Re: ODBC (was RE: Green screen - it's time is over )

["Brad Jensen" <brad@xxxxxxxxxxx>]

> I said message-based client/server,

Well, somtimes you did, and sometimes you didn't.

>  not "server based".  And for that
> architecture, my statement is absolutely true.

No it isn't and you seem to be ignoring the fact that it isn't.

>  Let's look at the situation:
> in a message-based client/server architecture, the client
program (such as
> an inquiry or a print report) requests data from a server.  The
client fills
> in some fields in a data structure, sends it to a server, then
receives data
> one message at a time back from the server.  In the simplest
case, each
> message corresponds to a record in the file being queried.
>
> Prior to Y2K, the dates in those messages would have been
six-digit dates.
> And that would have matched the way they were stored in the
file.  Now for
> Y2K, we expand the dates in the file, but we DO NOT EXPAND the
dates in the
> message.  When the server returns the message to the client, it
simply moves
> in only the YYMMDD portion of the date.
>
> So let's take a print program as an example.  Most print
programs did no
> date calculations;

Bingo. 'most' not 'all'. The point is that cleint programs, even
in a client server environment, even in a message-based server
environment, may have dependencies on data structure even if they
don't use ODBC.

And the only way to know is to examine them closely one by one.
The only way you will know for sure is if you find the dependency,
or in the case where you do not find it, if the program breaks AND
you catch it.

Meanwhile the extra layer of logic adds to the complexity of
debugging. It does not reduce it.

> they simply read data in a specific order and printed it
> out.  If they had used a server program to retrieve the data,
the server
> would have returned the data in the correct order.  Even though
the data on
> the record was eigth digits, the dates in the messages returned
to the
> client would have had six digits.  The programs would have done
their
> YYMMDD->MMDDYY comversion, and printed them with the appropriate
separators,
> and the reports would have looked perfectly normal.
>
> Without a single line of code in the report program changing.
>
> Does this make sense?

It makes sense as a hypothetical case of something that agrees
with your assertion, but only as a trivial example that doesn't
really test your hypothesis.