[game_preservation] Article about proper SNES emulation

[Christian Bartsch | softpres.org]

> That was a pretty fascinating article, especially outlining problems 

> with the preservation.

>

> One question I had is what the bottlenecks are for preservation. We've 

> identified decaying floppy disks and other magnetic media as one of 

> the most pressing concerns, but Kryoflux technology has been allowing 

> us to solve this problem.

>

I'd like to chime in here. KryoFlux is a start, and will help making a 
snapshot of data present on a floppy disk today. More below.


> However, the problem then becomes how to accurately run the software 

> again. From the sound of it, emulating the timing of the machine would 

> not be possible for early computer systems. To run the software on the 

> original hardware, it is possible to use Kryoflux to refresh the data 

> onto a floppy disk, effectively extending the life.


Beware. This is of course true for all disks that follow a well known 
encoding scheme, e.g. all disks that are standard DOS (pick your 
favourite OS here). But this is not true for disks that use copy 
protection mechanisms that create ambiguity (e.g. weak bits). If the 
original bit pattern is unknown, e.g. bits changing every revolution, 
you can not write this very pattern back to disk as you will remove 
ambiguity. In this case using KryoFlux to refresh a disk will ultimately 
destroy your original, unless you have kept a forensic dump that 
contains multiple revolutions. Based on this, an IPF could be created, 
which can hold the information necessary to truly rewrite the disk.

For protections that don't use ambigutiy, doing a blind write of raw 
data will also come with a drawback. Imagine scanning a photo at high 
resolution to do another print. Without cleaning it up, the print will 
contain noise and dust present on the original. Re-scanning and 
re-printing again will add  more noise and dust, apart from aliasing, 
that will also add up. In a world of bits, where a single bad bit in 
game code can render a program useless, such processing is also 
problematic. Therefore, while this will most likely work for the first 
gen copy, multi generations should be avoided. Again, converting the 
dump to something that is known-good, e.g. an IPF as the result of an 
analysation process, is very much recommended.

You will also basically destroy the option to use such a floppy as the 
source for preservation; the moment you rewrite it with KryoFlux you 
invalidate it as an authentic source. This is not a problem once the 
dump made in the first place has been confirmed good and authentic.


> Another option could be floppy disk emulation, but I'm not sure how 

> accurately this is, either.


The same objections from above would apply to emulation; the problem 
occours when reading. Data needs to be interpreted and analysed, after 
that it does not make a difference if you write it back to disk or 
emulate it.