Finally, I found that somebody is still using EBCDIC today!

Anthropos’ FYP need to try out the asynchronous I/O. However, this is an “advanced” technique and hence not much reference can be found. By searching in Google, the following link can give you a piece of sample:

By extracting the compress package, there are some source code and a readme file. However, all are in EBCDIC code. The actual variant of EBCDIC is unknown, but I have to get the ASCII version anyway. Thanks god that there is GNU iconv!

The following is the way I do the conversion all in once:

for x in * ; do
    echo $x
    cat $x |
      perl -pe 's/\x15/\x25/g; s/\xAD/\x4A/g; s/\xBD/\x5A/g;' |
      iconv -f EBCDIC-INT -t ASCII >
    mv $ $x

The file out of the source is EBCDIC but with some strange character encoding. The newline character is coded as 0x15, and the pair of square brackets are coded as 0xAD and 0xBD. I check the correct EBCDIC code for these three characters by using:

echo "[]" | iconv -f ASCII -t EBCDIC-INT

and put them into the perl one-liner.