The Y2K Cinderella Spread Sheet Torture Tests

Let me show you a worked example. I think it will also demonstrate why so few products have gone through my test mill at this time.

There has recently been a lot of shouting about Lotus and spreadsheets in general. The Bees have been disturbed in the Hive, and there is a Noisy Buzzing.

Now when I hear about this sort of thing my antennae start to twitch. It sounds "wrong". It does not gell with my Internal Reference Library. I vaguely remember dates were messy but not Terminal.

What about these problems with Lotus?.

I am not yet convinced there are actually any problems.

When Don Bricklin and Bob Frankston first produced Visicalc way back in 1979, I don't think they realised just what they were starting. Software Arts became Visicorp in 1981 and Visicalc got a fancy new manual with real photographs of the screen and DIF exportability. But it didn't have date support.

So between 1981 and 1986 and the Lotus takeover, spreadsheets acquired dates as an afterthought, and have remained more or less in that state ever since.

They say that the sincerest form of flattery is imitation. Bricklin et al must be feeling very flattered. Every man and his dog seems to have produced some form of spreadsheet or other, carefully replicating the original features (and flaws).

The earliest reference to Dates that I can find is in a 1988 Quattro manual.

And they knew about 4 digit solutions for Year 2000 even way back then.

What, we ask ourselves, is the present position?.

Lotus have produced a somewhat nebulous document at: URL: http://www.lotus.com/UK/220e.htm

But still the argument and confusion continues.

Fred Trigg produced a useful summary and guide (see Cinderella Topics)

So, in the face of this confusion, we must resort to the empirical method, and do some tests.

The concept begins to emerge. "Test spreadsheets for Year 2000 date support".

I generally just think about it all for a couple of days and jot down notes to myself in my System Log. Eventually I get round to doing something about it.

1) First I try to accumulate general references and references that I have seen on the Web, in one place. This is generally in my Eudora mailer.

2) Next I dig up all my old software relative to the topic. I do searches on my own index of freeware and shareware sources and catalog of available stuff. I am currently building Usenet archives on various things so I would scan these as well. This process quite often highlights related links to "sister" products not obvious at first glance.

This will also provide me with a "where to find" and "what to acquire/download" list.

I try to create extracts, summaries and definitions of terms, so that I dont have to scratch around in manuals while testing.

Bibliography and References for Spreadsheets

3) By this time I have a pretty good rough idea of the nature of the problem regarding size and method of attack.

4) The next phase is to design a test case or Test Strategy.

A Primary assumption with Cinderella Tests is that they have to easily performed by a Novice user at home. So KISS (Keep it Simple Stupid) is the rule.

Depending on the program under test, I might now draw up a schedule or procedure of what steps to take.

In this case I think a sample spreadsheet with accompanying Notes and instructions will do the job.

5) So then I find my old software, dislodge the Ancestral Spiders, dust it off and reinstall it.

(This reminds me of the origins of Grace Hoppers term "debugging")

Then I do some preliminary tests. This can best be described as "Fooling Around". The rough test is called "DATETEST.WK1"

6) At this point I have to consider the Cinderella population, and I usually need to map this against what products I have immediately available to me.

This might look like this:

Y2K Cinderella Spreadsheet Test:

                Lotus 2.0,     Lotus 5
                 Lotus 2.1        Quattro Pro
                      Quattro       Excel 5
DOS 3.2    XT      i  i  i
DOS 3.3            N
Dos 4.0            N
DOS 5.0            i

DOS 6.20  486      o  o  i
DOS 6.22
-------------------------------------------------
WIN 3.1            o  o  N       o   o   o
OS/2 Warp 3        o             o   o   o
OS/2 Warp 4        d             d   d   d

WIN 95                           d   d   d

WIN NT                           d   d   d

Key:  o - Online (installed)
      i - to install
      d - defer till later
      N - ignore
This at least gives me a guideline as to what I can do now or can do with minimal installation.

The results of the testing are added later in the form of the Cinderella Acceptability Index awarded to the product for compliance.

As you can see the list of things to do starts to increase quite fast.

In theory, the linkages between the items in the matrix should also be tested. But the problem has to be resolved this side of the time the Sun goes Supernova. So we do a few, not all.


Okay, we have defined our terms and collected our references. Now it is time to fool around with the products.

The Y2k Cinderella Torture Test for Lotus 1-2-3 and other Spreadsheets

Lotus 1-2-3 Version 2.0

One of the tricks is to try and find the common factor, the 'crux' or 'missing link' that points to a solution.

Most often you have to go right back to basic assumptions.

With Lotus 1-2-3 the key issue is that the dates are stored as numbers. This is the crux.

Now a great deal of confusion arises with the early versions of Lotus in that they do not differentiate between numbers and dates at the cell level. You have an all or nothing "General" or "mixed numbers and text" or Global Date "All numbers are displayed as dates" which is not particularly useful in the long run.

Not everything SHOULD be displayed in the spreadsheet. The working results can conveniently be hidden "off page" so that you have a "clean" display area and a "don't care" or 'dirty" display area.

This is what happens in the next test series 'TDATE1.WK1'.


TDATE1-The Generalised Spreadsheet test

Borland Quattro Pro Version 4 for Dos

The Windows Versions

Note: The prerequisite assumption for Windows is that using "Control Panel", The "International" setting has been made for YMD, Century year, leading zero day, leading zero month. I am now adding a new one, separator "-", as a result of the spreadsheet exercise.

Lotus Release 5.0 under Windows

Novell Quattro Pro 6.1 under Windows

MS Excel 5.0 under Windows

At this point everybody is weary of testing, me particularly, so we will stop.

However, I have now learnt two things.

1. Most spreadsheets will operate into the next century.

2. Conversion between spreadsheets is a minefield of traps for the unwary. One day the Users are going to get very tired of Vendors playing the "if you won't use my format I won't use yours" game.


The Next phase is to update the Lists of Acceptable Products and eventually publish them on the Web

Here are some samples:

MS Dos 3.2

MS Dos 6.2x

MS Windows 3.1


NOTE: DOWNLOADS

The spreadsheet tests are available as "TDATE.ZIP" via Anonymous FTP from ftp://www.cinderella.co.za/pub

or from the FTP link at http://www.cinderella.co.za


Usual Disclaimers Chris Anderson 1997, February 16


Sponsored in part by

Try Me?