Why do tar and gzip files usually have a file extension

If the file name is file-1.0.tar.bz2, the extension is bz2. The method you're using to extract the extension (fileext=${filename##*.}) is perfectly valid¹.

How do you decide that you want the extension to be tar.bz2 and not bz2 or 0.tar.bz2? You need to answer this question first. Then you can figure out what shell command matches your specification.

• One possible specification is that extensions must begin with a letter. This heuristic fails for a few common extensions like 7z, which might be best treated as a special case. Here's a bash/ksh/zsh implementation:

  basename=$filename; fileext=
  while [[ $basename = ?*.* &&
           ( ${basename##*.} = [A-Za-z]* || ${basename##*.} = 7z ) ]]
  do
    fileext=${basename##*.}.$fileext
    basename=${basename%.*}
  done
  fileext=${fileext%.}
  

  For POSIX portability, you need to use a case statement for pattern matching.

  while case $basename in
          ?*.*) case ${basename##*.} in [A-Za-z]*|7z) true;; *) false;; esac;;
          *) false;;
        esac
  do …
  

• Another possible specification is that some extensions denote encodings and indicate that further stripping is needed. Here's a bash/ksh/zsh implementation (requiring shopt -s extglob under bash and setopt ksh_glob under zsh):

  basename=$filename
  fileext=
  while [[ $basename = ?*.@(bz2|gz|lzma) ]]; do
    fileext=${basename##*.}.$fileext
    basename=${basename%.*}
  done
  if [[ $basename = ?*.* ]]; then
    fileext=${basename##*.}.$fileext
    basename=${basename%.*}
  fi
  fileext=${fileext%.}
  

  Note that this considers 0 to be an extension in file-1.0.gz.

¹ _{${VARIABLE##SUFFIX} and related constructs are in POSIX, so they work in any non-antique Bourne-style shell such as ash, bash, ksh or zsh.}

UNIX/Linux does not have the same early DOS / CP/M heritage that Windows does. So extensions are generally less significant to most UNIX utilities and tools.

I usually use a command-line only environment. Extensions in such an environment under Linux aren't really significant except as a convenience to the operator or user. (I don't have enough experience with KDE or GNOME to know how their filemanagers deal with extensions.)

But such convenience is usually important. If config.ini is really in Microsoft-standard ".ini" format, I'd let the extension stand. Plain old text files usually carry no extension in Linux, but this isn't universal for all programs configuration files. The programmer usually gets to decide that.

I think ".txt" is useful under Linux if you want to emphasize that it's NOT a configuration file or other machine-readable document. However, in source distributions, the convention is to name such files all caps without an extension (i.e. README, INSTALL, COPYING, etc.)

There are some standards and conventions but nothing stopping you from naming anything whatever you want, unless you are sharing things with others.

In Windows, naming a file .exe indicates to the shell (usually explorer.exe) that it's an executable file. UNIX builds this knowledge into the file system's permissions. If the proper x bits (see man chmod) are set, it is recognized as executable by shells and kernel functions (I believe). Beyond this, Linux doesn't care, most shells won't care, and most programs look in the file to find it's "type."

Of course, there's the nice command file which can analyze the file and tell you what it is with a degree of certainty. I believe if it can't match the data in the file with any known type, and if it contains only printable ASCII/Unicode characters, then it assumes its a text file.

@Bruce Ediger below is absolutely correct. There is nothing in the kernel or filesystem level, i.e. Linux itself, enforcing or caring that the contents of a file needs to match up with its name, or the program that is supposed to understand it. This doesn't mean it's not possible to create a shell or launcher utility to do things based on filename.