I wrote:
"Your 4.7004 bits example is impossible. There can only ever be an integral number of on/off states."

Euric replied:
"I gave such an example?  I don't recall giving such an example.  Of course there can only be an integral number of bits.  That could be specified in the definition of the bit.  There can be 4.700 4 Mibit or 4.700 4 Mbit." 

Indeed, you did -- in the paragraph  you quoted from Rowlett in USMA 27962. When you quote sources, perhaps you should read what they say first. <g>
 
However, in what you say above, you have compounded the error, in that you're now talking about a non-integer number of mebibits or megabits. You must do the rounding up first. Thus, 26 possible character values require 5 bits for their representation. As five bits can be used to represent a total of 32 unique values, 6 values are simply unassigned in that case.
 
There is, of course, a 5-level code that uses all 32 possible values, of which two are reserved for shift characters. That code is International Telegraph Alphabet No. 2 (ITA-2), which is a variant of Baudot Code. Following a shift character, 30 of the values have a new meaning (with 2 values still reserved for the same shift characters (letters shift and figures shift). Thus, in data transmission (which is the only purpose of ITA-2), you can, in theory, represent 60 unique characters. (In practice, there are a some values that mean the same in either shift mode.)
 
To respond to your "That could be specified in the definition of the bit," the answer is no, it couldn't. A bit is a bit is a bit. The fact that there can only be an integral number is a simple mathematical fact and is implicit, not specifiable, in the definition.

Bill Potts, CMS
Roseville, CA
http://metric1.org [SI Navigator]

Reply via email to