Hacker News Reader

Comment by Zigurd | Hacker News Reader

Pair programming isn't widely practiced for obvious reasons: It looks like you just made your code twice as expensive. There are studies justifying the cost by showing it can be recovered with less testing, code review, and bug hunting. But that evidently hasn't been persuasive. Also, anyone familiar with statistical quality control (SQC or SQA) would look at pair programming and see 100% inspection. Very suboptimal in manufacturing.

Still, pair programming makes sense in the abstract: Coding is not manufacturing. Applying SQC to code is nonsensical. Under the circumstances, 100% inspection is all you've got. Pair programming probably makes sense in cases where minimizing bugs at every step, right down to the creation of every line of code, is needed because the cost of a bug explodes the longer it exists. Probably rare cases. But worth knowing to know that coders are not working on a production line.

View on HN

This makes sense if you see programmers as cogs in your code factory that produce exactly 1 coder day of code per day.

If, however, you see them as humans with different skills that can be shared amongst each other, then pair programming makes a lot more sense.

Yes indeed. I remember a full team of old mainframers going together in a room with printed code and review it together. 500% inspections. Many pieces of code would hold for decades without major refactoring. A lot of the code being produced today is quite garbage IMO, and you can see that in the infrastructure requirements it brings with it, growing faster than our bandwidth.

have you ever done pair programming? like actually sit there and be forced to do it? It's absolutely infuriating to anyone who values anything. I spent 20min trying to convince my "pair" that it was ok for a comment to be a sentence fragment and not a complete sentence. Every keystroke is the same conversation over and over. After that experience, i pulled rank and said i will never pair program again and haven't since.

> It's absolutely infuriating to anyone who values anything

It's not really your place to say what's infuriating to others when your pair programming experience with a _single_ (somewhat neurotic) programmer was irritating. In my experience, pair programming has been a delightful two-way share of ideas that was both refreshing and productive (but I'm not going to claim universality or such an experience).

Pair programming probably does suck when one side is extremely impatient and overbearing.

It can be terrible in practice. Same with test driven development, also annoying if instituted by proclamation. A lot of code doesn't need these formalisms. By that I mean: Do it very carefully in circumstances where there is clear justification, and with an adequate budget for tooling and support. These techniques were invented before there was good tooling to support them, which makes bad outcomes a lot more likely.

Pairing does not work if you believe you have to convince the other person that you are right.

What you did wasn't "pair programming", it was an argument.

It sounds like you were just using your pair to type for you.

Reply on news.ycombinator.com