[TUHS] Algol 68 and Unix (was cron and at ...)

M Douglas McIlroy m.douglas.mcilroy at dartmouth.edu
Fri Dec 18 00:21:14 AEST 2020

> Sometimes I wonder what would have happened if A68 had become the medium-level language of Unix

My knowledge of A68 comes from reading the official definition back in
the day. It took effort to see the clarity of the design through the
fog of the description. Until more accessible descriptions came along
(which I admit to not having seen) it would have been a big barrier to

A68 was very much in the air (though not much on the ground) in the
early days of Unix, as was PL/I. Although we had implemented and used
PL/I for Multics, it was never considered for Unix due to Its size and
the rise of other attractive languages during the 6-year gestation of
Multics. BCPL had the most influence, particularly in its clever
identity a[i] = i [a] = *(a+i).It was OK to write 2[x], which served
to implement structs like this: field=2; field[x]=3. (Actually the
BCPL indirection operator was not *. Dennis borrowed the more
perspicuous *from the SAP assembler for IBM 700-series machines.)

>From Algol 68 Dennis borrowed addition operators like +=, though at
first he unwisely reversed their spelling, underestimating the
inherent hazard of  a=-b. He rejected A68's automatic dereferencing as
too inflexible. He considered whether semicolons should be separators
as in Algol, or terminators as in PL/I. (Separators are more
convenient for macro substitution, but macros were not in the original
design.) He also considered making statements and expressions mutually
recursive as in A68. My recollection is that his choices were finally
based on a guess about user acceptance--how many radical innovations
would prospective users buy into. Perhaps Ken would have more to say
about this,

I tried to persuade Dennis to provide simultaneous assignments like
a,b = b,a, In the end, the comma operator got hijacked for a partial
realization of embedded statements.(We could still get parallel
assignment by, interpreting the existing {a,b,c} syntax as an lvalue
thus: {a,b,c}={b,c,a}.)

Then came Steve Bourne, with real experience in A68. Its influence on
the shell, including the story of do...done, has often been told. It
shows up most vividly in the condition part of if and while


More information about the TUHS mailing list