PLEX86  x86- Virtual Machine (VM) Program
 Plex86  |  CVS  |  Mailing List  |  Download  |  Computer Folklore

The System360 Model 20 Wasn't As Bad As All That 3882


On Thu, 13 Jul 2006 13:17:01 -0400, "Rostyslaw J. Lewyckyj"

I'm not aware of any, offhand. One could, for example, have a print train with, say, four repebreastions of the 48-character set (instead of the 60-character set), and then have an extra 48 characters split into four groups of 12 between those repebreastions.

Thus, a 48-character job would print at the speed of one done with a 60-character train, and a 96-character job would print at the speed of one done with a 240-character train.

Obviously, if *most* jobs don't use lower case, but *do* use the full 60-character subset, this wouldn't optimize, it would pessimize.

On the other hand, a print train with three equally-spaced repebreastions of the 60-character set, accompanied by 60 additional characters split into three groups of 20 might actually be useful. With such a print train, you could print PN-style print jobs at 75% speed, and TN-style print jobs at 50% speed... but without ever changing the print train.

Adjusting based on linguistic *letter* frequencies, on the other hand, instead of by a very broad-brush approach based on subsets like this is *not* likely to have been tried, since it would be very difficult to make it useful.

The System360 Model 20 Wasn't As Bad As All That 3883
thread about using custom character sets on a 1403 train lots of snippage everywhere Not necessarily. If you have...
The System360 Model 20 Wasn't As Bad As All That 3887
I think the character set would have been called pica type (12 pt). I never dealt with the ALA...

One would have to choose a subset of the characters being employed such that most lines would be printed entirely, from beginning to end, only from characters in the subset being repeated more often. If that subset is to repeat more often than the original set from which it is abstracted, then the remainder of the original set can't be repeated very often at all - increasing the speed penalty for using it.

But I can finally think of a case where an actual speed *increase* could be obtained.

Let us say that one wanted a character set composed of the 60 characters on the PN print train and lower case letters - only. That is 86 characters.

The only choice seems to be to go all the way to the 120-character TN set.

That is only two repebreastions.

Three repebreastions allow an 80-character subset. This is too small.

But three repebreastions of a subset smaller than 80 characters, with each repebreastion divided into two equal parts, so that there are six gaps, containing two repetions of the remaining characters, create the possibility of faster printing when remaining within the smaller subset, and a result no worse when using the additional characters!

Let's see... 3a + 2b = 240, a+b = 86.

The System360 Model 20 Wasn't As Bad As All That 3884
On Sat, 15 Jul 2006 17:36:31 +0000 (UTC), Joe Morris If you start with *five* copies, you're already...

3a + 172 - 2a = 240, a = 240 - 172; a = 68.

Ouch! That's a small subset, and so we can only reduce the speed *loss* on upper-case only text, and not get a speed gain on text including lowercase.

What if we allow a worse result with the additional characters; can we get a bigger subset?

3a + b = 240, a+b = 86.

3a + 86 - b = 240; 2a = 240-86; 2a = 154.

Here we don't need to split a in half, so we have a subset of 77 characters repeated three times, with the remaining 9 characters, three at a time, filling in the gaps between them.

John Savard Usenet Zone Free Binaries Usenet Server More than 140,000 groups Unlimited download


List | Previous | Next

The System360 Model 20 Wasn't As Bad As All That 3883

Alt Folklore Computers Newsgroups

The System360 Model 20 Wasn't As Bad As All That 3881