Informationen/MiNT-Doc/

  FreeMiNT-Portal

System  Netzwerk   Tools   Programmierung   Distributionen    Informationen   Homepages

"unixmode.doc" from the official MiNT-Documentation of the latest kernal release.

For questions, suggestions, critism etc. please email me.


[PREV] timezone.doc [NEXT]

The UNIXMODE Extended Filename Standard


0. Summary

The UNIXMODE extended filename standard provides a uniform, user-controlled
way of mapping the user's name for files (which need not conform to the
GEMDOS restrictions on length and case) to the names by which GEMDOS
knows them. It is intended to be as compatible as possible, e.g. whenever
reasonable the GEMDOS name of a file should match the name that GEMDOS would
have assigned to the name by it's usual conversions. File names with
multiple extensions are permitted by translating the '.' character to a
user-specified replacement (e.g. '_'). Long file names, and file names
with mixed upper and lower case, are represented by entries in a special
file, ".dir". This file also contains entries for Unix-like symbolic links.
Programs use the ".dir" files in the appropriate directories to translate
file names to/from GEMDOS. All translations are controlled by the user
through the environment variable UNIXMODE.


1. Rationale

GEMDOS filenames are limited to an 8 character name plus a 3 character
extension. The names must be all upper case, and must not contain certain
characters (such as ' ' and '.'). Many other popular operating systems,
most notably Unix, have much more flexible naming conventions. Moving files
from computers running such operating systems to an Atari ST can sometimes
present problems.

Moreover, short filenames are sometimes not very descriptive at all. Having
longer filenames, with fewer restrictions, makes remembering what's in the
files *much* easier.


2. The Proposal

We're stuck with GEMDOS's rules, at least for the time being. However, there's
nothing preventing programs from storing extra information about filenames.
Thus, a program could know that when the user refers to "govt.letter.april1"
to actually use the file "GOVTLET.AP1".

The UNIXMODE extended filename standard provides a uniform method of
associating long filenames with shorter, GEMDOS compatible filenames. 
"UNIXMODE" is the name of an environment variable. This variable contains
information about how the user wishes file names to be mapped. File name
mapping can be turned off completely, or altered to reflect changes in
the file system. Thus, programs using the UNIXMODE standard will still work
correctly in future versions of GEMDOS (or GEMDOS emulators) which allow
longer or otherwise more flexible file names. Also, UNIXMODE standard programs
do not impose their will on the user; the user controls whether or not,
and how, extended filename mapping is performed.

Programs that use the UNIXMODE standard can allow file names up to 31
characters in length. These file names may be mixed upper and lower case,
with periods and even spaces in them (although the latter is not recommended).
The only characters *not* permitted in file names are the null, tab,
newline, and carriage return characters.

The UNIXMODE standard also provides for symbolic links. These are special
directory entries that "point to" other directory entries. They provide
an easy way to reference the same file under more than one name. For
example, if there are several versions of a document under revision
simultaneously, a symbolic link called, say, "current" could point to
the most current version. An editor that followed the UNIXMODE standard
could be invoked by, say, "edit current" and would correctly retrieve the
current version.

In fact, under the UNIXMODE standard, long or otherwise unusual file names
are implemented by a special kind of symbolic link, the "automatic"
symbolic links. These links (also called "aliases") are used in mapping
long file names to ones that GEMDOS understands. If the user has made
provisions for allowing automatic links (see below), then every time a
file is created whose name is not acceptable to GEMDOS, then the file's
name is modified to be acceptable, and a symbolic link with the user's
preferred name for the file is created.


3 Filename Translations

Programs that follow the UNIXMODE standard will translate file names in
certain standard ways. These translations may be over-ridden by various
options in the UNIXMODE environment variable, but in general it is best
*not* to override them.

When translating from user names to GEMDOS names:
(a) if a symbolic link is found with the same name as the user name,
    it is used.
(b) the character '/' may be converted to '\', depending on user options
(c) filenames are converted to upper case
(d) occurrences of the character '.' may be converted to a user-
    specified character, to prevent conflict with GEMDOS
    restrictions on extensions
(e) filenames are truncated to have an (at most) 8 character base and an
    (at most) 3 character extension.

When translating from GEMDOS names to user names (e.g. printing the name
of the current directory):

(a) '\' may be converted to '/'
(b) if an automatic symbolic link exists to the file, then that name
    is used instead of the file name
(b) filenames are converted to lower case
(c) occurrences of the user-specified replacement character for '.'
    are converted back to '.'.

If the user name is not already a symbolic link, and the conversion from
user->GEMDOS->ser again would change the name, then an automatic symbolic
link may be created for the file name. This allows the user to refer
to the file by its full name, and also allows the program to print the
name by which the user wishes to refer to the file, rather than the name by
which GEMDOS knows the file.

Examples:
If all of the above mappings are asked for by the user, and the user
specifies the character '_' as the replacement for '.', then

user name:		becomes:
foo.bar			FOO.BAR
foo.bar.c		FOO_BAR.C
longname1.extension	LONGNAME.EXT		*
ReadMe			README			*
a_file.doc		A_FILE.DOC		*

In the cases marked (*), the user name is not recoverable from the GEMDOS
name. In these cases, programs should (if the user requests it) create
automatic symbolic links from the user name to the GEMDOS name; such links
will allow programs to properly recover the user's desired name for the files
in directory listings, etc.

3.1 The UNIXMODE Environment Variable

The environment variable UNIXMODE contains the user's preferences for
file name mapping. Each character in the environment variable string
represents one of the options below. If the UNIXMODE variable is not
found in the environment, programs may assume whatever defaults are
appropriate. It is recommended that symbolic links *not* be active by
default. The empty string is a reasonable default, providing complete
GEMDOS compatibility. The GNU C library assumes a default of "/", for
compatibility with old code.

The characters in the UNIXMODE string have the following meanings:
(characters inside of angle brackets <> are supplied by the user; thus
"r<c>" means the letter "r" followed by any other character, which will
be referred to in the text as "<c>")

/	-- Allow the use of '/' as well as '\' to separate the directories
	   in a path name. If the "/" option is given, then the name
	   "foo/bar" refers to the file "BAR" in the directory "FOO".
	   Otherwise (the default) it refers to the file "FOO/BAR" in the
	   current directory.

.<c>	-- Translate multiple uses of the '.' character to the
	   character <c>. Normally, GEMDOS allows only one '.' in a
	   file name, and it introduces the extension. This option allows
	   extra extensions. All but one (or perhaps even all) '.' characters
	   are converted into <c>. At most one '.' will be left; this will be
	   chosen to minimize the number of characters lost to the 8+3 rule
	   (unless the "u" option is in force, see below).

	   Example: if "._" is found in the UNIXMODE string, then the
	   following translations will occur:
		foo.c.z    becomes  FOO.C_Z
	        foo.c.bak  becomes  FOO_C.BAK
		a.b.c.d    becomes  A_B.C_D
		foo.bar    remains  FOO.BAR
		.login	   becomes  _LOGIN  (rather than ".LOG")

	   If the "." option is present, programs should convert filenames
	   back to what the user expected, e.g. in the above, "A_B.C_D"
	   should be displayed to the user as "a.b.c.d".
	   Note that if a file name would contain the character <c> used
	   for mapping '.', and automatic symbolic links are active,
	   then an automatic link should be made for the name.
	   Example: if ".@LA" is in the UNIXMODE string, and the
	   user enters the name "foo@bar.c", then a link from
	   "foo@bar.c" to "FOO@BAR.C" should be made, so as to
	   distinguish "foo@bar.c" and "foo.bar.c".

	   The translation character <c> must not be an alphabetic
	   or numeric character (if it is, this option should be ignored).
	   It is recommended that users adopt either ',' (comma) or '_'
	   (underscore) for this character.

-	-- The minus character is guaranteed to have no meaning. Thus,
	   if the UNIXMODE string consists solely of "-", the effect
	   is the same as if it were empty. This may seem puzzling,
	   but is necessary because many programs are incapable of
	   correctly recognizing empty environment strings.

d	-- Provide a pseudo-device directory. Filenames that begin with
	   the characters "/dev/" are assumed to refer to devices.
	   The devices available may vary, but the following devices must
	   always be present:
		/dev/console	same as GEMDOS file CON:
		/dev/tty1	same as GEMDOS file AUX:
		/dev/lp		same as GEMDOS file PRN:

r<c>	-- Provide a default root drive. If the "r" option is present, then
	   the search for a file whose name begins with a directory separator
	   (e.g. '\') and which have no explicit drive letter present will
	   begin on drive <c>. If the "r" option is not present, then
	   the search will begin on the current drive (the default). Note
	   that the actual drive on which the file is found may differ,
	   if one of the components of the file name is a symbolic link.

	   Example: if "rC" is set in UNIXMODE, and the current directory
	   is "D:\FOO\BAR", then the filename "\bin\sh" refers to the file
	   "C:\BIN\SH", not the file "D:\BIN\SH" as it would normally.

L	-- Allow, and recognize, symbolic links. See below for a more
	   complete description of this option.

A	-- Automatically create links to files whose user names are not
	   recoverable from the GEMDOS names. Automatic links are much
	   "tighter" than normal symbolic links; for example, removing
	   an automatic link will also remove the associated file or
	   directory, whereas removing a normal link removes the file, only.
	   Also, automatic links should appear in directory listings in
	   place of the associated file; normal symbolic links appear in
	   addition to, but not in place of, the files they are linked to.

	   If an automatic link is created, and then the "A" option is turned
	   off by the user, the automatic link will behave like a normal
	   symbolic link until the "A" option is turned back on.

	   The "A" option is meaningful only if the "L" option is also given.

H	-- Hide the ".dir" file that symbolic links are contained in
	   from directory searches. The file may still be manipulated by,
	   e.g. "Fdelete", but should never be shown to the user in a
	   directory listing, file selector, or similar display.

The next two options are provided for compatibility with replacement
file systems running under GEMDOS, or similar extensions to GEMDOS. They
should not be used in the normal environment.

c	-- Pretend case is significant in file names. If this flag is
	   present, then UNIXMODE programs will not translate the case
	   of file names in any way, and will assume that case is significant
	   to the underlying operating system.
	   The "c" flag should *not* be used in present versions of GEMDOS,
	   since these are case insensitive. It is provided so that programs
	   will work correctly under GEMDOS extenders or future versions of
	   the OS.

u	-- Assume file name length is unlimited. Like the "c" option, this
	   should not normally be used under GEMDOS. If "u" is not present,
	   it is assumed that the operating system recognizes only the first
	   8 characters of the filename, and 3 characters of the extension
	   (if any). If "u" is present, then it is assumed that files may
	   be of any length, and may contain any number of '.' characters. If
	   the second assumption is false (e.g. a GEMDOS emulator running on
	   VAX/VMS) then the "." option must be used to translate extra
	   occurrences of periods.


4 Symbolic Links

Symbolic links are implemented by special GEMDOS files named ".dir". Such
a file should appear in any directory in which a symbolic link appears.
If the "L" flag has been given in the UNIXMODE environment variable, then
programs should look in any ".dir" files anywhere along the path of a
file being opened. This means that symbolic links to directories are
permitted, and function as expected.

Example:
Suppose a program attempts to open the file "c:\foo\ReadMe". If the following
symbolic links are present:

in directory "c:": "foo" is linked to "d:\sub\dir"
in directory "d:\sub\dir": "ReadMe" is linked to "..\Important"
in directory "d:\sub": "Important" is linked to "vital.txt"

and if the "L" option is active, then the name "D:\SUB\VITAL.TXT" is passed
to GEMDOS for the actual file open operation. The translation proceeds as
follows:

c:\foo\ReadMe -> d:\sub\dir\ReadMe -> d:\sub\dir\..\Important ->
	d:\sub\Important -> d:\sub\vital.txt (-> D:\SUB\VITAL.TXT).

[The last step (case translation) may actually be interspersed with the other
 steps; part of a name may be converted to upper case as soon as it is known
 that it is not a symbolic link].

4.1 Implementation of Symbolic Links

In every directory that contains a symbolic link, there is a file called
".dir". Each symbolic link occupies a single record in the file (records
are separated by a newline character). Each record consists of 2 or 3 fields,
separated by tab characters. The first field is the user's name for the
symbolic link, e.g. "ReadMe". The second field is the destination of the link,
e.g. "d:\sub\dir". This destination need not be a complete path, nor need it
be in GEMDOS format; i.e. the destination name will be converted just like all
other user-supplied names. The (optional) third field consists of flags for
the link. At present, the only character that has meaning in this field
is the letter 'A', which (if present) indicates that the link is an
automatic link. However, programs must preserve all characters in the flags
field when manipulating symbolic links, so that they will work correctly with
future extensions to the standard.

Carriage return characters are ignored everywhere in the symbolic link
files. Programs should neither require carriage returns after the line
feed ending a record, nor forbid them.


5 Extensions

There may be future extensions to the standard. Programs should ignore,
without reporting an error, any characters that they do not recognize in
the UNIXMODE environment string or in the flags field of symbolic links.

The GNU C compiler library supports the following additional option in
the UNIXMODE environment string:

b	-- Open all files in binary mode. With this option, new lines in text
	   files are indicated by a line feed alone, rather than the customary
	   carriage return/line feed combination.

[PREV] timezone.doc [NEXT]

Last Update: Thu Apr 27 22:12 MET 2000     by AltF4@FreeMiNT.de