Hacker News Reader

Comment by StevePerkins | Hacker News Reader

I'm being a more than a bit tongue-in-cheek here. Of course I do enjoy learning new languages, and accumulating at least a basic understanding of the technologies discussed here.

However, I notice that languages wash over HN like waves. A few weeks ago there were a glut of back-to-back Clojure posts. Last week it was all-Go-all-the-time, and over the past few days Rust has surged.

This wave pattern is a cultural quirk of HN, rather than serious indicators of what you should be pushing for within your company, or where you should be taking your career long-term.

View on HN

or where you should be taking your career long-term

Obviously, you have to learn Common Lisp if you're thinking long-term - the one language that people will still be using when all the ones you've named will be history!

I'm about to deep-dive into Erlang so consider me still susceptible to temptation :)

Erlang, in my opinion, is more of a Lisp or Smalltalk than, say, a Ruby, Python or Go: it gets some things really right, and some people will continue to swear by it, but it'll never see mass adoption, and sooner or later, other languages will rip off enough of the good ideas to be 'good enough'.

Erlang is an interesting hybrid between language and operating environment that offers a unique set of trade-off choices. The language itself is not as interesting as the result of these trade offs that translate into a reliability that seems hard to duplicate using other platforms, or at least the platforms that I'm aware of.

That said it is not a silver bullet, split brain problems and other nasties remain hard.

I don't think you can rip off the good bits of Erlang without using the Erlang VM and associated plumbing.

Right, what makes Erlang interesting is the VM and how it's put together, with its own scheduler. It's fairly unique, but something you could replicate elsewhere. Go has a scheduler too, although it's built differently with different goals in mind.

The problem is that you end up needing to do 'real world' stuff, and hit something like this:

http://www.erlang.org/doc/apps/xmerl/xmerl_ug.html

I just spent nearly 2 years on trying to replicate the 'good' bits of Erlang in a less exotic environment. It's been one hell of a ride (I never knew how much I didn't know) but ultimately fruitless because it is simply too much effort to build this for the one application we have for it. So, reluctantly, we have come to the conclusion that it is better to re-use an existing platform even if that means having to deal with the warts of those platforms.

Very frustrating. But then again, if it were easy it wouldn't make the money that it does. A defensible moat that is made out of hard work is a good thing in many ways.

I'd love it if you'd write something about what you learned through that process, what you tried that didn't work and so on. It's pretty significant that you couldn't make something good enough after two years and were driven all the way to Erlang. If that sounds skeptical, it isn't; I believe you, and that is what makes your remark enticing. I want to hear the story.

The reason why I haven't written this up so far is that this is mixed up in a bunch of other stuff, some of which is quite painful to those involved and it is very hard to keep it generic enough that fingers won't be pointed.

As these things go lots of money was blown, egos got frayed, potential customers were disappointed and time was wasted.

All in all a huge learning experience but ultimately mostly a harsh lesson in not biting off more than you can chew and how not to run a project. I was involved as a coder rather than in any other capacity but some of the responsibility for the way the project was run surely rubs off on me. For one I should have probably thrown in the towel earlier when I realized in how many dimensions we were deficit.

There wasn't a single point in time where I thought 'this is not the right way' other than right at the beginning, but once involved (originally just to fix 'a few bugs') I felt that I should give it my 200% to make it work.

Over the course of 2 years mountains were moved but alas that wasn't enough, and funds ran out before we reached a version stable enough that we could run our application on it. In some ways we got pretty close, in others we were still miles away from completion (most notably, the network core and the actual app). Precious time was spent optimizing things that did not need optimization and most of the lessons start-ups learn were re-learned, sometimes (very unfortunately) against better judgement.

I've been asked to look at salvaging this team/project/company and am working together with one of the founders to get things back on their feet in some form, so it is a very delicate situation.

Let me talk it over with the rest of the guys involved and see if they are willing to let me write it up, I wouldn't do it without their buy-in. One potential snag here is that there is still some lingering hope we can re-boot this thing on an existing platform side-stepping the bulk of the development and just concentrating on the app portion (which was my initial preference anyway so I don't mind that).

That makes sense. I thought I was asking a technical question, but it's never purely technical, is it?

> I thought I was asking a technical question, but it's never purely technical, is it?

Well, this one is technical in many ways and there really is an interesting angle there when you remove all the other stuff. That's still going to be a very harsh thing to see black-and-white in public rather than with context in private.

The tech part essentially boils down to 'the right tool for the right job' and don't try to fit square pegs in round holes when you're on a deadline.

The list of 'lessons learned' is rather longer than that ;)

Languages are a tool. They are fun to learn. I always want to use something new on a new project but then I realize that I need to get it done and I need to find people that can maintain it.

That takes the fun out of it. I do think that one can safely bet on scala or go and have fun.

If you have the right architecture you can be as polyglot as you'd like. It's just risky from a maintenance standpoint.

I think the wave pattern is just do to how the web is structured and the way people find new interesting articles. Nothing about the culture of HN.

There was a period of time when there was an absolute flood of Earlang articles posted here, perhaps a couple years ago.

iirc that was to drown out even less desirable articles.

Reply on news.ycombinator.com