Minutes of the FIGGY BAR RT Conference.

Date: 11/16/89   Time: 22:36EST

Attendees:
[[Gary] GARY-S]
[[Jax It Is] FIGCHAPTERS]
[[Mark] M.FOSTER3]
[[Dave] OLORIN]
[[Kevin] APPERT]
[[Steve] S.WHEELER]
[[Wil] W.BADEN1]
[[Len] NMORGENSTERN]
[DRMIKE]

 FIGGY BAR of 16 November 1989 (ARC'd).The question of the night:"ROLL and
 PICK have been removed from the 'required' wordset by the ANS TC. Do you 
 care ? Why ?". Other items discussed include Wheeler/Sarns February guest
 conference, Charles Curley's just completed conference, Wescon, Local
 variables, wire-wrapped planes in trouble, various chips, FOR/NEXT vs 
 DO/LOOP, TO , Pascal in Forth

Minutes:
<OLORIN> is here.
<APPERT> is here.
<[Dave] OLORIN> Answer: I don't use them...I don't care one way or another.
<S.WHEELER> is here.
<S.WHEELER> Hello.
<[Kevin] APPERT> enjoyed last week with CC very much, Gary
<[Dave] OLORIN> I personally find them rather inelegant, but that is a matter 
of prejuidice (I learned Forth while not using ROLL and PICK, and every
time I've ever felt the need, changing things around to make them unneeded made 
the code cleaner).
<[Dave] OLORIN> Hi Steve
<[Gary] GARY-S> Wheeler is half the February guest contingent
<[Dave] OLORIN> What's the temp like in Little Rock, Gary?
<[Gary] GARY-S> COLD !
<[Gary] GARY-S> Dropped 52 degrees in one day
<[Dave] OLORIN> Yeah..it's going down into the upper thirties here tonight.
<[Gary] GARY-S> Tornados out the kazoo
<S.WHEELER> Personally, I have never used ROLL, but I tend to use "2 PICK" a 
fair amount.
<[Kevin] APPERT> this old neighborhood isn't the same without Charles, and it 
was good to hear from him
<[Wil] W.BADEN1> is here.
<[Kevin] APPERT> I hear Huntsville got hit pretty hard
<[Dave] OLORIN> (Has the fire place roaring away)
<[Gary] GARY-S> Steve - I assume you and Sarns have accepted the Feb invitation 
<S.WHEELER> I have, but I'm not certain the boss knows he's scheduled, yet.
<[Gary] GARY-S> He will in time. right :-)
<S.WHEELER> Maybe JAX will get around to letting him know sometime :-)
<[Gary] GARY-S> Taking bets ?
<[Kevin] APPERT> anybody else go to Wescon?
<[Gary] GARY-S> Wil - care to answer the question
<[Dave] OLORIN> Wescon?
<[Wil] W.BADEN1> ok
<S.WHEELER> Western Con-tucky?
<[Wil] W.BADEN1> I think that they
<[Wil] W.BADEN1> are theoretically important.
<[Kevin] APPERT> Wescon is an anual trade show
<[Dave] OLORIN> Ah.
<[Kevin] APPERT> for the electronics industry
<[Wil] W.BADEN1> I programmed in Forth for 8 years before using ROLL.
<[Kevin] APPERT> alternately held in San Francisco and L.A.
<[Wil] W.BADEN1> For 7 years before using anything but
<[Wil] W.BADEN1> 2 PICK  except for .S .
<[Dave] OLORIN> (still cannot remember using PICK or ROLL in code)
<[Gary] GARY-S> I still don't use ROLL - PICK only rarely, so I could care less
<S.WHEELER> I use "2 PICK" often enough that I'd hate to replace it with
<S.WHEELER> " ROT DUP >R -ROT R>" , but I have yet to use ROLL in over 10 years 
of Forth programming.
<[Kevin] APPERT> : compost ( 1 2 3 --- 1 2 3 )   rot rot rot ;
<[Wil] W.BADEN1> I believe that much more useful words would be OVER2 (or PLUCK)
 and OVER3.
<[Gary] GARY-S> That would be truer to the 'self-defining' concept for sure
<[Wil] W.BADEN1> The word that would be most helpful to me would be +UNDER.
<[Dave] OLORIN> This is a very good reason to keep them out of the required set.
 If most FoForth programmers do not use them, making them required just
 swells the size of the kernel for the rest of us.
<[Dave] OLORIN> Yeah, old habits die hard...I use the return stack before I use 
ROLL or PICK or anything like that.
<[Wil] W.BADEN1> With OVER2 and OVER3 there wud be less cry for local variables.
<[Dave] OLORIN> Didn't a LOCAL definition just make it into the working BASIS? 
(I seem to remember something about that on comp.lang.forth)
<[Wil] W.BADEN1> Yes.
<[Dave] OLORIN> Which wordset?]
<[Kevin] APPERT> Wescon had a broad assortment of CAE tools (schematic capture, 
pcb autorouters,tdigital,  analog, thermal simulations, netlist->PAL c
onversion, etc.)
<[Wil] W.BADEN1> Let me look. (Minutes plus came today.)
<[Kevin] APPERT> a long time user of Forth, HiTech Equipment Co. of San Diego, 
had a product which may spell the end of the wire-wrapped prototype.  They
have a plotter which writes an etch-resist pattern to the copper directly.  
etch and done.
<[Dave] OLORIN> (long low whistle)
<S.WHEELER> Probably can't match the cost, though.
<[Dave] OLORIN> Now that is an idea whose time has come!
<[Gary] GARY-S> yet
<[Kevin] APPERT> cost of what?
<[Dave] OLORIN> Yeah...but who wants to try and wire-wrap for a bed-of-nails 
chip?
<S.WHEELER> Speaking strictly as a software type, I'd give it to a tech. :-)
<[Dave] OLORIN> Remember to figure in your time when considering the cost of 
wire-wrap...it gets expensive when you start figuring labor costs.
<[Dave] OLORIN> Heh...the book I used to learn 68k assembly language was 
talking about debugging tools...
<[Kevin] APPERT> who wants to try a wire-wrap prototype of any fast or analog 
logic?  they just are not valid.
<[Gary] GARY-S> wire-wrapping does afford the opportunity to make engineering 
changes (continuation engineering) in the field. You'd play hell doing that
with an etched back-plane
<M.FOSTER3> is here.
<[Dave] OLORIN> 1) ICE is second-best
<[Dave] OLORIN> 2) Debuggers are third-best
<[Dave] OLORIN> 3) Write statements are fourth-best
<[Dave] OLORIN> 4) Handing it off to junior engineers and forgetting about it 
took the number one spot. :-)
<S.WHEELER> That's the book by Krantz & Stanley, right?
<[Dave] OLORIN> Have a *lot* of blank cards?
<[Dave] OLORIN> Yeah. One of the best books in my library.
<[Dave] OLORIN> Especially since Z-80 and 8080 were my first Assembly dialects.
<S.WHEELER> I went 6502 -> Z80 -> 8086, and am trying to move on to 68k
<[Kevin] APPERT> you can make kaptan flexible circuits using this technique...
<[Kevin] APPERT> and simulate multilayer with a g-10 base board, soldered 
completely, with kaptan layers on top.  it's a little time-intensive, but it
 works well.
<[Dave] OLORIN> I went Z-80 (briefly) -> 8080 (sigh) -> 8085 -> 68000 
(Registers! Whee!) -> 8086 (mainly inline). Almost all of my assembly now (very
little actually) is inline 8086 at home or inline 68k on the mini-super at work.
<[Dave] OLORIN> The comparisons on the 68k work opposite from what you'd expect.
<[Dave] OLORIN> (I can't remember which way they work without a reference...
every time I try they work the other way... :-) )
<[Wil] W.BADEN1> (a very familiar experience with me, too.)
<S.WHEELER> Not familiar to me, yet.  JAX gets to do all the 68k stuff so far.
<[Dave] OLORIN> The 68k is a nice chip. I'd like to get my hands on the new 
32532 system Dave Rand and George Scolaro have designed and do a Forth for
 that (I want to play with that architecture).
<[Dave] OLORIN> I've been playing around with the idea of a slightly different 
dictionary structure (to make the module system more efficient).
<S.WHEELER> I like the NS32000 architecture.
<[Kevin] APPERT> you fellows at Vesta might want to give HiTech a call.  It 
sounds like this is right up your alley.
<NMORGENSTERN> is here.
<[Dave] OLORIN> The standard fields (i.e. name, code pointer, parameter pointer,
 immediate or no, etc) would be seperate, and then the dictionary entry
 would be a two pointer node, one to the body of the entry, and one to the 
 next pointer.
<[Kevin] APPERT> < still likes LSI-11's
<[Dave] OLORIN> The one thing this *would* break is code which tries to go, say,
 from the CFA to the LFA. Won't work (there's the possibility of a MANY->ONE 
 binding to the name entry, so there isn't any unique link field to go 
 back to)
<[Dave] OLORIN> But it would make modules very efficient.
<[Dave] OLORIN> (i.e. 2 cell cost instead of an entire dictionary entry cost)
<[Mark] M.FOSTER3> Seems mighty quiet here .../sta
<[Mark] M.FOSTER3> Sorry folks.  I don't know how that happenned.
<[Gary] GARY-S> Mark - did you get my reply off the BBS yet ?
<[Wil] W.BADEN1> LOCAL and TO are in the "locals extension word set."
<[Mark] M.FOSTER3> Nope.  Still waiting....
<[Mark] M.FOSTER3> I've got a question I can't seem to get anyone to answer on 
the ForthNet.
<[Dave] OLORIN> Shoot.
<[Mark] M.FOSTER3> Thanks.  Can someone explain why FOR/NEXT ...
<[Mark] M.FOSTER3> is better than DO/LOOP?  ...
<[Mark] M.FOSTER3> Pygmy and cmFORTH use FOR/NEXT ...
<[Kevin] APPERT> 1) Chuck thinks so
<[Mark] M.FOSTER3> but other systems don't seem to. ga and thanks.
<[Dave] OLORIN> I don't use FOR/NEXT so I'm not the best person to ask... :-)
<NMORGENSTERN> As I understand it, Do/loop
<NMORGENSTERN> works better on some machines, and For/next on others.
<[Kevin] APPERT> 2) the FOR/NEXT is better adapted for some machines...
<[Kevin] APPERT> Novix has it in hardware.  LSI-11s and 1802's have suitible 
opcodes to build it.
<[Mark] M.FOSTER3> What's best on an 8088?
<NMORGENSTERN> Do/loop
<[Kevin] APPERT> whatever you like.  this is Forth
<[Wil] W.BADEN1>  DO/LOOP will never be as efficient as FOR/NEXT.
<[Wil] W.BADEN1> FOR/NEXT will never be as effective as DO/LOOP.
<NMORGENSTERN> The action is "when the difference between the index...
<NMORGENSTERN> and the limit passes 0"
<NMORGENSTERN> That is easily detected on 8088 machines.
<[Mark] M.FOSTER3> I'll have to edit my capture file to figger out ...
<[Wil] W.BADEN1> FOR/NEXT is a good idea ... 18 years too late.
<[Kevin] APPERT> have both, use whichever appropriate.
<[Mark] M.FOSTER3> damn!
<[Dave] OLORIN> If you don't like the rules...change them. (This attitude 
undoubtedly gives the standards team fits when dealing with everyones favori
te variations)
<[Mark] M.FOSTER3> what you gurus are saying ...
<NMORGENSTERN> It's easy to write a word to change one type of limits to the 
other
<DRMIKE> is here.
<NMORGENSTERN> : basic-style  (n1 n2 -- n1' n2')  2dup > if swap then 1+ swap ;
<NMORGENSTERN> usage is 3 5 basic-style
<NMORGENSTERN> The result will set up indices for a do .. loop
<[Kevin] APPERT> when you need an index which corresponds to what you are DOing,
 use DO, for example, in most cases     10 0 DO    might just as well
be a for-next, but  512 256 DO must be a DO
<NMORGENSTERN> that will go from 3 to 5
<[Mark] M.FOSTER3> Thank you!
<[Wil] W.BADEN1> 10 0 DO is a terrible choice for FOR/NEXT ...
<[Wil] W.BADEN1> Try 0 9 DO -1 +LOOP
<NMORGENSTERN> : bigtop 2dup > if swap then ;
<NMORGENSTERN> The basic-style becomes : basic-style bigtop 1+ swap ;
<[Kevin] APPERT> if you want to step thru an array in the forward direction, I 
will point to the item or offset, rather than having to do arithmitic
   or increment a stack item each pass  in a DO loop
<[Kevin] APPERT> oops, that was not too clear
<[Kevin] APPERT> DO loop = index you control      FOR/next = index you have to 
manipulate (or use as is)
<[Wil] W.BADEN1> The trouble there is double indexing.
<NMORGENSTERN> What do you mean, Wil?
<[Wil] W.BADEN1> Pointer goes up, counter goes down.
<NMORGENSTERN> But isn't the counter hidden?
<[Wil] W.BADEN1> DO/LOOP often allows you to go the way you want.
<[Wil] W.BADEN1> I agree with Dijkstra. DO/LOOP and FOR/NEXT are both too 
restrictive.
<NMORGENSTERN> As we discussed last Sunday, Forth allows you to change the index
<NMORGENSTERN> in a do/loop while it is running.
<[Kevin] APPERT> what does he suggest
<[Kevin] APPERT> Dijkstra?
<[Wil] W.BADEN1> They force me to think in a narrow way.
<[Wil] W.BADEN1> Edsger (sic).
<[Wil] W.BADEN1> The Paul Revere of the structured programming revolution.
<[Kevin] APPERT> I've heard of the gentleman.  What does he suggest as an 
alternative to DO of For/next
<[Wil] W.BADEN1> "Guarded iteration."
<[Kevin] APPERT> yes. author of "GOTO Considered Harmfull" and other weitghty 
missives
<NMORGENSTERN> More like the Mao, Wil. The result of his efforts was a great 
rigidity.
<[Kevin] APPERT> like BEGIN  /  UNTIL   Wil ?
<[Wil] W.BADEN1> Do we still have "humble programmer" in the library.
<[Gary] GARY-S> what is the title -
<[Kevin] APPERT> what library?  what "humble programmer" ?
<FIGCHAPTERS> is here.
<NMORGENSTERN> A humble Forth programmer is a contradiction in terms
<[Gary] GARY-S> I never saw a file named 'humble programmer' - what title was 
it filed under ?
<[Wil] W.BADEN1> Try looking for "Dijkstra".
<[Wil] W.BADEN1> It was put up in 1987.
<[Gary] GARY-S> Unless Dennis pulled it for inactivity, it is still there
<[Wil] W.BADEN1> Nov or Dec.
<[Gary] GARY-S> I'll kill the steno if I leave the RTC, so I'll have to look 
later
<[Wil] W.BADEN1> I'd go look, but I don't want to miss a moment of this 
occasion.
<[Gary] GARY-S> I'll just check later and report my findings Sunday or next 
Thursday
  edit note : The 'humble programmer' file Wil refers to is #802 in the 
              GEnie Forth RoundTable software library.
<[Gary] GARY-S> care to tear into TO now ?
<[Kevin] APPERT> is it HiTech you are looking at?  you are probably aware they 
are a long time user of (guess what programming language)
<[Wil] W.BADEN1> Local variables reduce Forth to the level of Tiny C, but not 
as understandable.
<[Mark] M.FOSTER3> Doesn't miniforth use TO a lot?
<[Gary] GARY-S> I hope not
<[Mark] M.FOSTER3> How come?
<[Gary] GARY-S> TO is at least bad choice of terms and not the best way to do 
things
<[Wil] W.BADEN1> Local veriables fail to address the real problem of local 
names.
<[Len] NMORGENSTERN> Disagree: TO can be very useful. I prefer IS, but that's 
because
<[Gary] GARY-S> Wil - did you ever get edu forth going again ?
<[Len] NMORGENSTERN> I am used to it.
<[Wil] W.BADEN1> Still chasing fat rabbits.
<[Len] NMORGENSTERN> Right Wil. See my presentation at Austin (not out yet! 
disappointment)
<[Gary] GARY-S> have you examined Frank Sergeants PYGMY as candidate
<[Len] NMORGENSTERN> as the first-searched
<[Len] NMORGENSTERN> What I did was to interject a temporary vocabulary
<[Len] NMORGENSTERN> as the first-searched
<[Wil] W.BADEN1> (Well, that was just what I was going to say.)
<[Len] NMORGENSTERN> Pascal has the right idea on what a local is!
<[Gary] GARY-S> Len - there has been interest in PASCL routines on ECFB, have 
you posted your Forth-day stuff yet ?
<[Len] NMORGENSTERN> No, I am presenting it at Asilomar, and
<[Len] NMORGENSTERN> I would prefer to wait until after that.
<[Gary] GARY-S> no prob - I was just amused at the timliness
<[Kevin] APPERT> Len, in Pascal everything goes away whenever you leave a 
module.  then you have to build it all up again when you re-enter.  what a t
ime waster!  I don't think they have the right idea at all
<[Len] NMORGENSTERN> No, only the local things go away. The global ones stay
<[Len] NMORGENSTERN> It may be a time-comsumer but not a time-waster!
<[Len] NMORGENSTERN> What is a time waster is to accidentally redfine a word, or
<[Kevin] APPERT> yes, but you were saying they had the right idea about what a 
local is.
<[Wil] W.BADEN1> Not even a time-wastr.
<[Len] NMORGENSTERN> have to think up a poor name because the good one is used.
<[Kevin] APPERT> where do you keep track of the context so you know when it 
means what, though, Len?
<[Len] NMORGENSTERN> Often, you simply define your locals as A B C, etc
<[Len] NMORGENSTERN> In a really local context, that's usually good enough.
<[Len] NMORGENSTERN> You drop them when you don't need them any longer
=== End of Steno notes. ===