--- /dev/null
+
+#------------------------------------------------------------------------------
+# compress: file(1) magic for pure-compression formats (no archives)
+#
+# compress, gzip, pack, compact, huf, squeeze, crunch, freeze, yabba, etc.
+#
+# Formats for various forms of compressed data
+# Formats for "compress" proper have been moved into "compress.c",
+# because it tries to uncompress it to figure out what's inside.
+
+# standard unix compress
+0 string \037\235 compress'd data
+>2 byte&0x80 >0 block compressed
+>2 byte&0x1f x %d bits
+
+# gzip (GNU zip, not to be confused with Info-ZIP or PKWARE zip archiver)
+0 string \037\213 gzip compressed data
+>2 byte <8 \b, reserved method,
+>2 byte 8 \b, deflated,
+>3 byte &0x01 ASCII,
+>3 byte &0x02 continuation,
+>3 byte &0x04 extra field,
+>3 byte &0x08 original filename,
+>3 byte &0x10 comment,
+>3 byte &0x20 encrypted,
+>4 ledate x last modified: %s,
+>8 byte 2 max compression,
+>8 byte 4 max speed,
+>9 byte =0x00 os: MS-DOS
+>9 byte =0x01 os: Amiga
+>9 byte =0x02 os: VMS
+>9 byte =0x03 os: Unix
+>9 byte =0x05 os: Atari
+>9 byte =0x06 os: OS/2
+>9 byte =0x07 os: MacOS
+>9 byte =0x0A os: Tops/20
+>9 byte =0x0B os: Win/32
+
+# packed data, Huffman (minimum redundancy) codes on a byte-by-byte basis
+0 string \037\036 packed data
+>2 belong >1 \b, %d characters originally
+>2 belong =1 \b, %d character originally
+#
+# This magic number is byte-order-independent. XXX - Does that mean this
+# is big-endian, little-endian, either, or that you can't tell?
+# this short is valid for SunOS
+0 short 017437 old packed data
+
+# XXX - why *two* entries for "compacted data", one of which is
+# byte-order independent, and one of which is byte-order dependent?
+#
+0 short 0x1fff compacted data
+# This string is valid for SunOS (BE) and a matching "short" is listed
+# in the Ultrix (LE) magic file.
+0 string \377\037 compacted data
+0 short 0145405 huf output
+
+# Squeeze and Crunch...
+# These numbers were gleaned from the Unix versions of the programs to
+# handle these formats. Note that I can only uncrunch, not crunch, and
+# I didn't have a crunched file handy, so the crunch number is untested.
+# Keith Waclena <keith@cerberus.uchicago.edu>
+0 leshort 0x76FF squeezed data (CP/M, DOS)
+0 leshort 0x76FE crunched data (CP/M, DOS)
+
+# Freeze
+0 string \037\237 frozen file 2.1
+0 string \037\236 frozen file 1.0 (or gzip 0.5)
+
+# SCO compress -H (LZH)
+0 string \037\240 SCO compress -H (LZH) data
+
+# European GSM 06.10 is a provisional standard for full-rate speech
+# transcoding, prI-ETS 300 036, which uses RPE/LTP (residual pulse
+# excitation/long term prediction) coding at 13 kbit/s.
+#
+# There's only a magic nibble (4 bits); that nibble repeats every 33
+# bytes. This isn't suited for use, but maybe we can use it someday.
+#
+# This will cause very short GSM files to be declared as data and
+# mismatches to be declared as data too!
+#0 byte&0xF0 0xd0 data
+#>33 byte&0xF0 0xd0
+#>66 byte&0xF0 0xd0
+#>99 byte&0xF0 0xd0
+#>132 byte&0xF0 0xd0 GSM 06.10 compressed audio
+
+# Bzip from ulmo@Q.Net
+0 string BZ bzip compressed data,
+>2 byte x format v. %c,
+>3 byte x block size indicator %c
+