Characters Forbidden in File Names

[Jim Keenan]

For the purpose of enabling a Perl module, currently in development, to
function properly on the most common operating systems, can anyone provide a
link to a definitive list of:

1. Which characters are forbidden in filenames on Windows, *nix and Mac OS?

2. Which of these operating systems have limitations on the length of
filenames?

I ask these questions because my module creates plain-text files named
according to content extracted from the subject lines of e-mail. Such
subject lines may contain either too many characters for a particular OS or
characters forbidden in filenames on that OS.

Thanks in advance.

Jim Keenan

[Kevin Michael Vail]

In article <N6YTb.13593$KV5.251@nwrdny01.gnilink.net>,
"Jim Keenan" <no_spam_for_jkeen@verizon.net> wrote:

> For the purpose of enabling a Perl module, currently in development, to
> function properly on the most common operating systems, can anyone provide a
> link to a definitive list of:
>
> 1. Which characters are forbidden in filenames on Windows, *nix and Mac OS?
>
> 2. Which of these operating systems have limitations on the length of
> filenames?

Unix allows filenames to contain any character except /, which is used
as a directory separator. However, some characters can be problematic
in different ways. I think filenames are limited to 255 characters.

I don't remember exactly which characters aren't allowed in Windows
filenames, but definitely / \ * ? and I think < >. I think the max
length is 255 there, too.

Classic MacOS allows filenames to have any character except : (which
functions roughly as / does in Unix) and limits filenames to 31
characters.

Mac OSX supports filenames with up to 255 Unicode characters, but no
slashes (it's Unix underneath).

> I ask these questions because my module creates plain-text files named
> according to content extracted from the subject lines of e-mail. Such
> subject lines may contain either too many characters for a particular OS or
> characters forbidden in filenames on that OS.

You would probably be better off restricting yourself to a "least common
denominator" of characters known to be valid in all three OSes,
something like letters, digits, certain punctuation such as periods,
commas, dashes, etc.
--
Found Poetry (_Science News_, 14-Jun-2003): oldest _homo sapiens_ find
+-----------------------------------------+ ocean eddies' far-flung effects;
| Kevin Michael Vail <kevin@vaildc.net> | superior threads spun
+-----------------------------------------+ the pox from prairie dogs.

[Philip Newton]

On Tue, 03 Feb 2004 23:27:19 -0500, Kevin Michael Vail
<kevin@vaildc.net> wrote:

> Unix allows filenames to contain any character except /, which is
> used as a directory separator. However, some characters can be
> problematic in different ways. I think filenames are limited to 255
> characters.

Some very old versions have a 14-character limit.

> You would probably be better off restricting yourself to a "least
> common denominator" of characters known to be valid in all three
> OSes,

Do you also want to include VMS and DOS? What about ISO9660

> something like letters, digits, certain punctuation such as periods,
> commas, dashes, etc.

--
Philip Newton <nospam.newton@gmx.li>
That really is my address; no need to remove anything to reply.
If you're not part of the solution, you're part of the precipitate.

[Philip Newton]

On Tue, 03 Feb 2004 23:27:19 -0500, Kevin Michael Vail
<kevin@vaildc.net> wrote:

> Unix allows filenames to contain any character except /, which is
> used as a directory separator. However, some characters can be
> problematic in different ways. I think filenames are limited to 255
> characters.

Some very old versions have a 14-character limit.

> You would probably be better off restricting yourself to a "least
> common denominator" of characters known to be valid in all three
> OSes,

Do you also want to include VMS and DOS? What about ISO9660 CD file
system?

I think the latter is the most restrictive; IIRC up to eight uppercase
letters or digits (maybe underscores), optionally followed by one dot
and up to three more letters or digits. I believe there are also
limitations on the depth of nested directories.

DOS itself is only slightly less restricted - also "8.3" but with more
allowed characters.

> something like letters, digits, certain punctuation such as periods,
> commas, dashes, etc.

For maximum compatibility, not more than one period.

Cheers,
Philip
--
Philip Newton <nospam.newton@gmx.li>
That really is my address; no need to remove anything to reply.
If you're not part of the solution, you're part of the precipitate.