Hacker News Reader

Comment by eloff | Hacker News Reader

>We are generally at the forefront of trends and big changes in tech as they roll out, which keeps it fun and challenging.

Still phasing out jQuery and using php and mysql. I'm sure their streaming tech and video player are cutting edge, but otherwise their stack sounds like a large, legacy webapp.

View on HN

A "legacy" web app that happens to perform at the same scale as YouTube. Maybe there's nothing wrong with using simple, battle-tested technologies after all?

Seems I rubbed some people the wrong way. That's ok. PHP is a disgusting language. Of the over twenty languages where I have more than passing familiarity, PHP and Perl are the worst. There are better options almost irrespective of the use case. JQuery was fine back in the day, but there are much better options now too. If you find yourself working with these technologies, and you don't like it, find a new job. You couldn't pay me to work with them anymore.

I wasn’t actually commenting about the language itself, more about the fact that you can go very far with just a server-side language and powerful hardware without having to use overhyped stuff like microservices and/or Kubernetes which apparently everyone thinks is a must-have nowadays.

Sure, just about anything could work. With a dynamic language your hardware requirements will be about 10x more. But that's a tradeoff you might want for possibly faster development time.

What are you basing your 10x comparison on? What costs 10x less? Does it consistently cost as much?

I'm sure they would love to hear from you if you can cut 90% of their hardware costs without increasing other costs.

Anecdotal reports basically, obviously 10x is just a round number I picked. I've seen a company that switched citing up to 40x hardware reduction going from Ruby on rails to Go.

10x cost savings is hardly unheard of.

I think you're significantly underestimating how well-performing PHP is, in the narrow but important field of serving dynamic web pages. Go (or Java, or C#) can be faster but you might be surprised that it's less of a slam-dunk win than one might expect. Python and Ruby basically aren't in the same league. Nodejs can be said to be licking at its heels, at best. Yes, seriously.

I'd put PHP somewhere under the middle of the list of languages/ecosystems I'd like to work in again (RoR, specifically, though not Ruby generally; non-TypeScript JavaScript; and gnarly AbstractFactoryImpl-style Java are all probably worse than it, IMO, and of course PHP deep in anything like Wordpress is pretty miserable) and I like Go quite a bit, but there's no denying that if performance is a top criterion PHP should probably be on your short list when choosing a language/ecosystem, and if PHP's not going to cut it then probably no scripting languages are on your list, and you're very likely looking real hard at C++.

Based on these benchmarks - https://dizzy.zone/2018/01/23/Kestrel-vs-Gin-vs-Iris-on-EC2/ - Go could easily be reducing costs 10x at a PHP-based company. Just look at Go's memory requirements. It's insane.

When it comes to cost to performance ratio your best option is avoid cloud providers like the plague and go for good old bare-metal servers (from a third-party or on-premises if your scale warrants it). You'll get a 10x improvement without having to change any software at all. Then if that's still not enough you can consider switching languages.

Eh. K8 is for orchestration, right, not performance.

Also adult site’s performance is insanely read heavy. They can cache the shit out of content and whatever is behind that doesn’t have to be the fastest. That model doesn’t apply to all software needs.

Since when is jQuery, PHP and MySQL "legacy?"

Half the web runs on that stack.

And a significant number of financial and airline systems run on COBOL. Active usage doesn't make something more or less 'legacy'.

Brand new systems are being written all the time that use PHP and MySQL.

These are actively developed technologies that new programmers are learning all the time, with vibrant ecosystems of suppliers (like wordpress plugin and theme developers, or mysql performance consultants), that aren't showing any sign of fading.

Again, that doesn't compare to COBOL.

This is true. They're still subpar options though. I mean if you have to use them because that's what your team knows and you don't have time to learn something else, fine. But if you're building a new team and hiring for those skills you're making a mistake.

They're fine, as good as anything else right now. PHP can be very high-performance, and MySQL/MariaDB, when run in the strict modes (usually enabled by default now) it's ACID-compliant, fast, and still easy to administer.

I was part of a team that launched recently on C# (.NET Core) and MariaDB, but if more of us were comfortable with PHP I'd have had all confidence that that'd have been just as good a platform.

I don't see how it's a mistake. What wouldn't be? node.js? Something ruby- or python-based?

Use Python or Go. Better thought out languages, comparable or better tooling, libraries, and performance. A while back I would add ruby to that list, but Python seems to have won the market over it.

So, again if it's what your team knows best, it's fine. Otherwise it's a mistake.

I would like to argue C# is a well thought out language with great tooling, libraries and performance (not C(++) performance, but do we really need that if we could iterate 10x faster?)

The type system being static with dynamic support and reflection enables me to think about real problems instead.

I used to dislike GC languages, but working with C# changed my mind. (not saying refcounted systems aren't nice, I'm following rust from a distance)

I like C#, no problem with it as a pick.

I don't believe dynamic languages deliver on faster iteration over the long haul. I used to be the biggest dynamic languages fan, but now I avoid them unless it's a small project.

PHP is a mess of a language that's been cleaned up somewhat well, but it's still at such a disadvantage.

I would never advise a company to use it for a new project unless they have lots of talent very familiar with it already.

It's just a mistake. You can disagree, but know that you're wrong ;)

When I was a full time C# dev (after doing ruby for years), I found that ReSharper turbocharged my productivity to the point where I was faster with C# than I ever was with Ruby.

But credit to ASP.NET Web API being so solid too.

Now that I almost exclusively code in Elixir, I'd love to try F#.

i would like to see someone rewrite wikipedia in another stack and compare.

I guess I should do more of my own research (I am security ops, not dev) but when did jQuery become legacy?

Long time ago, even the creator of jQuery says don't use it.

With the advent of new frameworks like React (7 years old now) and Angular, jQuery has become less and less used. Anecdotally, I have interviewed at 7 or 8 companies over the last 7ish years and only one still used jQuery, and they were moving away from it.

jQuery provided a lot of quality-of-life functionality that is now part of the native Javascript language/runtime on modern browsers.

The most famous likely being the selector querying functionality, and the event management (and I'm sure others I am not thinking of).

jQuery still works, but under the hood it's just proxying those native methods, so you might as well just use those native methods directly.

if it ain't broke, don't fix it

Do suggest a better alternative.

I would use vanilla JS rather than jQuery now that most browser compatibility issues are sorted. But most of my work doesn't need to support versions of IE older than 11.

which is exactly what they are doing (also , nominally vanilla JS is even more legacy than JQ)

isn't jQuery nowadays also using all native functions of a browser? I still find the regex selector so powerful in jQuery.

What do you gain? What do you lose?

jQuery has a better syntax and smaller coding footprint.

Reply on news.ycombinator.com