Number: 2032

Date:  1-Sep-84 15':59':22

Submitter: Masinter

Source: Denber.wbst, LispCore^

Subject: want CLISPDEC(MIXED) default in standard loadup

Assigned To: 

Attn: Masinter, Release

Status: Open

In/By: 

Problem Type: Design - UI

Impact: 

Difficulty: Moderate

Frequency: 

Priority: 

System: Language Support

Subsystem: Other

Machine: 

Disk: 

Lisp Version: 31-Aug-84 16':17':57

Source Files: 

Microcode Version: 5124

Memory Size: 4096

File Server: 

Server Software Version: 

Disposition: '
["Masinter" " 2-Sep-84 15':23':30" Attn': Description':]

Description: RELEASE MSG': NOTE THAT THIS CHANGE WILL BE MADE IN THE FUTURE (also ZEROP change)'
'
The job here is not to set the default (which is simple) but to verify that nothing in the system breaks when we reset it. '
'
'
Date':  6 AUG 84 22':32 PDT'
From': MASINTER.PA'
Subject': proposal for incompatible default settings ....'
To':   LispCore^'
'
I propose making two changes in default system settings before Harmony':'
'
a) change DWIM to default (CLISPDEC ''MIXED) rather than (CLISPDEC ''INTEGER). That would mean that'
'
(A+B) would turn into (PLUS A B) rather than (IPLUS A B), and (add X 3) would use PLUS, and (for I from 1 to 10 sum N) would use PLUS and GREATERP for the iteration, and PLUS for accumulating the sum.'
'
I believe that all of our microcodes including 4K DLion now implement PLUS for integers exactly as efficiently as IPLUS. I think IPLUS is anachronistic and has resulted in a number of bug reports and obscure confusions.'
'
There *is* a possibility that some odd piece of the system will stop working with this change, but I sort of doubt it.... the only difference is in the case where you do'
'
(for I from X to Y ...) and Y/X are floating point arguments.'
'
There is also a possibility that some odd case of PLUS/TIMES etc *doesn''t* work the same or as fast, and that something will suddenly become dreadfully slower, but I also doubt it.'
'
Every other modern lisp defaults to generic arithmetic in lieu of other declarations. If you *SAY* explicitly IPLUS, you should of course get IPLUS, but if you don''t say or say implicitly, this would change the default.'
'
'
. . .'
Date':  7 Aug 84 18':06 PDT'
From': Kaplan.pa'
Subject': Re': proposal for incompatible default settings ....'
In-reply-to': MASINTER.PA''s message of 6 AUG 84 22':32 PDT'
To': MASINTER.PA'
cc': LispCore^.PA'
'
Although I think the changes that Larry is proposing are desirable things to do, I don''t think this is the time to fool around with pervasive system declarations.  This is the sort of thing that should be done not just before a release but just after, when there is a long time for all the hidden bugs to surface.  There is bound to be a long exponential tail on this.'
'
I think the harmony release message should warn users that we intend to do this on the following release.'
'
Date': 13 Aug 84 12':44 PDT'
From': JonL.pa'
Subject': Re': proposal for incompatible default settings ....'
In-reply-to': MASINTER.PA''s message of 6 AUG 84 22':32 PDT'
To': MASINTER.PA'
cc': LispCore^.PA, 1100Support.pasa'
'
The proposal to change the Interlisp default declaration from (CLISPDEC ''INTEGER) to (CLISPDEC ''MIXED) is certainly overdue.  Since this will be an incompatible change, I suggest that we resurrect the "announcement" letter about the pending ZEROP change, and extend it to include this one;  thus we can warn the user community some months in advance of the delivery of sysouts with these changes installed.'
'
'
-- JonL --'
'
'


Workaround: 

Test Case: 

Edit-By: Masinter

Edit-Date:  2-Sep-84 15':23':31