Hacker News Reader

Comment by preordained | Hacker News Reader

What did he gain? Some performance? Did he need it, and was there no simpler way to get it?

I get the premise, just trying things and having fun. But I get the impression that in a lot (most?) scenarios this would be a lateral move at best. His last comment about going with Phoenix instead of Rails for new projects, if he's not just being kind, that seems like a major case of throwing out the baby with the bath water.

Goodbye 10 years of experience, knowledge of pitfalls, etc...hello 1 month of promise and years of gotchas and relearnings for small benefit?

View on HN

I used to be very actively involved in the ColdFusion community, and heard the same criticism when I'd encourage others to expand their skillset.

Being aware and trying things is one thing, I meant more a seemingly cavalier attitude of taking all your eggs out of an old technology basket and putting them in a new one, with slim justification.

And then there is survivorship bias. What if by some strange turn of events, ColdFusion had continued to evolve and eaten its respective niche? While I wouldn't condone an attitude of not being open to new things, the ColdFusion stalwarts would look like sages, while people who invested in an upstart technology could well have wasted a lot of time.

The move "backwards" to older, more mature technologies happens often enough, too. Startup uses latest NoSQL thing, and then realizes an old timey SQL database would have saved them a lot of trouble, etc. It cuts both ways.

I used my ~9 years experience in ruby/rails and invested 100% in elixiR/phoenix. I've found elixir way more simpler and elegant for my tasks, to the point that I start my new projects in phoenix (and ported some rails app to it too).

I still respect and use ruby/rails, but it's not my first choice anymore.

I 100% respect that. It sounds like you didn't jump to that conclusion in 1 Phoenix project over a month, though...and when you say "my tasks" the situation may have been that you weren't churning out "normal" CRUD apps that Rails is so good at?

For sure its a lack of careful thinking and consideration of trade offs I'd take any issue with...it sounds like you didnt come to those conclusions out of emotion or a deep yearning for something new.

Yes, I approached elixir/phoenix (well, my very first project on it was a sort of ecommerce library), btw it's true that I didn't choose in only one month. Moreover, I tried other languages/techs during the years (Go and NodeJS/Express, for example).

And yes, sometimes my tasks go beyond the simple CRUD, but as of today, I'd choose Phoenix anyway because I find mysef more comfortable with it rather than with Rails.

Even the fact that I can think to use a 512mb DigitalOcean droplet without fear is a plus, but not the main reason :-P

The performance question is a good one. Erlang has never been a performance wonder, so I'd be surprised to hear that Elixir is noticeably faster than Ruby.

What are you talking about, its like an order of magnitude faster than ruby? It might not be as fast as Go or Java or whatever, but its not even a comparison to ruby, and particularly rails.

Is it really that much faster? I mean, Rails is a separate problem, but I would have put Erlang in the same perf bracket as Ruby/Perl/Python.

Perhaps in raw computation its more similar, but based on the context of your comment (and its parent), it seems more like you're comparing it to ruby+rails in a real-life scenario (say, a web app). (The original comment was about "what did he gain"?)

In that case, the multithreaded nature of Elixir and the VM there's no comparison.

I tried searching for raw ruby+erlang benchmarks but couldnt find much in the short time I looked.

Can confirm, elixir+phoenix outperforms ruby+rails by ~10x.

If we're talking about pure ruby and elixir, things become complicated, you can do some stuff a bit faster in ruby sometimes (where ruby is using native optimized C code)

Remember that the starting point for your comparison is Ruby. It's a great language, but not one you reach for if performance is a primary consideration.

DHH even acknowledges this: "But sure, I'd like free CPU cycles too. I just happen to care much more about free developer cycles and am willing to trade the former for the latter."

Reply on news.ycombinator.com