Ask HN: Any web startups using Java?

How times change - when we started in '95 the fact that we were doing a server side application in Java was seen as a very cool thing.

However, at some point about '99 or so Java seemed to endure a culture change where Java became popular as an "enterprise" platform and the values associated with that culture started to dominate a lot of discussions about how to do things. Suddenly everyone wanted to use every feature of J2EE in every project (largely for CV padding as far as I could see) - with predictably awful results.

However, Java has always been a pretty decent platform - we wrote search engines in it and things like Lucene continue to work pretty well. There are a lot of fascinating libraries for Java and performance can be excellent - it could be pretty fast in 2001/2002 and I assume it hasn't got slower....

If you can use Java and focus on "Keep Things Simple Stupid" then I can't think of many reasons not to use it - it might not be fashionable any more (those days are long gone) but it sure is capable.

I would totally disagree with you on that. If you use only the IOC controller for injecting your classes, maybe, Spring is not bloated, but also, its usefulness is not that great. But as soon as you begin to enter the wonderful spring-mvc world (which is the standard way to handle web apps using Spring), bloat is all around.

Take a look for example at the family of spring MVC controllers (http://static.springsource.org/spring/docs/2.5.x/api/org/spr...) : before doing any development, should you have to study such a family to know which controller should fit everyone of your use cases ? That, for me is the very definition of bloat.

Furthermore, the documentation is terrible, IMHO. That said, I have been spoiled by months of Django development, maybe it is not that bad according to Java Standards.

More generally, the main problem with Java is the emphasis on architecture astronautism, not the language itself, which is ok, if not very powerful (no function as first class objects, really ? ).

I'll give you my take on this as a Java programmer of 10+ years experience.

The problem you will have with Java in a startup is that Java's frameworks are by and large built for "enterprises". I put that in quotes because in one sense it doesn't really mean anything. In another, it's more of an idea of what governments and large companies want.

Even when using Spring (which, in my opinion, is a "must have" for pretty much any Java Web project), the amount of boilerplate required to set up an endpoint, map form data to objects, map those "presentation objects" into "business objects", map those "business objects" into a DAO (data access object), etc is pure torture.

Now this isn't completely the fault of the language. The language certainly doesn't help matters by being statically typed. This just doesn't gel well with how "fast" Web development works. Compare that to PHP (as just one example), where form data is just a map (hash or simply "array" in PHP parlance). Add a new field? Not a problem.

Add a new field in Java and you'll be making class modifications in about eight different places.

The bigger part of the problem is not the static typing though, it's the philosophies that dominate the Java landscape. There is a joke about the Java programmer's response to any problem is "just add one more layer and it'll be OK". It's funny because it's true.

It's fair to say that Fowleresque division of responsibility based layering is pervasive.

Of course the dynamic languages can have other problems (eg no error when misspelling a form field) and Java's static typing has, in my opinion, made Java's IDEs the best of any language or platform bar none (IntelliJ in particular). IDEs seem less able with dynamic languages because it's much harder (if not downright impossible) for an IDE to, for example, derive the members (let alone the types of those members, if that concept has meaning) on the fly.

Weirdly some in the PHP world have tried to mimic Java's deep layering with these horrific (imho) MVC frameworks that (again imho) simply combine the worst of both worlds. Dynamic class loading, bootstrappers, magic (and sometimes unpredictable) file loading, enforced directory structures and so on are just the wrong approach most of the time.

Java does have some benefits though. There are basically three tiers of languages in terms of performance (from best to worst):

1. C/C++/Assembler: the true compiled languages;

2. Java, C# (and the other .Net languages): the bytecode or virtual machine "semi-compiled" languages; and

3. Python, Perl, Ruby, PHP, etc: the scripted languages.

I qualify this by saying: most of the time (particularly for Web apps and especially if you don't have Google-level scale problems) performance doesn't matter. What matters is productivity and productivity is similarly tiered from worst to best in that above list. Hardware is now so cheap that it's far more effective to throw hardware at most problems than it is to write things in C. This is almost universally true when it comes to the Web where network latency and the speed at which the browser can process your page (and Javascript) will dwarf serverside performance most of the time.

I did a bunch of the Facebook puzzles awhile ago. The breathazlyer one was quite interesting. I started doing that in Python but it is somewhat difficult to get a solution to pass in that. Java is an order of magnitude faster. C is one or more orders of magnitude faster yet again.

So back to your original question (now that my answer has some context): I would say that using Java may make sense if everyone knows it and is comfortable with it. If not you will find many speed barriers along the way of learning a new language that may just slow you down or (worse) may create huge problems (eg security issues you weren't aware of because something didn't work the way you simply assumed that it did).

None of these problems are insurmountable but the way I figure it is this: when writing something new, chances are that within a year you'll be best off throwing it out and starting again anyway as the problem changes, you identify your bottlenecks (often not what you thought they'd be to begin with) and your scale changes.

So don't try and find the perfect language or platform or framework. Write something now and don't try and solve every problem you may ever have today.

There is play framework, which is very good and comparable to RoR.

"It just works - exactly as documented." -- I guess that's the one thing that Java is really good at.

I would even say you could go as far as saying it one end of the spectrum. Where you have frontend web apps (html/js/css) on a variety of platforms[1] on the one end and the big corporate backed languages and frameworks on the other end.

[1]: This can be a real pita. Though it's being mitigated more and more by adding compatibility layers on top (jQuery et al).