perm filename RUSS.1[AM,DBL] blob sn#462858 filedate 1979-08-08 generic text, type T, neo UTF8
 -- Messages from file: [SUMEX-AIM]<LENAT>MESSAGE.TXT;1
		 -- Wednesday, August 8, 1979 09:15:04 --

   65   2 Aug  Stefik                Misc
   66  03 Aug  Russell Greiner <RDG  What I've been up to    
   69   4 Aug  Stefik                Units fixes


65 -- ************************
Date:  2 Aug 1979 1543-PDT
From: Stefik
Subject: Misc
To:   lenat

Doug,
	Russ has been having a lot of problems with the Units stuff
on SCORE.  Some of the probs are of his own creation (e.g.,
not knowing the effects of certain INTERLISP actions like
START to the TOPS20 exec, trying to use the tranfer command from a
KB to itself (bug), not setting the generation counts appropriately,
...).  I can't really tell whether the final probs are caused by
his doings or by Unit bugs modulo TOPS20.  I think it would be
worth building a version of Units on SCORE afresh (ftping from
SUMEX), building a version of COMPACT on SCORE, and exercising the
hell out of it to try to uncover the problems.  I don't hve
time to do this now;  we've got to get Dave to do it or get
Russ educated to do it; or do it ourselves later.  The # of
probs he is having is truly extraordinary (I'm getting about
4 panic phone calls per day.  Something is definitely screwed up
 but I don't understand how Brutlag & Friedland have been faring
-- unless they've been only using SUMEX).  Mak

PS.  If you send a note to ADMIN.MRC at SCORE, he can set your mail box
so that mail gets forwarded to SUMEX automatically.  
-------


66 -- ************************
Mail from SU-AI rcvd at 3-Aug-79 0004-PDT
Date: 03 Aug 1979 0003-PDT
From: Russell Greiner <RDG at SU-AI>
Subject: What I've been up to    
To:   DBL at SU-AI

Doug:
	Well, I seem to be taking two steps backward for every one forward.
UNITS and I are still having our little "arguments", and I keep losing.
Just when I thought I knew the cause of all such misunderstandings, and
had resolved how to avoid them, I encountered another one (when I asked it
to add a new slot to one particular unit, UNITS decided to create a slew
of new units - and merrily entered an infinite loop).
Mark has been fairly available via phone; but three calls, on three distinct
(and apparently unsolvable) problems, in two days, has worn his patience a bit
thin.

At least until Dave Smith assures me the version of UNITS on SCORE has learned
its manners, I would recommend using the Eurisko DB stored on SUMEX.
See <CSD.GREINER>EURISKO for the SCORE data base (with BACKUP.* as backup),
and <GREINER>EURISKO on SUMEX, if I get a chance to FTP them over. (SUMEX
has been down, as this is Thursday.) Also, your WORK2.COM file should be over 
there as well. (If not, feel free to FTP them yourself.)
	While on that subject, I think I'll (one day) make the necesary modifications
to the full Unit Editor to only make an UPPER-Case/lower-Case distinction
for UNIT & SLOT names, when GLOBAL-NOSHIFTFLG is T. Based on recent experience,
most of one's time is spent typing commands, which have made the assumption
their input will be upper case. This change will not be major, I beleive.
Eurisko has been dutifully entered in both cases, so you should type
LISP <then> LOAD[<CSD.GREINER>WORK2.COM] <then> OK,
to access and/or expand the Knowledge Base.

	I hope this holds a good framework on which to base the full Eurisko
system. I certainly am getting a bit tired of retyping it. I do recommend
SAving often, and backing up copies at frequent intervals. The SUmmary command
(starting at EuriskoObjects, for all groups (ie No particular groups) ) is
a convenient way to store a version of the units so far.
(I believe <CSD.GREINER>EURISKO.SUMMARY is up-to-date.)
Also, it would not hurt to use the RECORDing facility, just incase of trouble.

	It would also be advisable to 
SET DIR GENERATION-RETENTION-... 3
(as well as SET FILE GENERATION-RETENTION-... *.* 3, for files created earlier)
to allow UNITS to delete most recent copies of files, which it enjoys doing.
	Also, I did finally talk with Peter, who both OKed the changes I proposed
to his Rule Input, and offered his continuing help towards testing such
updates.

	Hope you enjoyed LA.
Russ

PS One change which maybe should be made: Have the Relations slot of each
descendant of EuriskoObjects have a demon, which creates  a new instance
of RelationSet for this slot, (if needbe, or updates the one already there)
which points to the newly created UnitRelation (you'll have to create this)
indicating the new spec/genl/inst/proto relation in which this new unit is 
involved. (Ie this is bookkeepping  which will parallel what UNITs is doing.)
	By the way, the UnitName slot in Descriptor will be redundant with
this - just an economizing move.




69 -- ************************
Date:  4 Aug 1979 0843-PDT
From: Stefik
Subject: Units fixes
To:   lenat, greiner

Date:  3 Aug 1979 1726-PDT
From: De2Smith
Subject: RUSS'S UNITS PROBLEMS.
To:   STEFIK


Russ's problems with losing files and having trouble saving, should, I hope
be resolved by the different protocal for UF.  For one thing, if there is
not enough space for a second UNITS file (paging file) UF will not be able
to open the .PAGE file after trying to copy the .UNITS file to it.  Thus
it should never strangle itself by using the old version of the UNITS file
for a paging file.  Also, this way if you run out of room in trying to rewrite
the .rel* file or fnsfile you can hopefully save your work on another directory
which you might have access to.  The UF stuff has a few more bugs in it, but
should be done in another week or so.  I am working on modifying UE to be
compatible with it.  This shouldn't take long.  I have some test runs that
show what can be done with the new protocol.  Perhaps we can meet on Monday
and I can show them to you.

Dave.
-------
-------