filenames across OSes: portability checker

Nick Levinson · 02-02-2008, 02:55 PM

Linux is so flexible about what names are acceptable that admins have to restrict them in order to avoid problems. One area is with files that are moved between Windows and Linux platforms, especially when lots of files are being moved at once (for example, I back up my Win-oriented flash thumb to a Linux folder).

Some of us solve this by trying to use restrictive names in the first place, such as DOS 8.3- or POSIX-compliant names, but that's less convenient when I'm in Win and it doesn't work for old files.

Can any Linux program or command look at names and draw our attention to those that might violate some specified standard?

Or do we need a new program to do this?

Thanx.

--
Nick

unSpawn · 02-03-2008, 08:43 AM

Quote:

Originally Posted by Nick Levinson

(..) restrict them in order to avoid problems (..) less convenient when I'm in Win (..) violate some specified standard

While I agree convenience is nice and that it wouldn't be a bad exercise to have something iterate until you can write C:\Progra~1\Applic~1\File~199.DOC, the "standard" you describe does not affect GNU/Linux:

Quote:

The POSIX.1 standard allows implementations to put additional restrictions on file name syntax, concerning what characters are permitted in file names and on the length of file name and file name component strings. However, in the GNU system, you do not need to worry about these restrictions; any character except the null character is permitted in a file name string, and there are no limits on the length of file name strings.

(OK, not POSIX ISO but the Glibc manual, Portability of File Names).
IMHO compliance here is not a GNU/Linux thing but the Other Thing having made up its own incompatible rules.

Nick Levinson · 02-06-2008, 06:57 PM

Can any Linux program or command look at filenames and draw our attention to those that might violate some specified standard? I meant a standard we users/admins specify as an overlay on Linux, not a standard within Linux or GNU/Linux (http://www.gnu.org/software/libc/man...mono/libc.html, [§] 11.2, esp. 11.2.4, as accessed 2-5-08).

For example, Linux permits a hyphen in a filename, including in the leading position, but that can make a filename look like an argument to a command, and therefore, while Linux permits it, an admin might specify a ban on leading hyphens and might want a technical means of preventing filenames from getting leading hyphens and might want a way to catch files coming into a system that weren't restricted by other admins and might have a locally banned character.

In http://www.ibm.com/developerworks/li...ary/l-sp2.html (as accessed 12-8-07), search for "Linux" to find relevant filename discussions.

This is more than convenience. When I drafted this message, my flash was holding 2,127 files and folders. That's a lot of names to check if I have to check them manually one by one. If I eyeball an average of one per second, I'd lose over 35 minutes and what if I accidentally miss one? I already found one accented letter Win95a allows that Linux replaces in filenames, which would make finding the file or folder harder using Search for Files or equivalent.

I don't want Linux to have restrictions built in. I want each admin to be able to set their own or not set any, as they wish. Not setting any would amount to letting Linux do what it does now, which would be fine for many users. But, if constraints are helpful, posting a written memo to all users isn't enough for consistent adherence, and a technical means would be useful.

Windows has the baggage of Microsoft's business and programming practices, but Windows is out there, and many of us must let the 800-pound gorilla sit wherever it wants. I like Linux's power. I want a way that doesn't reduce its power but that allows compatibility of names with various operating systems and other kinds of rules.

Any ideas?

Thanx.

--
Nick

unSpawn · 02-07-2008, 05:28 AM

I thought LFN would chop things off to whatever it deemed presentable. Anyway. I'm not aware of that kind of "lint check". Maybe something like

Code:

 isValidName() {
 declare -r strRestr="[0-9a-zA-Z-_.]"
 [ -z "${1//$strRestr/}" ] || return
 [ "${1:0:1}" = "-" ] || return
 [ ${#1} -gt 11 ] || return
} 
isValidName "somefile\#n@me" || echo no

should give you the idea even though it's a crude, way too strict example only working on names. Expanding it to include your local "rules" shouldn't be that hard.

Nick Levinson · 02-10-2008, 01:27 PM

That's the basic idea, expandable. I guess we need an envelope, so to speak, so it can run as a background process alerting us when a name is wrong. I'm not a programmer at that level, and I guess there isn't a program to provide this kind of functionality yet, so maybe I'll propose this somewhere and hope someone wants to develop the concept further.

LFN can't be trusted to ensure compatibility. Possibly, the accented letter on my Win platform got auto-replaced or disappeared on my Linux platform because of an encoding or charset difference between Win and Linux, but I don't know. Besides, letting an OS rename a file by a protocol without user preview has a drawback: it can create unexpected names, as happens when you drop two LFN files into a Win directory:

verylongnameMonday.txt
verylongnameTuesday.txt

The order in which they're created in or added to a directory may result in either one becoming VERYLO~1.TXT with the other getting the ~2, leaving you in the dark as to which is which, which is why files called in Win by programs get DOS 8.3 names from the files' programmers regardless of LFN support. Better that a process alert us to problematic names; then you can rename to taste besides being assured of compatibility/compliance.

Thank you for giving the coding a start. I hope someone picks it up.

--
Nick