X-Message-Number: 7336
Date: Sun, 22 Dec 96 18:10:15 UT
From: "Robert Ettinger" <>
Subject: A wild idea

MONEY FOR CRYONICS--a wild idea:
 
There are quite a few computer people on Cryonet, and there is a (barely 
conceivable) chance for one or more of them to make a lot of money.
 
We have all heard of the "Year 2000 crisis" in many older, specialized 
computer programs. It is said that billions of lines of code must be reviewed, 
revised, and tested in context, at a cost of HUNDREDS OF BILLIONS of dollars 
(because of the previous practice of saving space by using only two digits for
year dates).
 
To begin with, I question the premise--that this enormous price would still be 
cheaper than jettisoning the old programs and installing new ones, or 
installing new hardware AND software. But let's assume the premise is correct.
 
 Once again I am questioning my betters. There are many thousands of computer 
people, all smarter and much more experienced than I am, who apparently concur 
that there is no "silver bullet" that will solve such problems in a neat and 
cheap way.  Nevertheless, I simply find that hard to swallow. 
 
After all, a date is a date. There must be some way to identify dates in 
memory, and (perhaps with more difficulty) in program instructions, as 
distinguished from other data. Once you have done that, many ways
to make the conversions suggest themselves.  
 
A second approach--leaving the old programs intact--might be just to regard 
(say) all years from 01 to 09 as 2001-2009, rather than 1901-1909. Surely 
those century-ago references will seldom appear. 
 
A third approach (which could be combined with the second) might be just to 
add a "plausibility" or "consistency" test before final entry. This would 
merely be an add-on program, short and cheap (I think, speaking from 
ignorance). Whenever the possibility of a century anomaly exists, check the 
final entry
for plausibility and consistency in light of other data. Any stand-out numbers 
would be reviewed by hand. (This should be a standard practice anyway, and if 
it is, there is small cause for concern.)
 
A fourth approach would depend on the style of storage, whether binary, 
decimal, or whatever. If, in the old program,  characters (letters of the 
alphabet) can be represented in the same space as the date numbers, yet 
differentiated from date numbers, then for the next century substitute letters
for numbers, with a trivial add-on program to change back to regular dates for 
final entry.
 
A fifth idea depends on whether the date involved is the year only, or also a 
month or day and month. If the date includes the month, or day and month, then 
all entries earlier than the date of changeover, and referring (say) to 
numbers < 9,  could be identified and handled accordingly. This should not
be very expensive in computer time.
 
A sixth approach might be to categorize the data in some manner, so as to 
identify probable frequency and severity of impact of the problem. The least 
and most vulnerable categories might be handled separately.
 
A seventh feasible approach might be--in some cases--just to ask the customer 
or next user to check for errors, shifting the financial burden from the 
computer owner without any great hardship to the next user, who might if 
necessary be compensated for his trouble.
 
Probably some on this list will be able to tell me why those ideas can't work. 
When experts are unanimous, they are usually right. But the amount of money 
involved may warrant second thoughts.
 
Robert Ettinger


Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=7336