On 11 Mar 99, at 9:54, James Kerr wrote:

> I was wondering, what are the people using to diagnose Y2K problems
> on peoples machines? I have software to test the hardware side but
> nothing to run to test all the software's, Norton 2000 looks pretty
> good and keeps getting updated all the time but are there others that
> do a better job?

  There are two basic approaches, as far as I can see:

1.  Try to locate .EXEs and .DLLs that call system date routines known
not to be compliant.  This can fail if the linkage is dynamic rather
than static (unlikely), but also can't detect if the application does
"windowing" -- effectively moving the critical date to 2040, 2050 or
2080.  Nor will it catch code that gets a compliant date, and proceeds
to use it in a non-compliant manner.
  So it will probably miss some problems, and give a misleading alert
on others.

2.  Try to match executables found on this computer against a database
of known good/known bad application versions.  The thing is that your
favourite shareware or game program might not be in the database (in
which case, approach 1 makes a reasonable fallback).
  Keeping the database up-to-date, and re-running the diagnosis as 2000
approaches, are critical to the success of this approach.

  Then there's the grey areas.  For instance, current versions of Excel
are *basically* compliant -- but theres nothing to stop a user from
using a 2-digit year display format, or throwing away the first two
digits of the year in a macro.

  Essentially, I don't think the technology exists to write a 100%
accurate tool for this.  A hybrid approach might get to somewhere
around 98% for most people, and I suspect that this is how Norton 2000
works.


David G

                         PCSOFT's List Owner's:
                      Bob Wright<[log in to unmask]>
                        Drew Dunn<[log in to unmask]>