Hacker News Reader

Comment by andrewflnr | Hacker News Reader

I find his point about the "economic need" for programming to be more efficient interesting: at the time, software was about as expensive as hardware, and hardware was about the get drastically less expensive, and so

  If software development were to continue to be the same clumsy and
  expensive process as it is now, things would get completely out of
  balance. You cannot expect society to accept this, and therefore we
  must learn to program an order of magnitude more effectively.

And yet, society has accepted it. It's now a truism that programmers cost more than hardware. Then again, it doesn't seem like his hoped-for revolution has occurred, either, so I guess he hasn't really been disproven and is merely guilty of being too optimistic.

View on HN

These are just engineering tradeoffs. Sure you need to weigh them carefully, but the fact that we spend more on software than hardware does not indicate a horrible failure in the industry.

For example, processing power, communication speed, and storage has improved by orders of magnitude, but battery capacity is perhaps better by one factor of ten. Much of the advance in battery life comes from reduced power consumption by other components. Much of the advance in software development has come by creating inefficient abstractions that let us create things faster but squander CPU, network, and battery life.

Most companies today are willing to burn up some hardware performance in order to reduce software development time. If they're burning too much, they'll fix it or they'll go out of business. Premature optimization is still not a good idea.

I guess my post sounds more pessimistic than I meant it to be. I don't think it indicates any horrible failure, I just find funny the exact way that Dijkstra's vision failed to come to pass. Honestly, as a programmer, I kind of like it that hardware is cheap and programming is expensive. ;)

This statement dates back to a time when people would buy a machine and hire some programmers to write a program for it; that is, it's about the enormous expense of one-off, in-house, custom programs. IBM's System/360 was around but portable operating systems hardly existed yet (In 1972, Unix was just getting started).

Since then, programs are shared and run on many more machines; the cost is amortized through portability and writing general-purpose tools. Dijkstra was right to say that something had to change, but he didn't anticipate the efficiency gains from sharing portable software, at least not in this essay. Instead he seems to have thought that programmers should learn to write programs faster. We've made some progress on that, but the great efficiency gains happened at a higher level.

It's sort of swinging back around with the scale of some of the systems in place today. Relatively small increases in program efficiency can save a lot of money in power consumption, physical server space needed, and other related data center costs.

Reply on news.ycombinator.com