The specification for the maximum length of parallel cables is 3 meters (ten feet).
In parallel transfer, a logical 1 will be specified, typically, by putting something between +2.4v and +5v on a wire. A logical 0 will be specified by a voltage of +0v and +0.8v on the wire. The voltages applied to the wires are actually controlled by the (hardware) parallel driver.
Voltages between +0.8 and +2.4 volts is a 'gray area' and is handled differently by different parallel drivers. Some drivers will ignore values in this range (typically causing higher level software to involke error correction), while others might split the difference, considering anything over +1.6v to be a logical 1. Still others might utilize a narrower gray area. In any case, voltages in the middle of the range are indeterminate in most hardware configurations and will lead to unpredictable problems.
The most significant physical constraints that limit cable length are therefore the capacitance of the cable and the voltage drop along its length.
This is why some higher-grade cables, with low capacitance and low resistance, will sometimes allow longer parallel cable runs. Many parallel-driven CCD cameras come with such a cable.
This also explains why some devices work fine with longer cables. A high grade cable in a relatively noise-free environment, connected to parallel devices that have a rather narrowly defined gray area, will allow longer cable runs. The problem is that even if an embedded parallel controller on a camera has a very narrow gray zone, there is no assurance that the port on the PC does the same. If a mismatch occurrs, getting the device to work over long cable lengths can be problematic at best.
Parallel port cameras are a bad idea. The allowable, reliable cable lengths are very short, requiring that the computer be right next to the telescope. In some cases, the swing radius of the telescope exceeds the allowable cable length (in a cartoon, the control computer would be hoisted up by the scope, dangling from the end of the parallel cable). In most others, it is a pain in the neck to site the computer so close to the telescope and route the cables safely so that they can't catch on something and pull the camera out of the telescope.
The problems are exacerbated by the poor programming practices of most of the camera manufacturers. SBIG cameras in particular are examples of very poorly written parallel I/O drivers. When downloading an image from these cameras, the computer interrupts will be completely taken over by the device I/O. This can time out your other software, freeze the user interface, and cause other problems. It would be one thing if this kind of interrupt handling were necessary, but it isn't. It's merely a bad holdover from DOS programming by engineers that haven't adapted to a modern multitasking operating system.
For all of these reasons, it is worth looking for a camera that interfaces using a network interface, a FireWire port, or a USB connection (in that order of preference).