cannot be made bootable, so a PFS disk cannot be
"installed".



6. Restrictions

PFS has the standard Amigados restrictions on file and disk naming:

- maximum disknamelength 32
- maximum filenamelength 108
- maximum commentlength 80
- almost unrestricted characteruse in filenames

These restrictions are forced by AmigaDos, they are not inherent to
the filesystem. The number of files in a directory or on a disk is
restricted by available diskspace only.



7. Faulttolerance

  If you reset or crash your machine during a write operation a
standard AmigaDos disk gets 'invalidated'.  When you insert that disk
later the disk will be revalidated automaticly.  After this the
directory is back in the state it was before the writeoperation.  This
actually is a neat feature, but it has two drawbacks:  revalidating
takes time and, if you were overwriting a file at the time of the
crash, you lose both the original and the new version, since the file
has been half overwritten.  The directory states falsely that nothing
happened, however.

  With PFS the directory is correct at all times.  This means that
whenever you remove the disk ¹), reset or crash the system, all the
objects (files and directories) listed in the directory will be there
as stated in that directory, including the filesize and date.  In
other words:  a PFS disk NEVER gets invalidated.  This does not mean
you cannot lose any data, however.

  This is how it works:  when you operate on the disk (creating files
and directories, deleting files etc) the filesystem keeps track of
all changes to the directory but doesn't write these changes to the
disk until the operation is completed.  It does write the data but
only to clean areas, that is, diskspace that wasn't in use the last
time the directory was updated.  If that's not possible anymore, or
you stop writing to the disk for half a second, the directory is
updated so that the area occupied by deleted files can be reused
without invalidating the disk.  So if the machine brakes down at a
nasty moment you will probably find the disk in the state it was
before you started your writing session.  This includes overwritten
files:  they'll be back in their old state, unless the free diskspace
didn't allow it (only in that case both versions could be gone ²).

  The directory and allocationtables themselves are also updated in
such a way that the disk cannot become invalidated.  So even a crash
during the update won't invalidate the disk.

  An extra advantage of this procedure is that files that are
overwritten by a larger version are not fragmented, as in most
filesystems.  The file will be fragmented only when there is no large
enough continuous free segment available.  A fragmented file is even
defragmented if possible.

  Disadvantages are that the freespace on the disk gets more
fragmented than strictly necessary and that the directory has to be
updated more often.  PFS does everything it can to limit
freespace-fragmentation as much as possible.

  A PFS disk normally cannot get invalidated but it can get corrupted.
Nothing can be done against bad tracks on a disk or external programs
(like virusses or diskeditors) that mess with the disk.  A read or
write error makes a disk invalid.  There currently is no utility
available to recover corrupted disks.


  ¹) You should not remove the disk during an operation on purpose.
If you do, you'll probably get a read/write error and a "You MUST
insert volume xxx" request.  After this a "Not a dos disk" system
request will appear.  Simply select retry and the operation will be
continued.

  ²) If you want to make sure you can't lose a file by overwriting
it you should keep in mind that the filesystem can only guarantee that
if the free diskspace is more than the size of the old version + the
size of the new version + 5K.  Then the directory will be updated
after the new version has been written and not between deleting the
old and writing the new version.


Virus damage and viruskillers

  Bootblock virusses destroy the bootblock of the disk.  For this
reason the filesystem doesn't store unrecoverable data in those
blocks.  It does use the bootblock to recognize a disk as being a PFS
disk, so if the bootblock is destroyed, the disk in question won't be
recognized anymore.  To repair such a disk you simple have to write a
PFS bootblock back to it.  You can do that by copying a bootblock of
an uninfected disk to the disk in question.  Most viruskillers have
functions to do that.  Remember not to change any blocks other than
the bootblocks (block 0 and 1).

  Most viruskillers have no problem with PFS disks.  You should check
if your viruskiller accepts uninfected pfsdisks as normal, however.



8. Tools included

You can find some tools that can come handy with PFS in the tools
directory.



9. Programmers info

The file 'programmersinfo' contains information about the
diskstructure of PFS. I encourage everyone who is interested to
implement tools for PFS. I am quite occupied updating PFS itself, I
don't have time left to spent implementing things like optimizers and
recover tools. I will fully support any programming attemps. You can
contact me anytime if there are any open questions or problems.



10. Bugs

If you find something that doesn't function properly please let me know.



11. Credits

Here I would like to thank everyone for their help and support.

Rougier Sylvian and Pierre Carrette for their numerous ideas and
support from the very start.

Holger Kruse for his in depth Amiga knowledge.

Pavel Troller for helping develop a multiuser version (not yet
finished).

Nicola Salmoria for his extensive betatesting and accurate
suggestions. And for allowing me to include his excellent tool MFS
in the PFS package.

Klaus Deppisch for helping bring his diskspare.device and PFS
together.

Andreas Baum for translating the docs to German.

And all registered users for their support.



12. Author

You can reach me at the following address:

Post: Michiel Pelt
      Waalstraat 328
      7523 RS Enschede
      the Netherlands

E-Mail: [e-mail]

Tel: [phone]



13. Registration

  If you like this product you should register. You will be the first
to get the next major update.

  If you wish to make utilities for PFS, like undelete, optimize or
repair, let me know if you need any help. I support all PFS related
projects.

  The registration fee is as follows:

    cash: f30 Dutch Guilders, US $20, DM 30, £15, FF 100 or the equivalent
          in another currency.
    eurocheque: f30 Dutch Guilders. No other currencies.
    money order: f50 Dutch Guilders. No other currencies.

  Please sent the completed registration form, which you can find in
this directory, to:

        Michiel Pelt
        Waalstraat 328
        7523 RS Enschede
        The Netherlands

  E-Mail address: [e-mail]

  All registered users will get the PFS version 10 as soon as its
ready. I expect this to be before the end of 1994. If there are any
minor updates, they will be published on Aminet.