Cell's programmer; 
from an old motherboard -> DB25 cable
you can make it in 4 minutes.

I made this new cable because I had problems with the old
(dapa) cable with avrdudes on win32 (same cable worked
fine lotsof times with uisp on linux - on another machine).
I didn't sort out completely
what the problem was (maybe an error in my selfmade avrdude.conf entry),
but this stk200 cable is more 
standard anyway:
programmer
  id    = "stk200";
    desc  = "STK200";
      type  = par;
        buff  = 4, 5;
          sck   = 6;
            mosi  = 7;
              reset = 9;
                miso  = 10;
                ;
I suggest you make a cable with this parallel-port pinout.
The header is documented on the schematic PDF.
Since I do not have a 74HCT245 buffer, I do not use parallel pins
4 and 5.
The PDI is used instead of MOSI,
PDO instead of MISO on atmega128. 
This is sure. (datasheet + it works this way)
-------------------------------------
OLD 'dapa' cable:

HDRpin#	AVRpin	parallelpin	parallelPin#	Note
1	PDO	Busy		11
2	VCC	-- 		--  		Used separate VCC
3	SCK	Strobe		1
4	PDI	data0		2
5	RESET	INIT		16
6	GND	GND		23		21..25

see file://localhost/usr/share/doc/hwb/html/connector/parallel/parallelpc.html
hardwarebook (use google)

$ uisp  -dprog=dapa --rd_fuses 
Atmel AVR ATmega128 is found.

Fuse Low Byte      = 0xe1
Fuse High Byte     = 0x99
Fuse Extended Byte = 0xfd
Calibration Byte   = 0x9f  --  Read Only
Lock Bits          = 0xff
    BLB12 -> 1
    BLB11 -> 1
    BLB02 -> 1
    BLB01 -> 1
      LB2 -> 1
      LB1 -> 1


# ATMEGA103 compatibility is a nightmare. TAKE CARE! TURN IT OFF!
$  uisp  -dprog=dapa --wr_fuse_e=0xff

External crystal oscillator, slow startup
# uisp  -dprog=dapa --wr_fuse_h=0x89
# uisp  -dprog=dapa --wr_fuse_l=0xff