kateveni@Shasta (05/25/84)
From: Manolis Katevenis <kateveni@Shasta>
------- Forwarded Message
Date: Fri, 25 May 84 11:50:52 edt
From: macrakis@harvard (Stavros Macrakis)
Message-Id: <8405251550.AA20833@harvard.ARPA>
To: ELLHNES@Berkeley, kateveni@Berkeley
Subject: Re: (forw-net-42): Glwssa gia anaptyxi systimatwn stin Ellada ?
Cc: macrakis@harvard
Oti i C ine hrisimi sto Unix, den amfivolo. Na exiyiso kalitera sta Anglika.
1. Machine-oriented programming model. Rather than encouraging thought in
terms of the concepts being programmed about, it encourages implementation-
level constructs: everything is thought of in terms of its constituent
bytes.
2. Very poor type checking. Although more and more is being done, partly by
Lint and partly by the compiler, still many type errors are undetected.
Worse, many existing programs depend on type punning for their correct
operation.
3. Terrible diagnostics in all compilers I know. One typically gets an
error message unrelated to the actual error, then a dozen errors after-
wards because of poor error recovery.
4. Difficult and confusing syntax. Many irregularities (e.g. aggregates exist
for initialization of static objects only).
5. Difficult and confusing semantics (consider semantics of array names,
for instance).
6. No nested declarations of functions to support top-down programming.
7. No runtime checking of ANYTHING (pointer dereference, array reference,
arithmetic overflow...) in any implementation I know.
C has the advantage, of course, that it is the Unix systems programming
language. But it's really just machine-independent machine language. Also,
future languages seem unlikely to be based on C, as they have been on Algol
for so long. An Algol-like language of some kind seems a better choice
if there is to be a single standard (even some dialect of Pascal). Ada would
be best if the schedule you have in mind allows the wait for better facilities.
Stavros
------- End of Forwarded Message