Paleo Retiree writes:
A hyper-ambitious new social-welfare program is in a seemingly hopeless tangle? O, who could have predicted such a thing! From Slate:
If you contract something out and get 500 million lines of code back, there’s no way it’s going to work correctly.
Blowhard, Esq.
Callowman
Sherbrooke
Eddie Pensier
Enzo Nakamura
Epaminondas
Fabrizio del Wrongo
Fenster
Brundle Guy
Glynn Marshes
Paleo Retiree
Atypical Neurotic
Question Lady
Sax von Stroheim
Sir Barken Hyena
The Mistaken
Uncouth Reflections 
But why in the world would it take half a billion lines of code in the first place?
As I understand it, this is what the website does:
1. Basic secure username/password account creation and login:
2. Have individual fill out information in a javascript webform. Some flow-charting to minimize the amount of time and info required.
3. Put it all in a databse
4. Calculate Legal Status (this isn’t as hard as it sounds, it’s mostly about household income)
5. Communicate with other networks to a. verify information, and b. provide quotes
6. Present Options
7. Web purchase of a policy.
Yes, I know the details are complicated. But they are not half a billion lines complicated. Try buying any other kind of insurance online and it’s essentially the same setup. And it’s been going on with orders of magnitude less coding for over a decade.
I suspect the 500-million-lines number is made up. 1,000 coders at 1,000 lines a day (net output after corrections and refinement) for 18 months? If it’s not, why so much?
LikeLike
Amazon.com does more business in a day. They should let Jeff Bezos handle this.
LikeLike
When building a system, the part that takes the longest is integration with other systems. It looks like the healthcare sites talk to a number of different government sites in order to verify identity and income, blood type, fantasy football standings, etc. I imagine that the interfaces between these sites were not clearly defined or tested. Multiply this by the number of sites involved and you have a predictable disaster.
LikeLike