About PayPal's Node vs Java “fight”

http://developer-blog.cloudbees.com/2013/12/about-paypal-node-vs-java-fight.html

by stephenconnolly • 13 years ago

220 207 13 years ago

developer-blog.cloudbees.com

puppetmaster32• 13 years ago

Summary: Most frameworks sux.

romanovcode• 13 years ago

Or PayPal programmers.

zabraxias2• 13 years ago

Wrong. PayPal refuses to use open source frameworks because of an internal culture (and self preservation interest) of some very old "architects" who really only care about being in charge of the next internal framework. The best part is they later get to showcase how much better the new one is compared to the old framework, even though it's multiples slower than the slowest open-source counterpart.

cgh4• 13 years ago

Spring is open source. From the sounds of things, they replaced Spring with a closed in-house framework.

ricardobeat• 13 years ago

Their node.js "framework" is really just a couple modules sitting on top of express, and they are all open-source.

auvi• 13 years ago

TL;DR from the post: """ My take-home for you, the persistent reader who has read all my ramblings in this post...

Don’t build your app on top of a pile of crap in-house framework. """

kcmarshall• 13 years ago

Which suggests that take-home #0 should be: Don't build a pile of crap in-house framework.

Sniffing that out before too late is the tricky part...

_random_5• 13 years ago

That is a refreshingly sober point of view. Blaming frameworks is easy. Speaking of technical debt is hard.

Xorlev2• 13 years ago

Well-reasoned.

I've always found framework overhead and especially tech debt to be the largest contributors to performance issues. Especially when said frameworks have lots of ways to do things and they're not all the best way to do it.

Modern JVM development starts to look a lot like Sinatra -- you don't get a lot baked in, but you probably didn't need it anyways.

pron13• 13 years ago

Yes. JAX-RS (the standard implemented by Jersey), which is part of the often maligned Java EE, is a (surprisingly?) well designed REST "microframework". Java EE in general has come a long way from the monster it used to be, and is becoming quite pleasant.

edwinnathaniel9• 13 years ago

You can use Servlet + Jersey. It's just one dependency away and minimal web.xml config (or spring-jersey...whichever). It's not necessarily tied to Java EE albeit Java EE app-server _must_ implement JAX-RS properly.

curun1r• 13 years ago

The biggest contributor to performance issues these days, from what I've seen, is synchronous code. The JVM can be blindingly fast at running your code, but when it hits a point where you make a database call or web services call that's implemented synchronously, it sits there, waiting, doing nothing. No amount of VM optimizations can increase of the speed of idling code.

This is why Node apps often out-perform Java apps despite running on a VM that's significantly slower. Node is pervasively asynchronous...it's baked into the DNA of the platform and almost all libraries follow that philosophy. And as organizations move to a micro-services architecture where one external call can result in multiple internal calls between systems, the need for reactive, asynchronous programming becomes all the more glaring.

That's where this critique, for me, fell down. It's spent too much time comparing the JVM to V8 and not enough time talking about how the applications being run were designed. Because CPUs are fast enough these days that any of the VMs will give you decent performance for workloads that aren't CPU-bound as long as you, the application designer, do the right thing. It's just that some environments, Node in particular, make it easier to do the right thing. This will get easier in Java with the introduction of lambdas in Java 8 and the inevitable asynchronous community frameworks and libraries, but right now it's still harder to write reactive applications in Java than it is in Node.

bad_user4• 13 years ago

Node is pervasively asynchronous

Personally I view that as a drawback. It's all fine, until you end up with CPU-bound tasks or until you end up talking with backend servers that don't support many concurrent requests.

The JVM support for multi-threading is absolutely stellar. You've got true 1:1 kernel-level threading, with no GILs and you've got access to the best primitives available for dealing with multi-threading issues, atomic references, re-entrant read-write locks, fork-join thread-pools, non-blocking data-structures, asynchronous I/O guaranteed to work well cross-platform, all of them backed by a guaranteed memory model that also works well cross-platform, plus garbage-collectors that are the most advanced garbage collectors in mainstream usage. G1/CMS are amazing for servers and if you've got cash to spend, Azure's pauseless GC is probably as close to real-time as you can get with a GC.

And on top of these you can build whatever concurrency handling architecture you want. Including Erlang-style Actors (e.g. Quasar, Akka), STM (Clojure, Scala-STM), persistent data-structures (Clojure, Scala), asynchronous streams (Play2's Iteratees, RxJava Observables, Scalaz Machines, Scalaz Streams).

Even when speaking of Node.js's strengths, like asynchronous I/O, I still think Java is better because of the foundations (e.g. NIO, Netty, Mina). The only reason people don't work much with asynchronous workflows is because the language is so freaking painful for that. But that will change with Java 8 and people are doing wonders in Scala/Clojure.

I've personally built an entire web-service, designed to handle thousands of requests per second per frontend instance, with responses generated in under 30ms, with a completely asynchronous workflow, on top of Scala/Akka/Play2. Node.js people don't know what they are missing.

smoyer• 13 years ago

Management won't take the blame but ... who allowed the architects/engineers to build two systems that failed the same way?

wpietri• 13 years ago

I agree that management is likely a problem here, but disagree with the framing.

If you're in a structure where management's job is to allow/deny, then I think the code is already doomed. Software development is work that can't be effectively micromanaged.

In software, I think management's job is to support and challenge. All the people coding should be professionals, thoughtful and aware of issues like tech debt, second-system effects, reinventing the wheel, and so on. Management's job isn't to be more professional than the pros; it's to support the pros and make sure they're living up to (and sometimes advancing) their professional standards.

I'm sure there are plenty of people who work with these frameworks that know that they're crap. Management didn't fail by allowing these to get built. They failed by creating power relationships that allowed the people who couldn't see the problems to control the building of the frameworks, and to suppress criticism and improvement.

sehrope• 13 years ago

Regarding the JVM performance:

Under minimum load the best page rendering time was 233ms

Regarding the node.js performance:

Under minimum load the best page rendering time was 249ms

What the heck are they doing that takes 250ms that could possibly be app code related? That's a lot of CPU usage if its actually doing CPU-bound work. Since these are web apps that's obviously not the case. The request time is going to be mostly waiting for external resources (DB, message queue, etc).

Without knowing how much time is spent on those external requests these numbers are meaningless.

Language choice is no where near as important as your DB-access patterns (and more generally DB performance, caching, etc).

If every request you process involves serially accessing a 50-100ms external resource four or five times then it will always take 250ms (though you could increase parallelism of multiple requests if done right).

cgh4• 13 years ago

The fact that both implementations are ridiculously slow tells me to look for commonalities. In this case, how much do you want to bet that there are some truly scary, gigantic and all-but-incomprehensible stored procedures underlying everything?

Xorlev2• 13 years ago

No bet. :)

camus23• 13 years ago

Language choice is no where near as important as your DB-access patterns (and more generally DB performance, caching, etc).

i'm pretty sure the paypal node app does not directly access the db , the node app talks to a private api.

Helianthus• 13 years ago

That doesn't really change the point, it just wraps it with a layer of abstraction.

kalleboo• 13 years ago

Language choice is no where near as important as your DB-access patterns (and more generally DB performance, caching, etc).

I bet they have a really crappy backend storage system for account history/data.

We have a PayPal transaction history with a low 6 digits record count. Searching for a specific email address, name, or transaction ID can take literally minutes. Exporting a transaction history CSV takes hours.

Negitivefrags• 13 years ago

It only gets worse from there.

When we exported a CSV of our transaction history there were entries missing. We knew this because each line has a running balance and it wasn't adding up with the line before it.

When we told our paypal account manager about this it triggered an investigation. They claimed they had never seen it before and they sounded very worried about it.

They then told us that they have multiple different databases with transaction histories and that they were not consistent. They said that they had investigated the problem and found that it affected only a small percentage of customers.

They then pointed us to another CSV export form that doesn't seem to be able to be accessed via the normal UI and were told that this one would give us correct results. Unfortunately we were unable to export an entire year at once with this form as it seems to have some kind of time limit.

We had to export our history in little chunks that would take hours each. We asked our paypal account manager to prepare the CSV for us after finding this out, but she refused.

Having a large number of transactions in paypal sucks very bad.

Compare this to Stripe that proactively emailed us saying that they had noticed that queries for filtered transaction sets were taking a very long time so they had done some optimizations and were asking us to see if it had improved.

Joeri• 13 years ago

Maybe we should assume the best instead of the worst. We could assume that paypal has capable developers who are knowledgeable about both the jvm and node.js, and who have concluded that their specific featureset can best be built using node. That is not a value judgment of the jvm vs node in general, as their situation is theirs and theirs alone.

spullara• 13 years ago

Two poorly written apps compared with blame for each placed on the underlying virtual machine rather than the part between the chair and the keyboard. I'd much rather compare VMs and systems with something a little more structured and open like the TechEmpower benchmarks. They still have their flaws and simplifications, but at least the source for all apps is open and you can see and contribute to them.

srean• 13 years ago

This was a good read. Minor nitpick, given the author's stress of not comparing apples to oranges I think the motivating anecdote on protein folding struck a dissonant cord. A fair comparison would have been a profile guided optimization with comparably aggressive compile optimization flags.

For these types of problems the profile characteristic is fairly static, so there is hardly ever a need to pay for the warm up time.

Java can indeed be very fast, and in my opinion JVM is the one of the most optimized VMs that we have now. That said Java code beating a C or a C++ code on a CPU bound task raised an eyebrow. In my experience Java usually gets to the 80% speed of C++ quite easily but at a cost of 2.5 to 3.5 the memory. For these types of applications memory tends to be an expensive resource. For our number crunching server, 80% of its cost is sunk in the RAM.

stephenconnolly• 13 years ago

I debated including that example, but at the end the point I was making is that the JVM can be faster than native code in some cases... this case had static allocation of data and the calculations could benefit from MMX instructions when available, plus there was a logical unrolling that I think it was able to infer from code path analysis

rossjudson• 13 years ago

And superword optimization, when possible...

pron13• 13 years ago

My experience is similar: Java will get you 80% of super-optimized-compiled C++ speed for the cost of more memory (although you may choose to give Java less memory and pay in speed), and will get much closer to C++ speed with some work (off-heap memory, etc.). However, Java has the upper hand in two scenarios: 1) long-running applications that are developed by a large team – those usually make heavy use of virtual inheritance for the sake of good engineering, and the execution profile is not static, so the program can greatly benefit from JIT optimization; and 2) fairly complex multithreaded code – Java's incredibly useful and well-implemented blocking and lock-free data structures, as well as an incredible work-stealing scheduler (all expertly programmed by Doug Lea) make great use of a general purpose GC, plus Java usually gets them about 5 years before C++ (if C++ gets them at all).

edwinnathaniel9• 13 years ago

Don't forget the tools and infrastructure that support good engineering (dependency management, build, javadoc, findbugs, checkstyle, PMD, cobertura for code-coverage, better unit-testing frameworks, the list goes on).

I'm sure C++ ecosystem has some of those but whether people are using them or not seem quite obvious: how many C++ projects do that out there? Not as many as the Java counterpart.

abritishguy• 13 years ago

Damn, paypal really suck at making fast sites.

tunesmith• 13 years ago

I wonder if a lot of headway could have been made simply by doing some constraint analysis on the Spring framework and figuring out how to improve the performance. I have a hard time believing that Spring itself was completely topped out. It might have been an architectural choice, or some bloat that could have been taken out incrementally.

Instead they seem to be going the route of "the unicorns will solve it". The overall cost of that approach seems higher - finding unicorns is rare. Finding unicorns that are very good at all areas is even harder. I know there are a lot of unicorns on this board, but there sure aren't a lot of unicorn resumes on a hirer's desk.

I think the value of Spring has always been not that it's super-high-performance out of the box, but that it does a lot for you, and that as you learn more and more of it, you can develop solutions fairly quickly. It's built to make it easier on the developer, not the servers - but it is still easy to improve performance by focusing on constraints.

I think it's easy to blame what's between the keyboard and the chair, but the availability of said unicorns is the one ingredient of the system that is not in an employer's control. What is in control is sponsoring more education, and focusing on simplicity and constraint analysis.

ebiester9• 13 years ago

I fantasize about rebuilding the java ecosystem. Create "JPM", don't allow anything from the old Maven system in it.

Create a new "hip" minimalist framework that ignores everything about java and its notions of OOP. Basic dependency management handled NPM-style. Developers encourage to put everything in the same package for their own stuff, and everything is package level. Getters and setters and encapsulation and java's poor implementation of OOP ignored.

Heck, fork Spark, split out the individual pieces as modules, and start there.

All "JPM" packages expected to be under 100 lines, a la NPM. Frameworks are discouraged. Take what you need a la carte.

I'd argue that java is a better back end language than javascript in and of itself, but that the community norms have led to the pain we see.

Have an extreme preference for 1.8 code (IE, the new java.time system) and away from anything bloated. Rethink for the common use cases, rather than trying to support everything under the sun. Spring still exists for that, after all. :)

The problem with java isn't that it's a BLUB language anymore, but that it's a BLUB ecosystem. I'm not saying the language is perfect, but Java 8 as good as ruby or javascript for web development if the ecosystem is ignored.

zaphar• 13 years ago

Java has always been a good language if you ignored the ecosystem. The ecosystem is a large part of what defines a language though. You have to take care in how you define that ecosystem.

I sometimes wonder if Node.js is making mistakes in the opposite direction as Java with their ecosystem these days.

gngeal4• 13 years ago

Java has always been a good language if you ignored the ecosystem.

You mean "no unsigned types (even though the language is essentially machine-level when it comes to integers), no compact memory layouts for aggregate data, no useful function values without jumping through hoops etc." is an approach that leads to good languages?

dustingetz• 13 years ago

A language IS its ecosystem. In 2013 java is a good platform.

pron13• 13 years ago

I think that much of your fantasy is already a reality. For example, look at how Dropwizard[1] assembled a best-of-breed collection of modern Java libraries into a small, yet very powerful, REST micro-framework.

So the bits and pieces are certainly there, and even the JCP does good work sometimes (JAX-RS) that becomes good standard APIs with many different implementations.

[1] http://dropwizard.codahale.com/getting-started/

nogridbag• 13 years ago

Ninja web framework [1] is also worth a look. It's similar to Dropwizard in that it's less of a framework and more like a well thought out archetype building upon the best Java libraries. Two main differences I see are Ninja uses Guice for DI whereas dropwizard is manual DI. Dropwizard uses JAX-RS while Ninja has its own Play 1 style routing system.

While I really like JAX-RS, the reference impl Jersey uses HK2 for dependency injection, which makes it odd to integrate into a Guice application (in fact Guice support is still lacking with Jersey 2.x).

[1] http://www.ninjaframework.org

candl• 13 years ago

Dropwizard is pretty great, but one unfortunate aspect is that it's stuck with Jersey 1.x for JAX-RS and as a result is missing some nice features, like async responses, etc.

ebiester9• 13 years ago

Dropwizard is a bit heavy by Node standards. The health checks, in the NPM-style system, would be its own package. The configuration class would be its own package. Resources would be their own package. Liquibase or Freemarker wouldn't be built in, nor would JDBI.

I think of Dropwizard much like express for node, and most of the "cool kids" in the node community think that that's just too much for a single library.

I like Dropwizard, don't get me wrong, but it isn't quite a micro-framework by other standards.

fat0wl5• 13 years ago

Hear hear. When I read about JVM performance issues I usually immediately think "team of 20+ complacent corporate devs who don't care if this succeeds or not because they can get a new job insanely easily at a moment's notice".

The issue with these other frameworks/languages is that even when modern frameworks are used there are usually some Java backend services (or another "powerful-outside-of-the-web-also" language) if the app does anything substantially complex. A ton of the hip web tech is basically for building front-end and cobbling together relatively inefficient backends.

I was an RoR guy myself but moved to Java & I see the correct hipster tech here but just lacking unanimity. I bet if the RoR community decided "Clojure hell yea" or "Scala hell yea" and contributed lots of plug-ins to those ecosystems, they would generally have similar development times with a lot of performance gain for free. I'm thinking about trying Play framework on Heroku just to see what the monthly bills & New Relic readouts look like in comparison.

cloverich2• 13 years ago

Here's[1] a comparison of the two from an RoR guy who used Play! (with Scala) for 16 months. I found the ensuing discussion pretty informative with minimal FUD / bashing.

[1] https://groups.google.com/forum/#!searchin/play-framework/pl...

radicalbyte2• 13 years ago

You've basically described C#:

Getters, setters fixed? Check. Functional features? Check. Most crappy OO features fixed? Check. Trendy package system inspired by npm/apt-get/gems? Check. Healthy community with libraries replacing frameworks? Check.

I'd recommend checking it out if you haven't already.

mynames• 13 years ago

Could you expand on "crappy OO features"?

lightblade• 13 years ago

I also like to know more about this C# package system you mentioned.

edwinnathaniel9• 13 years ago

Your package system is inspired by Maven, minus a good build system and other functionalities.

Try building a large C# projects with running unit-tests, integration-tests, packaging, code-coverage, code style baked into it, and a bunch of other plugins for your build systems without pulling your hair.

Your DateTime library ain't there yet sir so you're pretty much at the same situation with Java (although both ecosystems have NODA/JODA lib).

C# is just Java with 10% syntax improvement and -30% to -40% productivity lost when it comes to everything else.

mattgreenrocks5• 13 years ago

The "libraries over frameworks" notion is something I keep noticing in communities that seem to be more engineering-minded than rock-star-ninja-dev-minded. Where do you think this comes from? I think it's a mix of maturity on the part of the community members along with how often the language is used to do things that are harder than CRUD.

gngeal4• 13 years ago

And C# "fixes" the deficiencies by heaping features upon features. Already it looks like the MS version of Perl.

hansjorg• 13 years ago

Except that between the two, only the JVM is truly cross platform, Mono notwithstanding.

ebiester9• 13 years ago

Has the Mono story gotten better? How feasible is it to develop without thousands of dollars of Microsoft software these days?

WildUtah• 13 years ago

"You've basically described C#"

Unfortunately the CLR doesn't run on any useful platforms.

moomin2• 13 years ago

I'm sorry, but NuGet is a package manager that makes most Maven haters want Maven back...

rdmckenzie• 13 years ago

You've also described Clojure...

runT1ME2• 13 years ago

So why not scala? Scala 8 years ago is where java 8 is now, and you can ignore most of the java frameworks.

pron13• 13 years ago

Even putting aside my great dislike for Scala (which I discussed on HN ad nauseam :)), I think everyone would agree that Scala has a very different philosophy from Java. Java has been designed as a "blue-collar" language, with maintenance of a large project with possibly hundreds of developers as a central guiding principle. C#, Go, Dart, Kotlin, Fantom etc., have a similar philosophy. But Scala is a different beast altogether. You can love it or hate it, but even its most enthusiastic fans would concede that it is most certainly not appropriate as a replacement for Java in some/many/most (depending on your view) circumstances.

Also, Java is extremely popular, and some of its most painful points have been addressed in Java 8. Even an arguably better language (regardless of its name) will not easily replace Java. Hey, look across the browser at JavaScript. There are arguably better languages there; are they replacing JavaScript?

If you look just at GitHub – which is greatly skewed towards new languages – you see that all other JVM languages put together make about 10% of the JVM ecosystem, with Java being the 90%. Outside of GitHub, in "the real world", the situation is even more strongly in favor of Java. I don't know if this is going to change much in the next few years.

I don't mean that you shouldn't use an alternative JVM language if it makes your work better, and there are many viable and very useful alternate JVM languages – personally, I like Clojure and Kotlin – but you can't just "replace" Java so easily. I think efforts to reform the Java ecosystem are more likely to succeed – at least in the short term – than efforts that require switching to a different JVM language.

ebiester9• 13 years ago

Implicits, SBT being anything but simple, and compile times. :) People have suggested sticking to a subset of Scala for most purposes, but what happens when you run into a problem with a library? Suddenly, you're diving into Slick (from a friend's example) or Play (from a personal, painful example) and can't make heads or tails of the code because of how different the paradigm is.

Frankly, I'd rather just go to Erlang and skip all the compromises needed to stay with the JVM if I want that kind of language.

astral303• 13 years ago

Groovy with Gradle is a pretty good answer that is there today. Gradle way way way easier to work with than with Maven for your build system, publishes packages with dependency info, so you still get your dependencies sorted OK. Groovy has automatic setters/getters, a ton of functional-style helpers (closures, etc), dynamic & static mixins, delegates, etc.

ebiester9• 13 years ago

My day job is in grails. :) However, some shops still have an aversion to other JVM languages.

vorg3• 13 years ago

Groovy with Gradle is a pretty good answer that is there today

I've heard many good things about Gradle's functionality, but not so many about Groovy. Only yesterday ( https://news.ycombinator.com/item?id=6884355 ) I asked if Gradle will ever expose an API so developers can use the same language for the build scripts as they do for the projects they're building, i.e. Groovy with Gradle for Groovy projects, and Scala with Gradle for Scala projects?

usrusr• 13 years ago

Sorry, but in groovy, [[1 : 2]] - [["b" : "a"]] seems to be []. As long as they don't get fundamentals like this right i consider groovy only useful for demonstrating it's own deficiencies. The fact that this is in their Jira since 2010 and a fix would would be easy for anyone who understands the basics of hashcode and equals doesn't paint an all too rosy picture of the people behind groovy.

I'd rather work with a language that _i_ don't fully understand (modersky seems to provide me with an infinite supply of intellecutal challenges) than with a language that gives me reason to doubt that it is fully understood by the people who made it.

moomin2• 13 years ago

I think you want Clojure and leiningen.

al2o3cr• 13 years ago

"All "JPM" packages expected to be under 100 lines, a la NPM. Frameworks are discouraged. Take what you need a la carte."

LOL, I'm trying to imagine anything else in life being run like this. For instance, you go to the "car package manager" and are told that all parts must be less than 2kg and that you need to choose which ones you need...

Falling3• 13 years ago

I think your analogy makes a whole lot more sense when you keep in mind that you'd be a automotive engineer and not a consumer.

Periodic• 13 years ago

Except that there is implicit dependency management.

It would be like saying that all SKUs need to be comprised of other SKUs and less than 2kg of screws, nuts, bolts and custom components.

In your analogy the Spring might be an entire engine, spark plugs and all, built as a single unit. Want to use your own spark plugs? Tough luck. You get the whole engine or none at all, or you could just build your own from scratch.

m0skit02• 13 years ago

Forget about OOP. Check Clojure.

keypusher2• 13 years ago

Have you looked at Groovy/Grails/Gradle? You can use the existing Java code and ecosystem or you can choose to do things the new way, you have access to a web framework similar to Rails, it's pretty nice.

vorg3• 13 years ago

you have access to a web framework similar to Rails

Then why not just use Rails for Java, i.e. JRuby ?

lampe32• 13 years ago

I stopped reading at the part with the "Type information". Or better i should have stopped... Another post about(and the from paypal too) about my language X vs language Y...

It's the same flameware like Facebook started with its we can't get HTML5 as fast as native... then some people who know what they where doing did it...

No don't use java No don't use javascript

For god sake learn one language or two and learn it very well! i mean very well! and you will write good code and performant code...

EDIT: Why do i get down voted? Jesus HN gets worse and worse...

dangrossman2• 13 years ago

You're getting downvoted because this isn't a "language X vs language Y" article at all. Its point is that the language choice doesn't matter when you're building on top of crap, and jumping from one language to another probably won't solve platform problems. It's an article about the politics of being a software developer in enterprise where management is entrenched in legacy codebases, not about language features.

lampe32• 13 years ago

What? the first part is exactly that... for example: "So which is the faster virtual machine, V8 or JVM? My belief is that when you are coding in the most basic building blocks of either virtual machine (i.e JavaScript vs JVM Byte Code) that the JVM will win out every time. If you start to compare higher up, you may end up comparing apples with oranges and see false comparisons. For example, if we consider this comparison of V8 vs JVM performance at mathematical calculations. This blog post tells us that if we relax our specifications we can calculate things faster. Here is the specification that V8 must use for Math.pow and here is the specification that the JVM must use for Math.pow notice that the JavaScript specification allows for an “implementation-dependent approximation” (of unspecified accuracy) while the JVM version has the addition that"

and at the second part too: "... a limit of Node.js itself when used with the backing APIs that they have to call through to."

It's just hidden better... to me it reads like a java fan boy is crying why a company switched to language X from his language Y...

ExpiredLink3• 13 years ago

Spring brings a lot of functionality to the table. Likely far too much functionality. Most people are moving away from the monolithic application and moving towards smaller more lightweight frameworks…

It's high time to debunk Spring. In the name of 'Dependency Injection' many Java developers still create complex, monolithic and - first of all - untestable applications with this framework. In fact, Spring adds another layer of complexity on top of Java EE without real benefit. It fosters Anti-Patterns like 'Dependency Injection' that give you the illusion of modularity but instead create dependencies within the application ('Dependency Injection' needs to be taken literally).

mtrimpe2• 13 years ago

I'm pretty good with Spring and have even given trainings on it and I'd argue that Spring's main problem is that it needs to be seen in the right context: a poor man's DSL for declarative system configuration (it is last 10 years old after all.)

That it creates untestable applications seems like completely nonsensical FUD however since the entire point of Spring is to keep code easily testable. Spring is an almost direct response to the lack of testability of the ServiceLocator anti-pattern.

I'm not saying you can't make untestable code in Spring, but if you do the problem is most likely more with the people than with Spring.

penland• 13 years ago

I've done a lot of Spring side work and your first paragraph is the best explanation I've ever read for the framework.

I also agree that if you can't test Spring code you are doing something wrong. I get the frustration with having to create test harnesses for some things, and the SpringJ4Unit stuff oftentimes feels a bit too magical, but I've never seen well written Spring code that wasn't easily testable at the both the class and module functionality layer.

UK-AL2• 13 years ago

"It fosters Anti-Patterns like 'Dependency Injection' that give you the illusion of modularity but instead create dependencies within the application ('Dependency Injection' needs to be taken literally)."

What? Any system with separated modules is going to have dependencies. At least with DI, classes are written to a interface and different versions easily be swapped in and out.

porker2• 13 years ago

Any system with separated modules is going to have dependencies.

Thank you. I thought I was the only doing my nut in over "I've done DI, now I have all the same dependencies" ... adds DIC ... "Now I have all the same dependencies but higher up, and.. oh I need to add logging to 80% of the objects" [Goes to kill myself]

I have OCD about writing beautiful code. The more I try, it seems to real world prevents it from happening.

Eager to learn so if good demo code, resources, books or I'm missing something (obvious or not) I'm all ears.

iamleppert• 13 years ago

He forgets to mention one critical fact: developer productivity. Most applications (sans his terrible example of protein folding) are not that performance critical. Most of the code you write as a developer will do-away with any platform or language advantages itself.

It is not productive to write web applications in Java. Everyone hates it, which is why we have ruby, php, and now javascript (node).

I'd much rather trade a little performance for a scripting language anyday.

cloverich2• 13 years ago

I wonder if someone has written up comparisons on their productivity between the languages / frameworks. My previous job had me on the python stack (Pyramid / Mako / Sqlalchemy), which I loved, but have been on Java / Play! framework for the last 10 months. The one thing that has consistently surprised me is how productive I feel writing Java. Mainly, I notice I make far fewer mistakes / typos while coding; the IDE does a great deal of the work for me, and as the application grows, having static types has (for me) improved the readability of code I haven't looked at for a while.

No doubt a large part of any language choice is subjective, but I'm starting to believe that developer productivity has more to do with the tools and frameworks than the language itself. As these continue to improve in Java-land, I'm willing to bet that the disparity in productivity will continue to decrease.

goostavos• 13 years ago

The one thing that has consistently surprised me is how productive I feel writing Java.

I know exactly how you feel. I recently moved from a Java project to a Python one and felt a little amiss without the power that an IDE gives a statically typed language. The tooling for Python flat out sucks. And it pains me to say, as I love the language to death, but I do feel more productive in other languages. Whether or not that's true is another thing entirely, but it does feel like a lot of time is wasted calling `dir` on objects to figure out what the hell the name was of the function you need.

As of about a week ago I switched from my text editor + console to Eclipse + PyDev. It's not perfect -- intellisense functionality is really hit or miss, but it does make me feel a lot more productive. The background linting is pretty good at catching basic errors as you type them, which is really nice.

It's also does an OK job at renaming things -- which is generally a nightmare to do on a big project in Python.

So, point being, while I miss a lot of the other tooling that an IDE affords a statically typed language, PyDev gets you, like, 80% of the way there. Worth checking out if you move back to Python.

karlmdavis• 13 years ago

Everyone hates it ...

I don't; that's a bit of a sweeping generalization. I happen to enjoy writing webapps in JAX-RS and Wicket, and I know a number of folks who enjoy writing them in other frameworks.

UK-AL2• 13 years ago

Would really prefer java over php and JavaScript.

Closure, python, c#, f# over Java. But I quite like the java ecosystem. There's normally library for anything.

coolsunglasses• 13 years ago

Lots of people trying to unblub blub.

I'd venture a guess that mimicking Node.js isn't going to fix the deeper problems with Java.

gprasanth• 13 years ago

I was reminded that self deceit is the worst of its kind on reading the article. seeing a benchmark and deciding what to write your next application in kinda like that. you browse until you find the benchmark that aligns with what works out best for the lazy you and not works out best for the faithful app. this was just an observation of mine and not necessarily true or applicable to you.

neakor• 13 years ago

Application based comparisons are usually flawed, because the writer will never and cannot release enough info to reveal the real problem. Just because your bad-performing application runs faster on one platform doesn't necessarily mean that platform is superior to the old one. It simply means whatever problems you have in your bad-performing application manifests less in the new platform. In the end, fix your application!

kylequest2• 13 years ago

What a shocker... a Java guy says that you shouldn't abandon Java :-)

Lots of wild speculations... especially about the reasons PayPal picked node.js/JavaScript instead of a Ruby stack. He says that Ruby wasn't chosen because you can run it on JVM ignoring the fact that you can run JavaScript on JVM too using DynJS (he actually mentions DynJS earlier in his blog post).

pkolaczk• 13 years ago

So far I tried compiling some of my numeric Scala code to JS with Scala.JS and unfortunately running that code in Chrome is much, much slower (~10-20 times) than running the original version on Java 7. :( Not sure if it is because Scala.JS creates poor code, or just JS is still not as fast as they claim.

virtualwhys• 13 years ago

Current version of Scala.JS is an alpha release, don't expect miracles.

jroseattle• 13 years ago

This is more about PayPal's dirty laundry than an comparative evaluation of technologies. If a collective set of tools are used improperly, comparisons beyond the context of those walls means very little.

api• 13 years ago

TL;DR:

* Micro-benchmarks don't provide a good picture of real system performance and can be skewed to make almost anything look better/worse than anything else.

* Much of Java's "slowness," clunkiness, large amounts of boilerplate code, etc. is the fault of the Java ecosystem and libraries rather than the language itself or the JVM.

* The inverse is true for Node: the quality of its ecosystem helps make up for language and runtime shortcomings.

nswanberg• 13 years ago

This reads like someone reading a 1995 article about a company switching from writing Netscape Enterprise Server code in C++ on a Windows machine to Apache code in perl on a Sun server, and then writing a defense of the x86 processor against SPARC.

I might have missed something but I don't recall the original PayPal article arguing that the switch was motivated by moving from a slow virtual machine to a faster one.

But just as switching from C++ to perl would trade some computing power for productivity in getting a site built, I don't see it as controversial for an organization to say they achieved the same productivity gains going from java to javascript.

eddflrs2• 13 years ago

Does anyone know how vertx[1] fits into this picture? It gives you the JVM, JavaScript, and non-blocking I/O.

[1] http://vertx.io/

kylequest2• 13 years ago

Nodyn is an effort to run node.js apps on top of Vert.x. JavaScript alone is not enough to make Vert.x popular for node.js people. You need to be able to leverage the node.js module echosystem. Hopefully Nodyn will do that, so you can use your favorite node modules on Vert.x. It's still a work in progress though.

eddflrs2• 13 years ago

Cool, Nodyn looks interesting. Thanks for the info.

bhauer2• 13 years ago

I've had the misfortune of using the PayPal Java SDK to integrate PayPal credit card and Express payment into a web application. Based on that experience, I find it especially difficult to digest any opinion that PayPal has on Java. They may be experts on handling money, processing credit card payments, fraud prevention, and the like. (Or maybe not, given their reputation.) But their opinion on Java is about as meaningful to me as Microsoft's opinion on Java. Or more on point, a Rails developer's opinion on Java. Both are characteristically stale. And not "whew, that bread is really ripe, throw it out" stale. Rather, "what we have here is a new form of fungal life" stale.

I've worked with a fair number of Java APIs. I would put the quality of their Java SDK in the bottom 10%. It has the stench of really old Java. Remember how the first Java APIs reeked of C? The fact that the PayPal Java SDK still looked like hastily ported C code in 2011 when I last worked with it makes me worry that their entire manner of using Java is stuck in the early 2000s. That's being generous. It's the kind of Java you'd expect to see written by a company that still thinks about "applets" and "beans" and tries to come up with clever coffee double-entendres for their package names.

My theory: PayPal is big but they know they are vulnerable. They've acquired BrainTree and that helps. I suspect they want to improve their technical chops and felt a need to migrate away from Java in order to break free of their organizational quagmire. PayPal is from 1998 and if their Java SDK is any indicator, they probably have a Subversion commit log for their core systems going all the way back.

This sort of move is the kind of thing I could see myself doing if I were in a decision-making position at PayPal. Attempting to invigorate sensible modern Java practices within that organization would seem hopeless. I would have to start anew with an approach that is free of poisonous legacy.

As others have pointed out, the fact that a rewrite yielded so little performance gain is decidedly troubling. At least they got a fun and well-linked blog entry out of it. If I had just gone through that effort and reaped such a modest improvement, I'd be crestfallen.

ControllerLowDC• 13 years ago

Precisely.

I spent some time at PayPal and, one thing I think is being lost in all of these discussions is exactly what was "fixed."

All your points on introducing new practices to an older, establish (in the company) framework and technology are spot on. I see this as more a way to shake things up and improve things under the guise of new tech.

Also, this is in the top-most bit of their stack. All of the underlying services are the same. That's where the majority of the latency exists and a lot of it comes from old-school PayPal. But you can't shift a company by trying to tackle a deeply integrated bit of architecture first. The easiest place to demonstrate that change is feasible and actually quite beneficial (to a company that has been change-averse in the past) is your app tier. Get them on your side with an easy (easier?) win, then tackle the harder problems.

ariq• 13 years ago

Whose Java SDK would you rate at the top 10%? I'm curious.

bhauer2• 13 years ago

Just to compare apples to apples—by which I mean credit card payment APIs—Stripe's Java API is superior. I'd not say it's a thing of beauty, but then, I don't want a credit card payment API to be anything but immediately approachable and obvious. And it is precisely that. Easy to understand.

https://stripe.com/docs/api/java

As for something I am happy with in recent memory, the Undertow builder model is appealing: http://undertow.io/

But you ask a tough question. I've not spent much time thinking about the best of the best. I suppose part of it is that a great API gets out of its own way and lets you get to work. So I suppose citing Stripe's first may have been the right thing.

finnw• 13 years ago

PayPal do not care at all whether their APIs are developer-friendly. Why should they? They are a near-monopoly. You are completely dependent on them and they do not need you at all. I suspect their APIs are a result of minimising what they have to pay their own developers, because they win if they saved just one hour of development time even if 10 million vendors have to spend 10 hours each to work around the limitations.

In the very unlikely event that the developer has any influence in the choice of payment processor (i.e. you are a startup that intends to trade in only one country) they still do not care, because they still need to do the usual due diligence to check whether your company is a fraud/ML risk, but they earn little in fees from you. They would rather let a smaller payment processor deal with you, then try to take over later after your product is established.

puppetmaster32• 13 years ago

I think it comes to this: Java > JavaScript. But Java programmer < JavaScript programmer, due to culture. (ex: Spring, Maven, EJB light).

If Java programmer is true polyglot, then whatever tool you are using wins. That's why C programmers are the best.

tn13• 13 years ago

In our first lecture at graduate college, the prof told us how to read research papers.

Rule 1 : Louder the guest shouts of his honor, faster you should count your spoons.

Similarly if someone tells you about 50% savings just because you switched to some other technology you should be extremely skeptical. Besides those people who engage in `Node vs X` debate every now and then should spend some time educating themselves about compilers and interpreters to understand what makes some programming languages faster than others in certain circumstances.

peter_mcrae• 13 years ago

First post on HN. I'm not entirely sure how the PayPal Node vs. Java message has been packaged, but I think there is a lot of confusion. From my understanding (via blogs and conferences), Node was used to replace Java on the front-end. Meaning, the Node front-end is still calling Java + other stacks upstream -- which likely explains the performance benchmarks (the front-end wasn't the bottle neck). That said, even with similar performance benchmarks, I can see how Node might result in higher developer productivity for the front-end components of PayPal -- hence the enthusiasm they are showing. Lot's of negativity and blame on this thread, but I think it's because people are making assumptions.

exabrial• 13 years ago

Incredible insight... Here's the TLDR: Everything is faster when you engineer it a second time. Rewriting in a different language is not an apples to apples comparison.

About PayPal's Node vs Java “fight”

Related Stories