You're wrong in two different ways, which I kind of admire :)

Firstly, you have something that is between 7 and 8, so for arguments sake you call that '7 and a half' (this is regularly done).

Then you take another sample, and that scratches the '7 and a half' pieces, and not the 8, so it's between 7 and a half and 8. But that's a completely different scenario to arbitrarily assigning a '7.7'

There is another method of measuring hardness, the vicker's system which does indeed follow a linear scale, so you could potentially use the vicker's hardness of the specimens to determine an approximate decimal value for the hardness of your chosen sample. But that's completely stupid because the whole POINT of this is to show a relative hardness between two samples, ie to be able to measure it by scratching one piece against another. So my comment stands, the '7.7' value is pulled out of someone's ass and has no scientific merit.

The second problem is you then compare with software revision numbers... which are of course numbers pulled entirely out of someone's ass. And of course, version 5.9 is usually less than version 5.10 etc. so again you have no way of saying 'version 5.5 is about half way between version 5.0 and version 6.0 in features.'