diff --git a/tbxi/__main__.py b/tbxi/__main__.py
index a139afb..4f8b678 100644
--- a/tbxi/__main__.py
+++ b/tbxi/__main__.py
@@ -11,6 +11,7 @@ import macresources
 from .slow_lzss import decompress
 
 from . import dispatcher
+from . import binhex
 
 
 def main(args=None):
@@ -59,7 +60,6 @@ def main(args=None):
 
             base, ext = path.splitext(args.file)
             if ext.lower() == '.hqx':
-                import binhex
                 hb = binhex.HexBin(f)
                 data = hb.read()
                 rsrc = list(macresources.parse_file(hb.read_rsrc()))
@@ -99,8 +99,6 @@ def main(args=None):
             data, rsrc = data # unpack the resource list from the data fork
             base, ext = path.splitext(args.output)
             if ext.lower() == '.hqx':
-                import binhex
-
                 finfo = binhex.FInfo()
                 finfo.Creator = b'chrp'
                 finfo.Type = b'tbxi'
diff --git a/tbxi/binhex.py b/tbxi/binhex.py
new file mode 100644
index 0000000..56b5f85
--- /dev/null
+++ b/tbxi/binhex.py
@@ -0,0 +1,479 @@
+"""Macintosh binhex compression/decompression.
+
+easy interface:
+binhex(inputfilename, outputfilename)
+hexbin(inputfilename, outputfilename)
+"""
+
+#
+# Jack Jansen, CWI, August 1995.
+#
+# The module is supposed to be as compatible as possible. Especially the
+# easy interface should work "as expected" on any platform.
+# XXXX Note: currently, textfiles appear in mac-form on all platforms.
+# We seem to lack a simple character-translate in python.
+# (we should probably use ISO-Latin-1 on all but the mac platform).
+# XXXX The simple routines are too simple: they expect to hold the complete
+# files in-core. Should be fixed.
+# XXXX It would be nice to handle AppleDouble format on unix
+# (for servers serving macs).
+# XXXX I don't understand what happens when you get 0x90 times the same byte on
+# input. The resulting code (xx 90 90) would appear to be interpreted as an
+# escaped *value* of 0x90. All coders I've seen appear to ignore this nicety...
+#
+import io
+import os
+import struct
+import binascii
+
+__all__ = ["binhex","hexbin","Error"]
+
+class Error(Exception):
+    pass
+
+# States (what have we written)
+_DID_HEADER = 0
+_DID_DATA = 1
+
+# Various constants
+REASONABLY_LARGE = 32768  # Minimal amount we pass the rle-coder
+LINELEN = 64
+RUNCHAR = b"\x90"
+
+#
+# This code is no longer byte-order dependent
+
+
+class FInfo:
+    def __init__(self):
+        self.Type = '????'
+        self.Creator = '????'
+        self.Flags = 0
+
+def getfileinfo(name):
+    finfo = FInfo()
+    with io.open(name, 'rb') as fp:
+        # Quick check for textfile
+        data = fp.read(512)
+        if 0 not in data:
+            finfo.Type = 'TEXT'
+        fp.seek(0, 2)
+        dsize = fp.tell()
+    dir, file = os.path.split(name)
+    file = file.replace(':', '-', 1)
+    return file, finfo, dsize, 0
+
+class openrsrc:
+    def __init__(self, *args):
+        pass
+
+    def read(self, *args):
+        return b''
+
+    def write(self, *args):
+        pass
+
+    def close(self):
+        pass
+
+class _Hqxcoderengine:
+    """Write data to the coder in 3-byte chunks"""
+
+    def __init__(self, ofp):
+        self.ofp = ofp
+        self.data = b''
+        self.hqxdata = b''
+        self.linelen = LINELEN - 1
+
+    def write(self, data):
+        self.data = self.data + data
+        datalen = len(self.data)
+        todo = (datalen // 3) * 3
+        data = self.data[:todo]
+        self.data = self.data[todo:]
+        if not data:
+            return
+        self.hqxdata = self.hqxdata + binascii.b2a_hqx(data)
+        self._flush(0)
+
+    def _flush(self, force):
+        first = 0
+        while first <= len(self.hqxdata) - self.linelen:
+            last = first + self.linelen
+            self.ofp.write(self.hqxdata[first:last] + b'\n')
+            self.linelen = LINELEN
+            first = last
+        self.hqxdata = self.hqxdata[first:]
+        if force:
+            self.ofp.write(self.hqxdata + b':\n')
+
+    def close(self):
+        if self.data:
+            self.hqxdata = self.hqxdata + binascii.b2a_hqx(self.data)
+        self._flush(1)
+        self.ofp.close()
+        del self.ofp
+
+class _Rlecoderengine:
+    """Write data to the RLE-coder in suitably large chunks"""
+
+    def __init__(self, ofp):
+        self.ofp = ofp
+        self.data = b''
+
+    def write(self, data):
+        self.data = self.data + data
+        if len(self.data) < REASONABLY_LARGE:
+            return
+        rledata = binascii.rlecode_hqx(self.data)
+        self.ofp.write(rledata)
+        self.data = b''
+
+    def close(self):
+        if self.data:
+            rledata = binascii.rlecode_hqx(self.data)
+            self.ofp.write(rledata)
+        self.ofp.close()
+        del self.ofp
+
+class BinHex:
+    def __init__(self, name_finfo_dlen_rlen, ofp):
+        name, finfo, dlen, rlen = name_finfo_dlen_rlen
+        close_on_error = False
+        if isinstance(ofp, str):
+            ofname = ofp
+            ofp = io.open(ofname, 'wb')
+            close_on_error = True
+        try:
+            ofp.write(b'(This file must be converted with BinHex 4.0)\r\r:')
+            hqxer = _Hqxcoderengine(ofp)
+            self.ofp = _Rlecoderengine(hqxer)
+            self.crc = 0
+            if finfo is None:
+                finfo = FInfo()
+            self.dlen = dlen
+            self.rlen = rlen
+            self._writeinfo(name, finfo)
+            self.state = _DID_HEADER
+        except:
+            if close_on_error:
+                ofp.close()
+            raise
+
+    def _writeinfo(self, name, finfo):
+        nl = len(name)
+        if nl > 63:
+            raise Error('Filename too long')
+        d = bytes([nl]) + name.encode("latin-1") + b'\0'
+        tp, cr = finfo.Type, finfo.Creator
+        if isinstance(tp, str):
+            tp = tp.encode("latin-1")
+        if isinstance(cr, str):
+            cr = cr.encode("latin-1")
+        d2 = tp + cr
+
+        # Force all structs to be packed with big-endian
+        d3 = struct.pack('>h', finfo.Flags)
+        d4 = struct.pack('>ii', self.dlen, self.rlen)
+        info = d + d2 + d3 + d4
+        self._write(info)
+        self._writecrc()
+
+    def _write(self, data):
+        self.crc = binascii.crc_hqx(data, self.crc)
+        self.ofp.write(data)
+
+    def _writecrc(self):
+        # XXXX Should this be here??
+        # self.crc = binascii.crc_hqx('\0\0', self.crc)
+        if self.crc < 0:
+            fmt = '>h'
+        else:
+            fmt = '>H'
+        self.ofp.write(struct.pack(fmt, self.crc))
+        self.crc = 0
+
+    def write(self, data):
+        if self.state != _DID_HEADER:
+            raise Error('Writing data at the wrong time')
+        self.dlen = self.dlen - len(data)
+        self._write(data)
+
+    def close_data(self):
+        if self.dlen != 0:
+            raise Error('Incorrect data size, diff=%r' % (self.rlen,))
+        self._writecrc()
+        self.state = _DID_DATA
+
+    def write_rsrc(self, data):
+        if self.state < _DID_DATA:
+            self.close_data()
+        if self.state != _DID_DATA:
+            raise Error('Writing resource data at the wrong time')
+        self.rlen = self.rlen - len(data)
+        self._write(data)
+
+    def close(self):
+        if self.state is None:
+            return
+        try:
+            if self.state < _DID_DATA:
+                self.close_data()
+            if self.state != _DID_DATA:
+                raise Error('Close at the wrong time')
+            if self.rlen != 0:
+                raise Error("Incorrect resource-datasize, diff=%r" % (self.rlen,))
+            self._writecrc()
+        finally:
+            self.state = None
+            ofp = self.ofp
+            del self.ofp
+            ofp.close()
+
+def binhex(inp, out):
+    """binhex(infilename, outfilename): create binhex-encoded copy of a file"""
+    finfo = getfileinfo(inp)
+    ofp = BinHex(finfo, out)
+
+    with io.open(inp, 'rb') as ifp:
+        # XXXX Do textfile translation on non-mac systems
+        while True:
+            d = ifp.read(128000)
+            if not d: break
+            ofp.write(d)
+        ofp.close_data()
+
+    ifp = openrsrc(inp, 'rb')
+    while True:
+        d = ifp.read(128000)
+        if not d: break
+        ofp.write_rsrc(d)
+    ofp.close()
+    ifp.close()
+
+class _Hqxdecoderengine:
+    """Read data via the decoder in 4-byte chunks"""
+
+    def __init__(self, ifp):
+        self.ifp = ifp
+        self.eof = 0
+
+    def read(self, totalwtd):
+        """Read at least wtd bytes (or until EOF)"""
+        decdata = b''
+        wtd = totalwtd
+        #
+        # The loop here is convoluted, since we don't really now how
+        # much to decode: there may be newlines in the incoming data.
+        while wtd > 0:
+            if self.eof: return decdata
+            wtd = ((wtd + 2) // 3) * 4
+            data = self.ifp.read(wtd)
+            #
+            # Next problem: there may not be a complete number of
+            # bytes in what we pass to a2b. Solve by yet another
+            # loop.
+            #
+            while True:
+                try:
+                    decdatacur, self.eof = binascii.a2b_hqx(data)
+                    break
+                except binascii.Incomplete:
+                    pass
+                newdata = self.ifp.read(1)
+                if not newdata:
+                    raise Error('Premature EOF on binhex file')
+                data = data + newdata
+            decdata = decdata + decdatacur
+            wtd = totalwtd - len(decdata)
+            if not decdata and not self.eof:
+                raise Error('Premature EOF on binhex file')
+        return decdata
+
+    def close(self):
+        self.ifp.close()
+
+class _Rledecoderengine:
+    """Read data via the RLE-coder"""
+
+    def __init__(self, ifp):
+        self.ifp = ifp
+        self.pre_buffer = b''
+        self.post_buffer = b''
+        self.eof = 0
+
+    def read(self, wtd):
+        if wtd > len(self.post_buffer):
+            self._fill(wtd - len(self.post_buffer))
+        rv = self.post_buffer[:wtd]
+        self.post_buffer = self.post_buffer[wtd:]
+        return rv
+
+    def _fill(self, wtd):
+        self.pre_buffer = self.pre_buffer + self.ifp.read(wtd + 4)
+        if self.ifp.eof:
+            self.post_buffer = self.post_buffer + \
+                binascii.rledecode_hqx(self.pre_buffer)
+            self.pre_buffer = b''
+            return
+
+        #
+        # Obfuscated code ahead. We have to take care that we don't
+        # end up with an orphaned RUNCHAR later on. So, we keep a couple
+        # of bytes in the buffer, depending on what the end of
+        # the buffer looks like:
+        # '\220\0\220' - Keep 3 bytes: repeated \220 (escaped as \220\0)
+        # '?\220' - Keep 2 bytes: repeated something-else
+        # '\220\0' - Escaped \220: Keep 2 bytes.
+        # '?\220?' - Complete repeat sequence: decode all
+        # otherwise: keep 1 byte.
+        #
+        mark = len(self.pre_buffer)
+        if self.pre_buffer[-3:] == RUNCHAR + b'\0' + RUNCHAR:
+            mark = mark - 3
+        elif self.pre_buffer[-1:] == RUNCHAR:
+            mark = mark - 2
+        elif self.pre_buffer[-2:] == RUNCHAR + b'\0':
+            mark = mark - 2
+        elif self.pre_buffer[-2:-1] == RUNCHAR:
+            pass # Decode all
+        else:
+            mark = mark - 1
+
+        self.post_buffer = self.post_buffer + \
+            binascii.rledecode_hqx(self.pre_buffer[:mark])
+        self.pre_buffer = self.pre_buffer[mark:]
+
+    def close(self):
+        self.ifp.close()
+
+class HexBin:
+    def __init__(self, ifp):
+        if isinstance(ifp, str):
+            ifp = io.open(ifp, 'rb')
+        #
+        # Find initial colon.
+        #
+        while True:
+            ch = ifp.read(1)
+            if not ch:
+                raise Error("No binhex data found")
+            # Cater for \r\n terminated lines (which show up as \n\r, hence
+            # all lines start with \r)
+            if ch == b'\r':
+                continue
+            if ch == b':':
+                break
+
+        hqxifp = _Hqxdecoderengine(ifp)
+        self.ifp = _Rledecoderengine(hqxifp)
+        self.crc = 0
+        self._readheader()
+
+    def _read(self, len):
+        data = self.ifp.read(len)
+        self.crc = binascii.crc_hqx(data, self.crc)
+        return data
+
+    def _checkcrc(self):
+        filecrc = struct.unpack('>h', self.ifp.read(2))[0] & 0xffff
+        #self.crc = binascii.crc_hqx('\0\0', self.crc)
+        # XXXX Is this needed??
+        self.crc = self.crc & 0xffff
+        if filecrc != self.crc:
+            raise Error('CRC error, computed %x, read %x'
+                        % (self.crc, filecrc))
+        self.crc = 0
+
+    def _readheader(self):
+        len = self._read(1)
+        fname = self._read(ord(len))
+        rest = self._read(1 + 4 + 4 + 2 + 4 + 4)
+        self._checkcrc()
+
+        type = rest[1:5]
+        creator = rest[5:9]
+        flags = struct.unpack('>h', rest[9:11])[0]
+        self.dlen = struct.unpack('>l', rest[11:15])[0]
+        self.rlen = struct.unpack('>l', rest[15:19])[0]
+
+        self.FName = fname
+        self.FInfo = FInfo()
+        self.FInfo.Creator = creator
+        self.FInfo.Type = type
+        self.FInfo.Flags = flags
+
+        self.state = _DID_HEADER
+
+    def read(self, *n):
+        if self.state != _DID_HEADER:
+            raise Error('Read data at wrong time')
+        if n:
+            n = n[0]
+            n = min(n, self.dlen)
+        else:
+            n = self.dlen
+        rv = b''
+        while len(rv) < n:
+            rv = rv + self._read(n-len(rv))
+        self.dlen = self.dlen - n
+        return rv
+
+    def close_data(self):
+        if self.state != _DID_HEADER:
+            raise Error('close_data at wrong time')
+        if self.dlen:
+            dummy = self._read(self.dlen)
+        self._checkcrc()
+        self.state = _DID_DATA
+
+    def read_rsrc(self, *n):
+        if self.state == _DID_HEADER:
+            self.close_data()
+        if self.state != _DID_DATA:
+            raise Error('Read resource data at wrong time')
+        if n:
+            n = n[0]
+            n = min(n, self.rlen)
+        else:
+            n = self.rlen
+        self.rlen = self.rlen - n
+        return self._read(n)
+
+    def close(self):
+        if self.state is None:
+            return
+        try:
+            if self.rlen:
+                dummy = self.read_rsrc(self.rlen)
+            self._checkcrc()
+        finally:
+            self.state = None
+            self.ifp.close()
+
+def hexbin(inp, out):
+    """hexbin(infilename, outfilename) - Decode binhexed file"""
+    ifp = HexBin(inp)
+    finfo = ifp.FInfo
+    if not out:
+        out = ifp.FName
+
+    with io.open(out, 'wb') as ofp:
+        # XXXX Do translation on non-mac systems
+        while True:
+            d = ifp.read(128000)
+            if not d: break
+            ofp.write(d)
+    ifp.close_data()
+
+    d = ifp.read_rsrc(128000)
+    if d:
+        ofp = openrsrc(out, 'wb')
+        ofp.write(d)
+        while True:
+            d = ifp.read_rsrc(128000)
+            if not d: break
+            ofp.write(d)
+        ofp.close()
+
+    ifp.close()