On Sun, 6 Jan 2002 13:44:39   
 Bill Potts wrote:
>Marcus:
>
>I'm sorry, but you are still missing the point.
>
I'm afraid not!  I guess it's you who continue to miss the point I was trying to make!

>(By the way, the II in ASCII (written without a space) isn't a Roman
>numeral. It stands for "Information Interchange." There was no "ASC I"  and
>there will never be an "ASC III." The current 8-bit version of ASCII is
>simply a subset of the more-encompassing 8-bit ISO code.)
>
I just wanted to come up with a term that would convey the idea of a "next generation" 
to the ASCII thing (I could have picked EBCDIC instead or any other scheme, I chose 
that one because of its popularity and the fact that it ended in II, I felt it would 
have had a nice... tune to my point.  ;-)   ).  But thanks for bringing that up.

>As for a greater number of bits, that's already been done. The new
>international standard is Unicode, which has 16 bits. I recommend you do a
>web search (with the Google search engine, for example) for Unicode. You'll
>turn up a wealth of sites.
>
That's the problem, binary mentality.  Unless they'd also want to include all other 
foreign language characters to the pool, 10 would have been enough.  My proposal would 
have been to make that table 20 bits long, then.
...
>We didn't arrive at base 10 for rational reasons. The rational choice would
>have been base 8 (if only we'd omitted the thumbs). For very practical
>reasons, I can see no change from the use of bases that are powers of 2 for
>computer arithmetic (at the machine level) and memory addressing.
>
?  I'm not sure I understood you here.  But if you concede that they should have used 
base 8, instead of 2, you should also be open to the idea of revamping hardware's 
architecture to move to decimal.  And not that I'm advocating that, BTW.

>Finally, a computer with a 100-bit bus would still be binary. It's the bit
>that's irrevocably binary.

?  Why do you continue to bring that up, Bill, when I already mentioned twice now that 
I was NOT discussing this point?  Didn't I make myself clear on my reply to your post 
already?  What else do you need from me to make that even clearer?

I can't care less if internally computers are binary in nature.  I, as a *human* user, 
am used to decimal-base.  If computer hardware could descend (or ascend, depending on 
how one sees it...) to my level at least in some aspects, it would make my life a 
little simpler and easier.  That was all I was trying to point out.

 The number of bits simply represents a power of
>two. For example, a 32-bit memory scheme will yield 2^32 possible unique
>addresses. A 100-bit scheme would yield 2^100.
>
?  You're preaching to the choir here, Bill.  I'm fully aware of that, obviously.  I 
have a degree in computer programming/analysis, so you don't need to remind me of 
that...  ;-)

>In any case, bus width is a speed issue, not an addressing one. A 64-bit bus
>can be used to retrieve twice the same amount of data as a 32-bit bus, given
>the same clock speed. Machines with 64-bit buses still use 32-bit addressing
>schemes.
>
Again, I'm also fully aware of that.  But I guess our audience would benefit for your 
opportune clarification above.  And indeed machines in the market with 64-bit buses 
are not "pure" 64-bit machines.  But one day they will (and I guess that will be 
pretty soon, less than 3-5 years, I'd say).

>On the subject of binary and decimal values, counting from the low order
>bit, we give the bits values of 1, 2, 4, 8, etc. Numbers, whether binary,
>octal, decimal or hexadecimal, have to be represented by groups of bits. All
>possible single-digit octal values can be represented by the seven possible
>values of the low-order three bits (1, 2, and 4). It takes four bits for
>either decimal or hexadecimal.

This is very educational to our audience here, and I'm glad you brought that up.  I 
also learned this kind of stuff in my 1st year of my Prog/An program.

Actually, I'd like to expand on that a bit.  3 bits can be used to represent up to 8 
numbers (2^3) in binary, i.e. 000, 001, 010, 011, 100, 101, 110, 111, hence all 
numbers from 0 to 7 would be covered, i.e. perfect for the octal system.  Evidently a 
group of 6 would be required to represent numbers in octal of two digits, and so forth.

With 4 bits you can represent 2^4, or 16, hence perfect for hexadecimal, but somewhat 
inefficient for decimal, since decimal base is not a perfect multiple of 2 of binary.  
I.e. we would have a waste of 6 numbers for that base if using a binary "half-byte" 
word length, as you rightfully pointed out below.

 The problem is that, in decimal, six of the
>possible values of the bit group are wasted. In doing decimal arithmetic on
>a computer, we cheerfully accept such waste.

Very good, true.

> To extend that to the
>computer's memory addressing scheme (which is the main application of binary
>numbers within a computer) wouldn't be very smart.
>
That's the reason why I didn't want to be perceived as advocating that here.  Let 
computers be binary all they want, but let's bring a more user-friendly interface with 
humans, even if they come at the expense of some inefficiencies, which may not be 
terribly bad.

>Finally, even if one could justify such a wasteful scheme, the absolute
>requirement for backward compatibility would rule it out.
>
?  Now you lost me.  I can't see why one would lose that, a higher number of bits 
should not interfere much with that aspect, you would just be using more bits to 
represent more addresses.  It's like with decimal places, you can represent a number 
as 10., or 10.0, or 10.00; or 10, 010, or 0010.  An address which was represented in, 
say, an 8-bit machine, as 00100101 would continue to be represented in a 10-bit 
machine the exact same way, but now as: 0000100101.  Therefore, would you care to 
elaborate, please?

>Might I recommend that you do some reading with respect to computer
>architecture. Conclusions that seem obvious from afar often turn out, on
>close examination, to be wrong ones.
>...
?  Why?  I've done enough during my school years.  Thanks but no thanks.  I'm passed 
that point.  But you're right on your last sentence above.  And I don't see we have 
any disagreement overall.  Perhaps another classical example of us talking the same 
thing while focussing on different aspects or something.

Marcus


Is your boss reading your email? ....Probably
Keep your messages private by using Lycos Mail.
Sign up today at http://mail.lycos.com

Reply via email to