Ralph Corderoy:
ed(1) pre-dates pipes. When pipes came along, stderr was needed, and
lots of new idioms were found to make use of them. Why didn't ed gain a
`filter' command to accompany `r !foo' and `w !bar'?
===
I sometimes wonder that too.
When I use `ed,' it is usually really qed, an extended ed
written by the late-1970s UNIX crowd here at U of T. (Rob
Pike, Tom Duff, David Tilbrook, and Hugh Redelmeier, I think.)
qed is something of a kitchen sink, full of clumsy programmability
features that I never use. The things that keep me using it are:
-- Multiple buffers, each possibly associated with a different
file or just anonymous
-- The ability to copy or move text (the traditional t and m
commands) between buffers as well as within one
-- The ability to send part or all of a buffer to a shell command,
to read data in from a shell command, or to send data out and
replace it with that from the shell command:
>mail user ...
<ps -ef
|tr a-z A-Z
I use the last quite often; it makes qed sort of a workbench for
manipulating and mining text. One can do the same with the shell
and temporary files, but using an editor buffer is much handier.
sam has similar abilities (but without all the needless programmability).
Were sam less clumsy to use in its non-graphical mode, I'd probably
have abandoned qed for sam.
Norman Wilson
Toronto ON (for real now)
> From: Ralph Corderoy
> Then the real definition, ending in an execution of the empty `q'.
> qq/4$^Ma2^[@qq
Gah. That reminds me of nothing so much as TECO (may it long Rest in Peace).
Noel
>Speaking of which, am I the only one annoyed by Penguin/OS' silly coloured
"ls" output?
Syntax coloring, of which IDE's seem to be enamored, always
looks to me like a ransom note. For folks who like colorized
text, Writers Workbench had a tool that can be harnessed to
do a bang-up job of syntax colorizing for English: "parts"
did a remarkable job of inferring parts of spechc in running
text.
Doug
I started with V6 Unix at UC Santa Barbara in 1977. I remember
that when V7 came out, I learned about the 'make' program and
started using it with great success to help me build a large
Fortran package for signal processing.
For its size, there was a lot going on in Santa Barbara at that
time. It was one of the first 4 Arpanet nodes, and there were
a bunch of companies making networking products and doing speech
research as a result.
I was a student at UC Santa Barbara but I started toying with
the idea of finding a real job, mostly to make more money.
I found several possibilities and went to interview at one.
This place had an a need for somebody to, in essence, be a
human 'make' program. The computer they used, some kind of
Data General, was so slow that they couldn't do a build more that
once or twice a day. So, in an attempt to speed up the build,
they wanted to hire somebody who would, by hand, keep track
of the last modification date of all the components in the
package they sold, and do a build that only performed
the necessary steps to generate the package - in other
words a human make program. Apparently they figured that
this would save enough time to justify the $24K salary they
were willing to pay. $24K in 1978 wasn't a bad salary at all.
I didn't take the job, but I've often thought that what I should
have done would have been to take the job under the condition
that I could mostly work remotely. Then, I could have used the
'make' program on our V7 Unix system to generate the optimal
script to build the package, and then taken the script back
to the company to run on the Data General computer. I figure
this would have taken maybe an hour a day. The rest of the time
I could have spent on the beach thinking about ways to spend that
$24K.
Jon Forrest
The infamous Morris Worm was released in 1988; making use of known
vulnerabilities in Sendmail/finger/RSH (and weak passwords), it took out a
metric shitload of SUN-3s and 4BSD Vaxen (the author claimed that it was
accidental, but the idiot hadn't tested it on an isolated network first).
A temporary "condom" was discovered by Rich Kulawiec with "mkdir /tmp/sh".
--
Dave Horsfall DTM (VK2KFU) "Those who don't understand security will suffer."
Hi All.
In 1983, while a grad student at Ga Tech, I did some contract programming
at Southern Bell. The system was a PDP 11/70 running USG Unix 4.0 (yes,
it existed! That's another story.)
Beside ed, the system had a screen editor named 'se' (not related to the
Ga Tech 'se' screen editor). It apparently was written within AT&T.
ISTR that it was written mainly for Vaxen but had been squeezed and made to
run on the PDP 11.
Did anyone else ever use this? Know anything about it? I never saw it
or heard it about it again; it didn't ship with System V.
Thanks,
Arnold
I am somewhat embarrassed to admit that this just occurred to me. Is the
reason that SIGKILL has the numeric value 9 because cats are reported to
have nine lives? Clearly the connection between 'cat' and 'kill -9' would
make for a irreverent but harmless inside joke if so....
- Dan C.
> I especially liked the bit in which Tom's virus infected a multi-level secured UNIX system that Doug McIlroy and Jim Reeds were developing which they didn't spot until they turned on all their protections ... and programs started crashing all over the place.
That's not quite right. The system was running nicely with a
lattice-based protection system (read from below, write to above)
working fine. Processes typocally begin at lattice bottom, but
move to hivel levels depending on what data they see (including,
of course any exec-ed file.) All the standard utilities, being
usable by anyone are at lattice bottom.
Cool, until you realize that highly trusted system programs
such as sudo are at lattice bottom and are protected only by
the old rwx bits, not by the read-write rules. So, following
an idea of Biba's, that integrity rules are the opposite of
secrecy rules. You don't want to forbid writing to high-integrity
places, nor read from low-integrity places.
This was done by setting the default security level away from
the lattice bottom. High-integrity stuff was below this floor;
high-secrecy above.
The Duff story is about the day we moved the floor off bottom.
An integrity violation during the boot sequence stopped the
system cold. Clearly we'd misconfigured something. But no, after
a couple of days of fruitless search, Jim Reeds lit up, "We
caught a virus!" We were unaware of Duff's experiment. He had
been chagrined when it escaped from one machine, but was able
to decontaminate all the machines in the center. Except ours,
which was not on the automatic software distrutioin list, since
it was running a different system.
> From: Andy Kosela
> That is why MIT and Bell Labs represented two very different cultures.
Oi! Not _everyone_ at MIT follows the "so complicated that there are no
obvious deficiencies" approach (to quote Hoare's wonderful aphorism from his
'Emperor's Old Clothes' Turing Award Lecture).
My personal design mantra (it's been at the top of my home page for decades)
is something I found as a footnote in Corbato and Saltzer, 'Multics: The First
Seven Years': "In anything at all, perfection has been attained, not when
there is nothing left to add, but when there is nothing left to take away..."
No doubt some people would be bemused that this should be in a Multics paper,
given the impression people have of Multics as incredibly - overly -
complicated. I'll avoid that discussion for the moment...
I've often tried to understand why some people create these incredibly
complicated systems. (Looking at the voluminous LISP Machine manual set from
Symbolics particularly caused this train of thought...) I think it's because
they are too smart - they can remember all that stuff.
Maybe my brain isn't like that (or perhaps I use large parts of it for other
stuff, like Japanese woodblock prints :-), but I much prefer simpler things.
Or maybe I'm just basically lazy, and like simpler things because they are
easier...
Noel
