Re: evolution of bytes, The joy of FORTRAN

John Levine <johnl@taugh.com> wrote:

According to Ted Nolan <tednolan> <tednolan>:

In article <vq2j3r$v1q6$2@dont-email.me>,
Lawrence D'Oliveiro  <ldo@nz.invalid> wrote:

On Sun, 2 Mar 2025 20:34:09 -0000 (UTC), John Levine wrote:
 

S/360 brought us the addressable 8 bit byte packaged into 16 bit
halfwords and 32 bit words, using the same addressing for each.

 
Did any machine offer “byte” addressability with “byte” having
any meaning
other than “8-bit quantity”?

 
As late as the last half of the 1980s, we ran some network operations
on a BB&N C-70 machine with 10 bit bytes. 

 
Good point, I'd forgotten about it.  It was a C-30 with two extra bits in
each byte to increase the address space from 16 to 20 bits.
 
I talked to one of the developers who told me with considerable frustration
how much C code implicitly assumed 8 bit bytes.  Well, duh.
 

C is supposed to be agnostic about data sizes, with a few specifications
such as “sizeof(short)<=sizeof(int)”. This was probably a reaction against
PL/I where sizes are specified explicitly, such as “bit(8) unsigned”.  As
usual programmers managed to sabotage this, just like they introduced
character code dependencies .

--
Pete