Slashdot Mirror
FBI's New Info-Sharing Software Project Fails
Spy Handler writes "After 4 years and half a billion dollars, FBI's attempt to create new information sharing software - called Virtual Case File - simply didn't work.
← Back to Stories (view on slashdot.org)
Since the attacks, Congress has given the FBI a blank check, allocating billions of dollars in additional funding.
And that blank cheque is the problem. Whatever happened to accountability? It's the tax payers money to begin with.
Free XBox, PS2
Having worked at Andersen Consulting little more than a decade ago and seeing the dismal IT failures of EDS has had in England, when I here of vast amounts of money wasted on failed IT projects these companies immediately come to mind.
Perhaps you should try this?
:)
Hm, more seriously.. They must really have tried to make something special. Otherwise WebDAV+SSL would have proven to be a bit cheaper..
Probable impossibilities are to be preferred to improbable possibilities.
Aristotele
From a position of a tax payer this frustrates me.
However, as a programmer I can understand them wanting to scrap the program. If the design has been shot to hell, if their using technology several years past its prime, it's time to start fresh.
And as a tax payer, I'd prefer the FBI to use a system that works, rather than a system that doesn't.
-Teiresias
When I worked for the Department of Justice, a case might have 5 different case numbers: one case number for the DOJ, one case number for the FBI, one case number for the Defense Criminal Investigative Service, one case number for the Naval Criminal Investigative Service, etc. If I only had the DOJ jacket number, it could take me 15 minutes to get the case number for another agency, just so I could talk to one of the investigating agents.
Spend money to fix that larger-scale problem, before flooding the FBI with money to squander on a software application that they will be terminating and starting afresh on.
The bureau is no longer saying when the project, originally scheduled for completion by the end of 2003, might be finished. ...
...
...
A prototype of the Virtual Case File was delivered to the FBI last month by Science Applications International Corp. of San Diego. But bureau officials consider it inadequate and already outdated, and are using it mainly on a trial basis to glean information from users that will be incorporated in a new design.
Science Applications has received about $170 million from the FBI for its work on the project.
A spokesman for Science Applications, Ron ollars, said via e-mail that the company had "successfully completed" delivery of the initial version of the Virtual Case File software last month.
The stripped-down prototype will be running for three months. The bureau plans to then "shut it down, take all the lessons learned and incorporate them in a future case management system," a person familiar with the bureau's plans said.
An outside computer analyst who has studied the FBI's technology efforts said the agency's problem is that its officials thought they could get it right the first time. "That never happens with anybody," he said.
Some sources sympathetic to the FBI defended the process, and said that what has been learned in designing the software has given the bureau
valuable design and user information.
The first time they saw the software was a year after the delivery date. So they must have been using waterfall. Then they defend the process by saying the only good thing they got out of it was the information for the next pass of iterative development. So the best thing about waterfall is that when it fails you can turn it into iterative. Pure genius.
Stop-Prism.org: Opt Out of Surveillance
Why don't they do the same for software?
These are the same feds who treat copyright infringement as "theft"; who tack on all sorts of costs to the cost associated with a breakin (where a kid just pokes around the system); and yet they turn the other cheek when these companies waste billions of dollars on badly-executed projects.
As a taxpayer, I am thoroughly pissed at this waste of my money.
Expect the Prime to pay a token couple of million dollars as a "fine" and walk laughing all the way to the bank...
i think you fail to see how a gov project works.
steps to gov project.
1. create team of 4-5 people to outline requirements
2. get team staff of 2-3 to support them
3. get place for them to work for up to 2-3 years developing requirements
4. subcontract out requirements analysis to someone
5. hire subcontractor to verify requirements analysis by the first.
6. hire gov people to oversee both contractors
7. hire people to support people overseeing subs.
8. release requirements out for public "auction"
9. review company responses to "auction" by s team of 12-20 people
10. hire crew to support 12-20 people on the responses review.
11. except bid.
12 start project.
there you have approximately 100-200 people working on just ensuring the requirements documentation and bid for 3-4 years before a project even starts. that alone could cost 200 mil or more.
they got off cheap.
In theory there's a guy involved in the process who reads the bids and rejects those that seem infeasible.
In practice this guy is a manager, not a software expert, and he's usually an idiot. I've written dozens of proposals and it's monumentally clear that your job is to impress this idiot. Coming up with an intelligent design is something you spend time on after the bid, not before. And there's usually not time then, because you're busy fulfilling this idiot's pipe dreams.
Not that I'm bitter or anything.
One of the hardest things in working for government is that in order to write software properly, you need to get a good look at the data you're working with. You can't see this data; it's heavily, heavily classified.
It's classified two ways: first, a lot of this data is privacy protected (the FBI spies on American citizens and that data is heavily controlled). Second, one of the things it needs to store is sources&methods, which are protected even more closely than the data itself. (The most classified stuff is always about sources&methods, not the data itself.)
The open-source community could write pieces of it, but the hard work on a project like this is adapting it to the particular requirements of the customer.
The problems involved aren't abstract ones that can be solved byu an incredibly clever person like Bram Cohen. They're involved in getting a gazillion people to all buy off on a data format, and convincing them that they really can share information without violating their security requirements (which is really just code-speak for "if I let you have this information I won't be the only one with it, and therefore I become less important.")
The security clearance requirement means that they're working with a drastically reduced pool of programmers. Corners get cut, ideas go unused for lack of implementers, internal oversight is practically nil. (They have code reviews but they're an immense waste of time.)
I'm not sure I've ever worked on a government project of even a tenth this size that I considered to be successful, even if it did get deployed. But throwing it out to the open-source community isn't an option.
Imagine the learning curve! Here we have this monolythic blob of stuff that can do EVERYTHING. Here are the 12 billion source lines of code and a make file that takes 3 centuries to run. The project is 4 years behind schedule and we fired the 10,000 coders that were working on it before, but we asked them to comment their code with nice flower boxes.
From my experience, the bigger the project the more likely it is to fail. Making lots of little bits out of one big one may result in some integration hiccups, but at least there will be useful pieces and refactoring can be addressed on a priority basis.
When the people fear their government, there is tyranny; when the government fears the people, there is liberty.