Re: Reversing errors is trivial?

[Ed Gregory <edgregory -at- HOME -dot- NET>]

        It depends on the architecture of the system or systems involved.
        In some cases, an keystroke-by-keystroke undo is impossible. In many
others, impractical.  One of the jobs of the programmers is to get the
entire process to work as cleanly and quickly as possible. They will not
slow a major system to a crawl in order to accomodate sloppy data entry.
        OLTP systems do not routinely offer a grand "undo" function. Your bank's
ATM network is a perfect example of how this doesn't happen.    From recent
experience, I can tell you that:
        1.      When you enter the amount you want to withdraw from a network ATM and
hit the OKAY key, a message to your bank is generated instantly. It asks
that a hold be placed on that amount of money.
        2.      If the machine is out of cash or the network at your home bank is down
for some reason, there is no "undo" that says "don't place a hold on that
because I didn't give out any cash."  The hold message is created and on
its way BEFORE the cash is dispensed.
        3.      The same is true when a retail clerk swipes your card more than once.
Don't believe them when they say you will only be charged one time.
        4.      These "bogus" charges will fall off within a given period, usually 3 to
5 business days, if there is no followup confirmation message  that says
you actually got the cash or made the purchase.
        Apparently, somebody decided at some point that the overall system was
better served with this delayed error trap than building in an instant
"undo" function.

        However, the bottom line here is that as a tech writer, your job is to
explain how the system works in the best and clearest terms possible. Can
you imagine, in a help file under the topic "undo", seeing an entry like:
"Sorry. The programmers were too lazy to build in anything useful like
that. Type more carefully."
        Unless you have the kind of access that makes you part of the development
process, leave the kibbitzing for software reviewers.  Write about your
client's products as they are, not as you wish they would be.






At 02:46 AM 12/24/98 -0700, Geoff Hart wrote:
>Charlotte Hybschmann Jacobsen writes <<Another one of Cooper's
>statements are that every action a user take should be reversible,
>that is no matter what you do, you can undo it again. My developers
>agree but they also tell me that this often takes a lot of
>programming effort to effectuate.>>
>
>To effectuate? <g> Actually, in the grand hierarchy of programming
>tasks, implementing an "undo" function is the next best thing to a
>trivial task. All you need to do is save the entire file to disk
>after each keystroke. If you don't like what just happened, reload
>the file from the previous copy upon pressing cmd-Z. In fact, if
>operating systems routinely did this sort of thing, losing data to a
>crash would be rare enough to be a remarkable event. Online
>transaction processing systems (OLTPs) do something similar, at least
>in intent. Now before anyone takes me to task on this statement, let
>me emphasize that there are unquestionably more elegant ways to do
>this. But "a lot of programming effort"? Not by any stretch of the
>imagination.
>--Geoff Hart @8^{)}
>geoff-h -at- mtl -dot- feric -dot- ca
>
>"Patience comes to those who wait."--Anon.
>
>From ??? -at- ??? Sun Jan 00 00:00:00 0000==
>
Ed Gregory
http://www.members.home.net/edgregory/search1.htm

 From ??? -at- ??? Sun Jan 00 00:00:00 0000=