The Joel on Software Discussion Group (CLOSED)

A place to discuss Joel on Software. Now closed.

This community works best when people use their real names. Please register for a free account.

Other Groups:
Joel on Software
Business of Software
Design of Software (CLOSED)
.NET Questions (CLOSED)
TechInterview.org
CityDesk
FogBugz
Fog Creek Copilot


The Old Forum


Your hosts:
Albert D. Kallal
Li-Fan Chen
Stephen Jones

Evidence for most bestest programmers

I have to admit, Joel's evidence is compelling. Based on CS-323 evidence, the Yalies have got a neat cirriculum.

10 programs in a semester (that's 15 weeks?), writing huffman tree compressors, command line and makefile interpreters - pretty cool stuff.

I am very impressed with the uncorrelatedness of the data. I think the most shocking point is how the top 25% sample set matches the overall sample.

So, whatzat mean?  That you can hire a very good programmer from Yale, that likely is an 'A' student - and yet you did not necessarily hire _THE_#1_PROGRAMMER_. Or as Joel says the top 1%.

OK, so say that I concede the point that the top 1% can specifically write software 10x faster (the data clearly bears this out and I think is really an amazing example to prove the point*).

What I love about the CS323 problems is that they are generic parsing CS core assignments - with some math thrown in. Not some "technology" (like Java or C#) specific infomercial.

Well, I'm rambling and I cannot create a conclusion because the implications... well... harumph.

But I cannot argue with what was presented. Uncle.

*I'm fully aware that proofs are not made by example. Software profit is rarely made on proofs.
hoser Send private email
Monday, July 25, 2005
 
 
Here is my evidence for best programmers.  No correlation data.

1. Linus Torvalds - Created the Linus operating system in 1991.

2. Slava Pestov - created jEdit at 14? is now 21, a Canadian developer, now working on a programming language: Factor Programming language.  Google on jEdit, there are a couple of open source IDEs, Eclipse is one, jEdit is another popular one.

3. Blake Ross(born June 12, 1985) - "is a software developer who is known for working on the Mozilla browsers, started FireFox"

4. Arfa Karim Randhawa of Pakistan - "10 years old believed to be the youngest person in the world to have earned Microsoft Certified Professional status, visits the company's Redmond campus."

5. Daniel Robbins - Linux Gentoo Founder: "Daniel Robbins joined the Redmond, Washington-based company on May 23. He had given up responsibilities as a Gentoo developer last April."

6. Marc Fleury, JBoss Founder: Born in Paris in 1968, Marc Fleury got his Ph.D in physics from the Ecole Polytechnique in Paris. Marc started the JBoss project in 1999.

7. James Hong - HotorNot creator
8. Google Creators
9. Blogline Creator (created early Bloglines.com in 3 months)
Berlin Brown Send private email
Monday, July 25, 2005
 
 
I am pretty sure Steve McConnell gave evidence showing (or at leasted stated) massive speed differences between different programmers. In his book Code Complete I believe (excellent book - must read)

(No doubt there will now be fifty responses saying 'but what about the quality'. Well, the studies factored that in - so fast but crap was not viewed as better than slow and quality. So, please don't waste your time replying if that was what you were planning to say).
Matthew
Monday, July 25, 2005
 
 
The surprise to me is the time differential between what are essentially the top 25% of the class.

The time differential among the rest is what I expected.
hoser Send private email
Monday, July 25, 2005
 
 
Perhaps Joel would post the raw data - at least for the top 25%. But that might also be construed as a breech of students' privacy.
hoser Send private email
Monday, July 25, 2005
 
 
The time differentials within the top 25% shows why it's important to look for somebody who was active in other activities while getting their high GPA.

Somebody who worked 20 hours/week and was involved in other stuff like the university radio station or student government probably didn't have the time to be slaving away at an assignment for 40 hours in a week.  That's the type of person you want to hire ... the type that gets things done and gets them done *fast*.
T. Norman
Monday, July 25, 2005
 
 
Here is some anecdotal evidence:  I am a somewhat slow coder.  In school I spent many long hours in the lab working on assignments.  At work I used to get pressure from my boss about not producing code faster.  The flip side is that my code has a very low bug count.  In my last project I measured the total bug count per developer and mine was the lowest by far.  When the product released I was the only developer to have no open bugs, and I was assigned other developer's bugs to fix.  A lot of their code had to be completely rewritten because it was rushed garbage that they clearly wrote without truly understanding the problem.  So I have a new boss who understands that I may seem slow, but over the course of the entire project I'm much more productive than my peers.
Productive Coder
Monday, July 25, 2005
 
 
What you want is someone who is fast and provides the correct solution. All these were the top students. You can also get someone who was slow and provides the correct solution. Many of these were also top students, though some of the slowest who got it correct after the midnight deadline scored badly.

You do not want the one who is fast and provides the incorrect solution. Yet this character is the one most in demand by corporations.
Rich Rogers
Monday, July 25, 2005
 
 
The article has an unfortunate vagueness in blurring between design, humor and software development.  The end result is a oddly prosaic elitism.

In a field like mine--computational linguistics and machine learning--the level of expertise and topical knowledge far exceeds trivial training via programming classes.  Spreadsheets and bug tracking software require nothing more than reasonable programming skills combined with an understanding of the appropriate frameworks.  Business logic compares similarly.  Harder problems in harder domains just don't compare and can't be commoditized in the same way (or outsourced, for that matter).

I have to admit to being an MS Alumni myself (from server apps), but really treasure the intellectual depth during my time at a Xerox PARC spinoff much more, where hard problems required more than just writing efficient Huffman trees.
Mark Davis Send private email
Friday, August 05, 2005
 
 

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics
 
Powered by FogBugz