7 Person Year Application Rewritten in 2 Person Months | Hacker News Reader

7 Person Year Application Rewritten in 2 Person Months

metromodemedia.com

33 pointsstartupcrazy43 comments

A lone programmer rewrites an entire project in 2 months. It uses executable specifications. It is thrown away because nobody else understands it. (LISP, YACC. or something else?)

Open articleView on HN

He's talking about SCADA (http://en.wikipedia.org/wiki/SCADA)

Sounds really impressive. If anyone has ever worked with PLCs you know what a nightmare it is to interface with them and deal with their primitve data types. Not to mention programming them in ladder logic (yeah, most industrial machines are still programmed in ladder logic straight from the 1800s.)

So he's not kidding, this is a BIG deal.

I hate, no I HATE that language...

I thought for sure this would mention another "innovation" that came out of Chrysler, "Extreme Programming":

http://c2.com/cgi/wiki?WasChryslerComprehensiveCompensationS...

Short story: Chrylser had been trying to unify its various systems and subsystems for payroll and had failed and failed. Eventually, a group of consultants were brought in who had invented a particular agile technique, extreme-programming. Of course, the project failed, but that didn't keep the consultants from spreading the XP word throughout the world.

Two of my developers were on the original XP team at Chrysler. They don't to XP at Chrysler anymore either.

This sounds familiar. I went through something like this at a Fortune-100 company which I won't name here, except they rejected the solution because it wasn't written in .Net and budgeted around seven figures to rewrite their existing horrible system in .Net. I quit somewhere along the way but I still feel the frustration, pain, and perhaps even anger whenever I think of the stupidity and waste. I hope I never have to go through that again.

Executable specifications. The new compiler was about 5000 lines of Java. The developer says he can probably reduce most of 3000 lines of it to about 700 productions in YACC. Of course, nobody is interested!

Uhh...where does the $50 billion come from again?

Tracking down-time means knowing where your bottlenecks are. If you know your bottlenecks are you can improve cycle time. Improving cycle time by just a few seconds mean massive increases in output.

$50 billion might be a stretch, but it's not far off.

Note: Cycle time is the amount of time it takes to produce a part. It's sort of like the MHz of the manufacturing world. Usually car plants are setup such that there are n manufacturing lines in operation at any given time producing/assembling parts. So you're always at the mercy of the slowest line. It takes months to comission a new line so it's not just a matter of "adding servers" to improve production.

Thanks. I've read The Goal.

My point is that Chrysler itself was recently sold for less than $10 billion. Maybe Cerberus bought because they knew about this code that was written in a day and now they are going to turn them into a $50 billion dollar software company?

Look out SAP & Oracle...

Ok, well then your question is even easier to answer if you know that data tracking is serious business in manufacturing.

From the article: "Did you know that there is an IT breakthrough hidden inside of Chrysler that in the right hands could be worth 50 billion dollars or more?"

In other words all he's saying is exactly what I replied to your post: that data collection and tracking (not just for Crystler but as a standalone product) is worth a lot of money. He's not saying that Chrysler could or would sell this - in fact that's kind of the point of the post.

I see your point, but the title of the article is "Chrysler, Want To Make an Extra 50 Billion?"

My point is that if it was written by some dude in a day, it's not going to make anyone $50 billion, especially Chrysler.

Well, the project took 1 month by two guys, so we can assume it was non-trivial.

As a personal note, I've got a couple of years in manufacturing, with much of that in automotive. The amount of stupidity that goes on in that industry is shocking. It doesn't surprise me at all that a couple of smart guys were able to build something like this. It surprises me that Chrysler was actually able to get a couple of smart guys to work for them.

yes...that's probably the most confusing part of the article. :)

Sounds like a prisoner's dilemma kind of problem... individual firms won't benefit from open sourcing unless everyone open sources.

Not at all. Open sourcing your software not only means you share the benefits, but also the costs:

http://web.mit.edu/wwwdev/cgiemail/buybuild.html

Indeed. All of the Detroit automobile companies are now off shoring to India and China to try to reduce costs. Open sourcing factor floor control and monitoring systems is simply another way to achieve the same results.

Anyone have any idea what this consultant might have been doing to get this kind of productivity?

As far as I can tell from the article: Magic.

Or, maybe he called it "agile", but I'm not really thinking gymnasts and ballerinas are magically great developers, though they are strikingly agile. This article needs some kind of detail--it doesn't answer most of the W's of good journalism: Who, what, when, where, why, and how. We've only got a really vague what ("a good developer built something that I liked a lot") and a vague where (Chrysler). Somehow all of the good bits are completely left out.

Yes. You have run into the nature of a non-disclosure agreement the big three have everyone sign. You cannot tell the good parts.

But, that means you don't have a story. You strung me along expecting a payoff, and there was none. It's more than a bit rude. ;-)

I told you too much already. :-) Several clever grammars were written. If you want any more, you have to buy Chrysler.

I'm afraid Chrysler poisoned me as a customer before I was even old enough to drive. My folks owned a couple of Chrysler products that were the worst vehicles I've ever dealt with: my mom's Dodge van once stopped steering while she was driving...yeah, the steering wheel just spun right around. On another occasion the brakes simply stopped working, pedal to the floor and nothing happened. Those are seriously dangerous catastrophic failures in dangerous circumstances and without warning. The engine also had to be entirely replaced within weeks of the brand new vans purchase, but at least no one was in danger in that case.

I got to work on them quite a bit growing up--they at least gave me lots of experience fixing things. But it'll take another generation before folks who experienced Chrysler in the 80's and 90's, like me, start buying Chrysler products again...and then only if they are actually making good cars now (which I'm not confident of). I'm afraid I've been well-trained to buy only Japanese cars. I've had two fantastic Nissans in the 20 years I've been driving, and one mediocre (but never broke in dangerous ways or left me stranded--it just had some annoying quirks) Toyota...I like to drive my cars into the ground, and the ground comes at me way too fast with American cars, and Chrysler cars in particular.

Sorry, I know you weren't looking for a rip on Chrysler, but suggesting I buy a Chrysler product sent me reeling into the past for a moment...I could smell the burning oil, feel the claustrophobia of laying on concrete under a jacked up minivan, and everything. I just thought I'd share my pain with you. ;-)

I wasn't suggesting you buy a Chrysler product. I was suggesting you buy Chrysler. You get it all, along with 5000 lines of very clever Java code that they probably couldn't even find if they wanted to.

Heheheh! Much better suggestion. Maybe I could make Chrysler profitable. I've got forty-one crisp American dollars in my wallet. Who do I talk to? ;-)

Cerberus. They just bought it though; wait a couple years and they may reconsider.

BTW, those of us who follow the financial news are laughing because Daimler (the previous owner) literally paid Cerberus to take Chrysler off its hands. The purchase price was nominally $7.4B, but it was structured as an investment into Chrysler itself. In other words, Cerberus "bought" Chrysler by putting cash into a business it now owns, effectively paying itself. In addition to that, Daimler paid Cerberus $650M in cash. Yes, that's the seller giving money to the buyer.

The reason for this is that Chrysler is currently worth less than nothing. Daimler's earnings are up $1.5B since getting rid of it, and Cerberus took on Chrysler's $18B of pension obligations. Cerberus probably figures that it can renegotiate with the UAW and default on Chrysler's pension plan, which'll help return it to profitability. Daimler couldn't do this because its profitable Mercedes division was effectively subsidizing workers in Detroit, costing it bargaining power in any attempt to stiff the unions.

Chrysler is currently worth less than nothing, but Chrysler Financial has real value. Basically, they acquired the unfunded liability of Chrysler as the price of Chrysler Financial. Chrysler Financial may be work 18B of pension obligations to Cerberus, especially if they get synergy with GMAC.

Put your money away. As I understand it the last people to "buy" Chrysler got paid to haul it away.

Edit: curses, beaten by 2 minutes. And the other post brings the details, too.

I feel your pain. My parents owned a dodge Omni hatchback. My mom thought that the best thing about it was that 'Omni' started with 'Om' (she is Hindu). Om? The car provided little peace with its uneven engine and body exquisitely designed to rust more efficiently than the competition.

In an incomprehensible display of customer loyalty, my parents next purchase was a 4 door dodge Aspen. This car had a 6 cylinder engine which produced a hamster-threatening 90 horsepower. The best thing about this car was the horrible white paint which perfectly obscured the bird crap...

The design and production of these cars ranks as the worst example of capital deployment in American corporate history. I mean, efforts like this turned off an entire generation of customers.

Nowadays, that doesn't seem all that hard... ;-)

Writing two or three very clever grammars.

How would it benefit Chrysler to open the source? According to this article, their software gives them a competitive advantage. Why would they want to share it with other companies?

I was half expecting this to be a story about Lisp.

I was wondering when someone would finally make a comment about Lisp. In fact, this could very easily be a story about Lisp. Using Lisp to create grammars to read and execute specification files. This developer was forced by the client to write entirely in Java, but would probably have used Lex and Yacc, plus Java, if given a choice. I suspect, folks around here could get the exact same results in Lisp. The key is, Lisp programmers have likely been taught to think about the problem in a similar manner. The key is how you think about the problem, not the specific tools you use.

Actually, this writer's conclusion, open source, is the EXACT OPPOSITE of what I believe, and what I think is the premise of this forum.

Mega-institution isn't aware of its own golden needle in its haystack? Good! Another opportunity for a couple of hackers in a spare bedroom.

I don't see the opportunity.

It doesn't sound like a hacking problem; it sounds like a sales problem. If your customers are so hidebound they won't even use the software they already have, how will you convince them to pay for, let alone use, yours?

Or are you planning to duplicate that effort and sell it to other manufacturers (who I hope can still act rationally)?

If your solution is so compelling and the demand is there, find better customers.

"Good! Another opportunity for a couple of hackers in a spare bedroom."

Right-O.

Agreed. The problem is, hackers don't tend to even have access to this type of problem because you need access to the very real environment of a large factory floor. And, even if you did build a better solution, there are relatively few customers, so selling is really about selling into large corporate bureaucracies, not getting someone to download a cool new plug-in. This type of selling is very expensive, and they will not buy from a couple of hackers in a spare beadroom

Right. Now, change "billion" to "million". Find a little work in one of the thousands of small businesses in your neighborhood, and within 2 weeks you'll find yourself screaming one of two things, either, "Who wrote this &^#$^#&!" or "I can do way better than this."

Congratulations. You just found the basis for a viable start-up.

That I can do!

Those 4 words still send chills up my spine, especially with the exclamation point. (I know, what a nerd.) Hate to get all Oprah on you, but sometimes I think having that attitude is half the battle.

Given enough incompetent developers or bad managers any project can be a "7 Person Year Application".

give me 2 months and I could do away with AOP by showing how (comparatively) trivial it is to implement in Lua

Reply on news.ycombinator.com