Hacker News Reader

Comment by pslam | Hacker News Reader

The only possible ways you can believe you can write issue-free large scale projects in C, which don't suffer from all the usual pitfalls of C, are:

* You've never deployed one.

* You've never had to maintain one.

* You wrote the last line of code and handed it over to the maintenance team, who then never gave you feedback.

* You have no customers.

* You have done a never-before-seen formal analysis of a large scale project down to the individual source line level, and made no mistakes with the formal specification.

The author is correct: use C only if you must. There are still an enormous number of reasons why you must use C, but it should be a constraint imposed on you, and not a language you actively want to start a project in.

View on HN

True.

Parent seems to believe in nginx. But as a guy who actually tried to do a few of those things with nginx I can confirm that nginx's code base is pretty awful and working around all the usual pitfalls of C in nginx is a huge PITA and enormous amount of work.

(same username on github)

> There are still an enormous number of reasons why you must use C, but it should be a constraint imposed on you, and not a language you actively want to start a project in.

It's the same as with, say, goto or #ifdef's - one has to consider the alternatives and make an informed choice. Most of the time, the alternatives are preferable, but sometimes there either aren't any alternatives, or they are even worse.

Which is not to say that C can't be a whole lot of fun. I love it dearly. But I, too, tend to avoid it when I can. It's just too easy to shoot yourself in the foot.

It's what they paid me to do in the 90s. That was a compelling reason to use C.

(of course, one employer appropriately used C as a portable assembler, and respected it as such, for the purposes of developing a cross compiler and tool emulations for an older minicomputer environment; the next was doing (batch) "business applications" and foolishly spent too little on hardware and too much wasted effort on application development)

Pascal (Modula) did much the same, SAFELY, but the money wasn't put into compiler optimization :-(

Interesting that the GNU compiler collection now includes a very efficient Ada compiler.

> Interesting that the GNU compiler collection now includes a very efficient Ada compiler.

GNAT has existed for ... at least ten years, I think. The DOD apparently paid for the development so there would be at least one open source/free software implementation. (According to Wikipedia, development started about twenty years ago, and it was merged into GCC in 2001.)

Reply on news.ycombinator.com