March 5, 1988

Dear Bob:

	I'm immensely flattered that you should be using my editor for
anything at all.  I'm sending you a new version; it has a few added capa-
bilities, a few fewer bugs, & greatly improved color if you have an EGA, as
I wrote you in my last letter.  It cannot execute programs out of a window,
however; I'm not quite smart enough to do that, & executing anything costs
you about 250 k of memory, which I'd rather not spare.  Maybe that has
changed in Turbo 4.  Epsilon (a version of Emacs) lets you run programs out
of a window; however, Epsilon is big & awkward & has a user interface that
I was never able to get used to.

	I'm also including the C version; it's not bug-free, & it doesn't
have all the features that the Pascal version has (I only update the Pascal
version), but you may like to fiddle with it.  (Make sure you do any seri-
ous editing with ed.com, on the Maxell disc, not the C version.)  Please
feel free to criticize my C style; after a year or so of programming in C,
I still do not consider myself a C programmer.

	Two further observations about C: (a) Nat has suddenly declared
himself for Pascal again, especially Version 4; this afternoon, in the
thick of a frustrating debugging session, he suddenly burst out, "All
right.  I'll admit it: I hate C!"  (b) this quote from a recent Computing
Reviews: "If the mental investment in learning a language is greater than
the return in mental leverage for the production of correct programs, the
language fails."

	I'm also including a runoff program which you can use until you get
Write back again.  Its name is BIP: I originally wrote a bi-directional
printer for the Diablo, which I named BI; then when I wanted one for the
Epson, I modified BI & named the modified version BIP.  It is geared to the
Epson FX-80, but you can easily modify the commands as needed.  It supports:
	-automatic pagination (which can be turned off)
	-pause for sheet feeding (which can also be disabled)
	-continuous & non-continuous underscore
	-italics
	-phantom page-break (breaks if phantom is within 6 lines of
		bottom)
	-page numbering with optional headers (can also be disabled)
	-automatic centering
& a few other things.  I'll include a file (BIP.DIR) with directions for
running it.  (These programs are for Turbo 3, of course.)  The normal pro-
cedure is, first edit the file with Ed, then run it off with Bip.  (This is
like the Unix world, where you edit with something & then use Nroff.)  The
source is generously commented & should be easy to adapt; I imagine all you
would need would be an option switch for NLQ.  When my Diablo broke down a
year ago, I modified Bip to work with the HP LaserJet (called it, stupidly,
BIPL) & ran off essentials on D&B's laser printer: so I know it's not
difficult to customize for another printer.

	Your comments about the 386 were very illuminating.  As for hanging
multiple users on a PC, that's silly.  I agree with Pournelle's philosophy:
at least one computer per user.  (One way to get more than one is by way of
a LAN.  We're using IBM's PC-LAN at D&B; as long as you don't try to do any-
thing out of the ordinary with it, it's OK, but it's a memory pig, renders
the server virtually unusable for anything else, & hung on me repeatedly
when I tried to read/write to a 1/2" mag tape drive while using the net-
work.  It's termed a token ring, but curiously all the computers seem to be
connected in a star configuration.)  The question in my mind is how much
good a 386 will do you if you use ordinary software, like Turbo or (in
fact) most compilers, where the target machine is presumably still the
8088.  Of course, you get a fast clock speed, & I suppose a wider data
path, but what else?--unless you hand-optimize, as you have done.  Perhaps
we will see alternatives on the market where the back end of the compiler
generates 386 code, but I don't think we'll see them soon.

	Microsoft's assemblers are one huge mess, in my opinion; they've
always stuck me as having been designed by academics rather than computer
professionals.  Even making full allowance for the brain-damaged architec-
ture of the '86 family, it's a clumsy language with an ill-thought-out
instruction set.  I particularly dislike assemblers in which all data
transfers are indicated by Mov instructions.  I know it's beautifully
uniform.  So are upper-case letters, but we know from psychologists that
people read the slightly irregular lower-case characters more easily.  A
certain amount of non-uniformity actually facilitates comprehension: look
at human languages.

	Of the assembly languages I've used, I still regard the IBM 370's
as ideal.  I tell students that they will look back on the 370 with longing
when they program in assembly language for micros; I always get stares of
stark disbelief.  (Which reminds me of the gasp which went up from a data
structures class one day when they learned that I had programmed my first
quicksort in 370 assembly language.  I had been talking about removal of
recursion by means of a stack & had said that I had had to do that once.
Someone then asked me what language I had been using, that that should have
been necessary, which is how the story came out; I hadn't been aiming to
impress them, but I sure did.  When things like that come out these days,
I usually lessen the impact by pointing out to them how different it is in
the workplace, where a young programmer has 8 hours a day in which to do
nothing but write that program, with no other classes, no competing home-
work assignments, no term papers...etc.)

	The fights over improving the graduate program continue.  Joseph
Fulda & Phil Panzeca got in a shouting match two meetings ago; Phil kept
saying, "Where's your documentation [for all these claims]?"  I hope he
asks this again next time; I've run off a copy of your letter describing
all the stupid questions that were asked in Lincoln's class (suppressing
the names) & I'm going to read the whole thing aloud in the meeting if I'm
asked for documentation.  (I won't mention your name, although I doubt if
it would make any difference if I did; it would probably make you well-
liked by a number of people.)  I just got a letter from Poly, bragging
about how highly their program has been rated by some organization or
other.  I want to find out how one gets rated; it might give us some
valuable ammunition.

	Joseph didn't tell me he was using Hofstadter.  I find that daring
& original.  As for Hastings' class, I will include my Turbo-callable FFT
library; you may be interested in the games I play with the 8087 stack.
The file is called FFTLIB; it's called by a Pascal program, FFTDEMO, also
included.

	Joseph is up for tenure this Summer, & he is having nervous fits
over it.  He wants it very, very,  v e r y  badly & is sure he's going to
get shot down.  He may be, too: he's eccentric & argumentative, & he has
Phil Panzeca against him.  On the other hand, he has a better publications
record than anyone else in the department, with the possible exception of
Tom Steel, & he's brilliant & a demon for work.  He has published a couple
of very interesting papers recently, especially one quite original contri-
bution on estimating the semantic content of logical expressions.  I
pointed out to him that he might draw on information theory for this & gave
him a quick & dirty tutorial on it; he got all excited & would now like to
do a collaboration this Summer.  Wait till he sees how little information
theory I know.

	Joseph also inadvertently got me involved in another project.  Some-
how, in the course of a conversation once, I told Joseph that I thought the
book of Job was the only honest book in the Bible--& in January he found a
remaindered copy of some famous classic French biblical scholar's commen-
tary on Job & got it for me.  (A goyish commentator, by the way.)  There's
a famous parody of annotated copies of Shakespeare, which has two lines of
Shakespeare at the top of the page & commentary from there on down.  That's
the way this book is--a smaller text/commentary ratio even than in most
Hebrew books I've seen.

	I think he sent it me as a joke, but of course I started reading
it.  The commentary is in English, not French, but it's studded all over
with little bits of Hebrew.  So that reminded me of how long it's been
since I studied Hebrew, & how lousy I was at it when I did, & I started
trying to trying to puzzle out the little bits.  The upshot of all this was
that I found a really good Hebrew grammar & I've been at it since shortly
before the beginning of term.  I suppose it's weird for goy who's not a
clergyman to learn Hebrew, but I have the example of the critic, Edmund
Wilson, before me.

	The problem isn't the grammar, but the vocabulary, which is just
about impossible for a native speaker of English, since there are so few
cognates to aid the memory.  I've prepared a set of flash cards, & I drill
myself with them on the train mornings.  If I learn the vocabulary presen-
ted in the grammar, which is of course quar, which is of course questionable (I am not noted for
persistence in studying languages, & I'm surprised that this exploration
has lasted as long as it has), then I'll know every word that appears in
Tanach 100 times or more.  Considering the repetitiousness of much of the
writing, this should take one quite far.

	Predictably, when I see a word which has other associations, I
learn it almost immediately.  I learned "yada`," "he knows," [you'll have
to forgive my clumsy transliterations, & I must use quotes instead of
italics because the latter don't show up on the terminal], from a story I
once heard from a rabbi who was my office-mate at Federal Scientific, which
depended on the fact that Rashi says "lo yada`ti" three times in his com-
mentary.  Some of the words I recognize from Yiddish.  I got "ohel" right
away, because there was a shul in Riverdale called Congregation Ohel Torah,
& I was glad to understand, at last, that that meant Tabernacle of the
Torah.  (I believe "tabernaculum" means tent in Latin.)  But most of the
words are simply brute force.  I must be making some progress, however,
because Joseph also lent me a book of writings by Hirsch, & this had clumps
of unpointed text in it.  Some of those clumps were almost intelligible,
since I recognized words from my flash cards among them.

	When I first studied Hebrew, in my twenties, the textbook taught
a pronunciation that was neither Sephardic nor Aschkenasic, but some Cam-
bridge scholar's conjectural reconstruction of the speech of biblical
times.  Tav, in particular, has always given me trouble, because we were
taught to pronounce it as -th- (without the daghesh, that is).  It was no
better when I found that Jews themselves weren't unanimous on the letter,
some pronouncing it -s- & some -t-.  I now find myself wobbling irregularly
between -s- & -t-, which is probably going to get me in trouble some day.

	We were taught the "binyanim," & I can still remember most of their
names--the "pi`el," the "pu`al," the "niph`al," the "hiph`il," the
"hoph`al," the "hithpa`el" (note that -th-)--but the only one whose meaning
has stuck with me is the "niph`al," & that's because it occurs in the Pass-
over question, "Mah nishtaneh ... ?"  "What causes-to-be-different ... ?"
In my twenties, Piel's was one of New York's principal beers, & I never saw
the name without mentally pronouncing it "Pi`el's."

	To answer a possible question, I am  n o t  planning on writing a
Hebrew character-generation program for my HP plotter.  I did that once,
years ago, for a plotter at Federal Scientific, & that was enough.  (I had
written a general character-generation program, back in the days when
characters weren't stored in the plotter's ROM, & I designed the Hebrew set
to show how flexible the program w