Roy Schestowitz wrote:
> __/ [ BearItAll ] on Saturday 02 September 2006 08:46 \__
>
>> Roy Schestowitz wrote:
>>
>>> Software Delay Said to Cost IRS $318 Million in Overpaid Refunds
>>>
>>> ,----[ Quote ]
>>> | The Internal Revenue Service gave away $318 million in improper
>>> | refunds this year because a computer program that screens tax returns
>>> | for fraud was not working, according to a report released yesterday.
>>> |
>>> | [...]
>>> |
>>> | The IRS had contracted with Computer Sciences Corp. to update the
>>> | program, but the contractor could not produce a working program by the
>>> | deadline. The old program could not be put back into operation in time
>>> | for the spring 2006 tax-filing deadline.
>>> `----
>>>
>>>
>>
>
http://www.washingtonpost.com/wp-dyn/content/article/2006/09/01/AR2006090101507.html?nav=rss_technology
>>
>> That contract needs a kick up the arse. Who would take on a critical job
>> that runs anually and run it live for the first time in place of the
>> previous version?
>>
>> You would run it beside the orriginal, testing and comparing samples. So
>> that in a worst case you have the data from the previous method to fall
>> back on.
>>
>> The agency has to take an equal part in the blame because they shouldn't
>> have left themselves in such a vulnerable possition. The problem there is
>> that some software houses sales staff tend to be a little exagerated in
>> their claims.
>>
>> I call an outside software house in to here now and then, the number of
>> times they have suggested routes to completion that would have put the
>> current system at risk. When I say to them that they must prove the test
>> systems on the mirror machine first, they always try to claim that
>> constitutes a second install when they put it onto the live system. I'm
>> never fooled by that, but I can imagine some company managers being
>> persuaded to allow them to work directly on the live systems.
>
> What caught my mind were two things:
>
> (1) The Internal Revenue Service is in the hands of a sole vendor who has
> access to the source code. If it were an Open Source project, there would
> be wider options. Think, for example, Red Hat versus Novell. The Internal
> Revenue Service cannot mend the code, either.
>
> (2) The deadlines and timeframe makes the IRS a hostage. It's very much
> like software houses that get clients paralysed owing licensing, formats
> that prevent migration (non-standard), and contracts. Traditional Open
> Source doesn't have these problems.
>
> Think could be worse for the IRS if its systems were (are?) vulnerable to
> attacks and the responsible software company 'shelved' the known bug
> rather than address it with a quick fix.
>
> CSS - company owns program, controls schedule.
>
> OSS - customer own program, can fix at will, third parties can help.
>
> Then come the issues of liability and control over destiny. Not every CSS
> house survives. And that's where disaster strikes... because of lockins
> and dependence.
>
> Best wishes,
>
> Roy
>
In this case I wouldn't really think that open/closed is the reason for
failure. The project failed because it was badly implemented. It doesn't
even matter that the software itself failed, the first live runs of any
software can fail however much pre-testing you did, but they has to always
be a safety net.
|
|
