Why I will use Java EE (JEE, and not J2EE) instead of Spring in new Enterprise Java Projects in 2012
The question comes up often. It came up in my new project in November 2011, too. I will use Java EE (JEE) instead of the Spring framework in this new Enterprise Java project.
I know: Several articles, blogs and forum discussions are available regarding this topic. Why is there a need for one more? Because many blogs talk about older versions of Java EE or because they are not neutral (I hope to be neutral). And because many people still think thank EJBs are heavy! And because the time has changed: It is Java EE 6 time now, J2EE is dead. Finally! Finally, because not only JEE 6 is available, but also several application servers (not just Glassfish as reference implementation). I do not want to start a flame war (too many exist already), I just want to describe my personal opinion of the JEE vs. Spring „fight“…
Therefore, I think it is very important to start with a short overview and history of both alternatives. Afterwards, I will list the differences of both and explain why these differences lead me to JEE instead of Spring for most new Java projects. I am explicitly talking about new applications. If you have to extend an existing application, continue using the existing framework!
One more disclaimer: I am talking about mission-critical Enterprise Java applications. I am not talking about a little internal application or other uncritical stuff. I also would prefer using a combination of Scala, Groovy and Clojure persisting to a NoSQL database while being deployed at a PaaS cloud service such as JBoss OpenShift or VMware CloudFoundry…
General Information about JEE and Spring
First, I want to summarize some general information about JEE and Spring:
- In the end, both alternatives consist of several libraries which can be used by developers to create enterprise applications.
- Both can be used in most use cases, they have very similar functionality (business logic, transactions, web-frameworks, whatever…) – they only differ in realization (e.g. declarative transactions in Spring vs. conventions in JEE).
- You also can use only one or some of the available libraries. You can even combine JEE and Spring stuff.
- Usually, the crucial question is: „Should I use JEE (i.e. especially EJB, JPA, CDI, etc.) or the Spring core framework (i.e. especially Spring Application Context, Spring beans, etc.) for realizing my new application? Mostly, you can choose both, it does not matter from the point of view of the end user. But you should not merge both, this only creates higher complexity.
- There always was a debate about which alternative to choose. It is very difficult to discuss this question in a neutral way. That’s why almost all discussions end up in praising one framework and bashing the other one (I hope to be neutral in this blog post).
History: J2EE was horrible, so Spring helped!
J2EE was horrible. So much XML configuration, so many interfaces, and so lame application servers. This is why the Spring framework was created. It solved many problems of J2EE. It was lightweight, easy to use, and applications could be deployed in a web container (such as Tomcat) instead of a heavy J2EE application server. Deployment took seconds instead of 15 minutes. Unfortunately, JRebel did not exist at that time. The Spring framework is no standard as J2EE, nevertheless it became very widespread and an large community arose.
Today: J2EE is dead. JEE „stole“ the lightweight Spring ideas!
Everything started with a little shortcut change. J2EE was dead. The new shortcut was JEE. JEE 5 was born in 2006. It „stole“ many good, lightweight ideas such as „convention over configuration“ or „dependency injection“ from Spring and other frameworks. Yes, JEE application servers still were heavy, and testing was almost impossible. Nevertheless, developing JEE applications was fun with JEE 5. You did not have to write 20 interfaces when creating an EJB. Wow, amazing!
Then, in 2009, JEE 6 was released. Development is so easy. Finally! For example, you have to add only one annotation and your EJB is ready! Of course, the developers of the Spring framework did not sleep. Much new stuff was added. Today, you can create a Spring application without any one XML file as I have read in a „No Fluff Just Stuff“ article some weeks ago. Besides, several really cool frameworks were added to the Spring stack, e.g. Spring Integration, Spring Batch or Spring Roo.
Today (November, 2011), both JEE and Spring are very widespread and have a large community. Much information is available for both, e.g. books, blogs, tutorials, etc.
So, after I have described the evolution of JEE and Spring, why will I use JEE in most new Java projects?
Pros and Cons of JEE and Spring
A decision must be made. Which alternative to use in new projects? Let’s look at the pros and cons of both. I will add a „BUT“ to the Spring advantages – these „BUTs“ are the reason why I prefer JEE to Spring.
Advantages of JEE
- JEE is a set of standard specifications, thus it is vendor-independent. Usually, several implementations exist of a specification.
- Sustainability: Well, this is the advantage of a standard which is supported by several big players.
- Yes, believe it or not, testing is possible! Lightweight application servers and frameworks such as Arquillian arrived in the JEE world!
- Convention over Configuration is everywhere instead of explicit (I know that some people will disagree that this is an advantage).
Advantages of Spring
- You do not need a heavy JEE application server, you can deploy your application in a web container such as Tomcat.
BUT: JEE application servers are not as heavy as they were some years ago. Besides, the JEE web profile can be used, too. You do not have to use a Tomcat or Jetty to be lightweight!
- Spring offers features which are not available as JEE standards, such as Spring Batch.
BUT: You can add such a library to a JEE project without problems. You can also add other Spring libraries such as JDBCTemplate or JMSTemplate (which help reducing some boilerplate code) if you want.
- Spring offers much more flexiblity and power, e.g. aspect-oriented programming is more powerful than JEE interceptors.
BUT: In most projects you do not need this flexibility or power. If you do need it, then use Spring, not JEE – of course!
- Faster Releases (because it is no standard and only one vendor). The reaction to market requirements is much faster. Some current examples: cloud, mobile, social computing.
BUT: All enterprise projects – of many different clients – which I have seen, are not that flexible. Enterprise applications do not change every month or year. If there is a project, where you can change your version very easily, Spring might be better than JEE under some circumstances. But in most enterprise projects, you cannot simply upgrade from Spring 2.5 to Spring 3.x or from JEE 5 to JEE 6. I wish this would be possible, but low flexibility and politics rule in large companies with thousands of employees.
Conclusion: I will use JEE in most new Enterprise Java Projects
Due to the reasons I explained against Spring in the „BUT“ parts, I will choose JEE in most new Enterprise Java projects. Nevertheless, I will sometimes use a Spring libraries, too (such as Spring Batch). Sometimes, I will even have to use Spring (if I need its flexibility or power), but only then I will choose it. Of course, for existing projects, I will continue using the framework that is used already. I would probably not migrate a Spring 2.5 application to JEE, but I would migrate it to Spring 3.x instead!
So, I have described my reasons why I will use JEE in most new Enterprise Java projects. If I have missed something, or if you have got another opinion (probably many guys have), you can bash me in the comments. I appreciate all „non-flame-war“ discussions…
Best regards,
Kai Wähner (Twitter: @KaiWaehner)
Tags: Application Server, J2EE, Java, Java Enterprise Edition, JEE, JEE 6, JEE6, JPA, Spring, Spring Framework, spring roo
This is the answer to many questions. I agree with you!
I agree with everything you said. But spring has a great autentication and authorisation framework (old acergy) integrated. And its help us a lot. JEE has something like that?
Hey Daniel.
Spring Security may be one of the features which is more powerful than the JEE standard. In our last projects we had to implement low-level LDAP or NTLM integration for some reasons. Usually, we use the Atlassian stack (Jira, Confluence, Crowd) whenever possible because IMO Atlassian makes awesome easy-to-use products for a fair price. Crowd is used for Single-Sign On.
Best regards,
Kai Wähner (Twitter: @KaiWaehner)
Thanks for writing this up.
Even though you concluded with JEE, seems like a pretty even fight. I think developer experience will probably sway the vote in most EE environments. Dev. teams are probably even harder to switch then frameworks.
@Danny:
You are absolutely right! The “fight” will not end, because you can realize most applications with both alternatives. I also know developers who prefer Spring. Of course, developer experience is important, too.
Though, the trend goes in the direction of JEE (in my opinion). In a while, there will be much more JEE developers than Spring developers available. One reason is that all guys coming from universities will learn the lightweight standard and there is no reason to learn Spring at the university anymore (contrary to J2EE times, where Enterprise Java was too complex to solve problems quickly).
Best regards,
Kai Wähner (Twitter: @KaiWaehner)
Isn’t Oracle the Vendor who controls JEE? Sure they maybe a commitiee but isn’t certfication only possible from Oracle?
I would also argue that as SPRING can do what JEE can do, then it comes down to where the most developers are at. And there are more that have learned SPRING than JEE. Agreed that is due to the legacy days of J2EE, but that is the way it is.
Also what about using TomEE?
http://jaxenter.com/tomee-be-small-be-certified-be-tomcat-38434.html
Using SPRING with this you get the best of both worlds no?
Hi,
I think a comparison between JEE and Spring for the sake of comparison has no real value. I support the approach of using the minimum libraries you need for the project as you need them, and not decide in advance to go all-JEE or all-Spring.
You should first start (at least in your mind) with a main() method, and ask yourself what else you need: DB access? REST endpoints? Transactions? Then pick the tools that you think fit best for your specific needs. You may end up deciding that you don’t need JEE/Spring, or even a web container at all. For example you can do amazing stuff with Guice.
JEE containers may be a lot better today than J2EE, but that doesn’t mean you should use one if you don’t get a real benefit from it over the alternatives. As for the specific JSRs, you may or may not use each of them in your project (some JSRs are also supported by Spring), but that too shouldn’t necessarily lead to all-JEE architecture.
Personally I agree that the more flexibility you need, the better you choose Spring and other open source libraries. If you stick to JEE, you may need to wait for years to get features that are available today in such libraries.
@Gabriel:
As I have stated at the beginning of the article, I had to choose between JEE and Spring for future projects – I did not say that I wanted to choose
In theory, your suggestion may be possible. Also in some companies, this may be possible.
In practice, best-of-breed is not possible in many large companies. Politics and other keywords such as “technology consolidation”, “know-how transfer”, and so on prohibit your suggested approach.
I have seen this in several companies – and often it is not a bad idea IMO. My company also does Enterprise Architecture Management. There are so many application silos with different technologies, and the people with know-how already left.
Best regards,
Kai Wähner (Twitter: @KaiWaehner)
@LazyFan:
Well, you could use TomEE and combine Spring with JEE. But why would you do that? IMO, it would only increase complexity. If you want to use Spring, you should use its tc Server (and buy support for this server). If you combine Spring and JEE, you have to buy support for both, as Adam Bien stated some months ago in a blog:
http://www.adam-bien.com/roller/abien/entry/java_ee_6_xor_spring
You may argue that you do not need that support, but in mission-critical systems, this support is necessary.
I am no expert in this TCK certification stuff, but many other vendors made the certification for JEE in the past.
Best regards,
Kai Wähner (Twitter: @KaiWaehner)
Hi Kai! It’s not bad to have a bias in this discussion. If view points don’t differ then where is the value add in discussion. Seriously, I believe you have a strong JEE bias (you only have the “buts” against the Spring option) and you explain why. Again, that’s OK and its the only way I see to add value here.
I do agree to many points you mention. To me its not a matter of features to be honest in the end ’cause features are well sorted in both architectural options. It’s also not a question of “heavy weght” or so. In the end (to me) it’s a decision where you want to place your framework libs: in the server lib directories or with the WAR/EAR file in WEB-INF/lib. Now what’s the consequence: if you have all the stuff in the server lib and you use those APIs directly then
(1) migrating JEE servers will become non-trivial due to the migration and testing efforts associated
(2) therefore not all the clients want to migrate cause they don’t see a reason -> result: multiple JEE server versions in IT landscape, multiple local dev env versions due to different Java versions in production, more complex ALM solutions – all in all: more complexity and more heterogeneous landscape
(3) shift of responsibility: in Spring scenario the development or application architecture guys decide what to use und also have the knowledge to fix problems in production. in a JEE case the operations team is in charge of the application framework libs, they typically don’t have the knowledge and need vendor support
(4) less flexibility: e.g. you need to talk to procurement guys before you can add a new framework feature of a new server version (licensing costs), or e.g. you need to talk to operations team to upgrade. as with Spring you just compile a new version of your WAR/EAR file and deploy that to a java runtime of your choice
(5) snowball-effect: you need one feature but ’cause the app server you’re using (and in JEE you’re not the only stakeholder that can decide which server to use!) doesn’t have that feature you need to upgrade. Ups, you need to migrate and retest all your application ’cause all the API versions changed!
Now, all in all I think I do have a strong Spring bias. As soon there is a solution to these extensive problems I will change my mind ’cause obviously my live would be much easier if I could just say: we use the JCP “standard” programming model. Unfortunately, this is not the case right now.
Also, I do believe it’s not JEE against Spring, it’s CDI vs. Spring IoC in particular and its also about like EJB or not. Also, it’s a lot about independence of your business applications from the underlying application server runtime. E.g. we use EJB but the business app doen’t know it. The key to platform independence is not using either Java EE or Spring but making sound decisions which APIs to use directly in your business apps. And sometimes its better not to use some app server lib dir derived ones
JEE is moving towards a good direction. More modularity (as a standard) would be desireable to make it more applicable to complex and large IT environments.
Cheers, Niklas
All comparisons between JEE and Spring i have seen were biaised and not serious.
A good comparison should involve every single module/package of the spring framework. And explain how you do it with pure JEE.
A lot of third libraries offer a spring namespace extension. Will it be simpler with JEE only ? For exemple : BPM, ESB, nosql databases (all the spring-data-* stuff) …
And what are “entreprise java projects” ? Is it webapps … or complex integration systems ?
And today we have cloud platforms. Do you want to use them NOW, or wait for java EE 7/8 ?
So what happened to all the new ideas of spring which come with Spring 3.1. The Spring Data and Cloudfoundary efforts (agreed that you can integrate things around, but then who is to stop you from integrating things with .NET or RoR then)
The It seems that you are comparing stolen_from_spring_of_yesterday_framework to spring_of_yesterday. I am not so clear about standards.
I think standard is some thing which is widely used and spring beats JavaEE in 2011 IMO.
@Kai
Yes I totally agree on support. TomEE+ is very new and will no doubt get support. As we know Tomcat already has support from not only Spring Source but Mulesoft as well. And your article was about 2012. I think you will find that it will get that commercial support in 2012
Let us say you need Spring feature X, well just use it. TomEE+ is clearly the future. Which I think you will find is 2012 as per your article.
Also as a side note I am assuming Oracle control the certification for JEE. And so far, they have been reasonable. So, far! Let us suppose they get heavy with say your JEE container (Larry is already in an official War with IBM on hardware) get it black listed. Now you will have to move or you could just swap to Spring in TomEE+.
I lol’d @ “stolen_from_spring_of_yesterday_framework to spring_of_yesterday”
So true. We should have respect for Jurgen Hoeller and Rod Johnson’s work. Through spring, they’ve popularized so many concepts and patterns …
For people interested in an approach of migrating your Spring legacy apps towards Java EE, have a look at our JavaOne talk “Best Practices for Migrating Spring application to Java EE 6″ on slideshare: http://www.slideshare.net/ertmanb/javaone-2011-migrating-spring-applications-to-java-ee-6
Both of these:
EE is a set of standard specifications, thus it is vendor-independent. Usually, several implementations exist of a specification.
Sustainability: Well, this is the advantage of a standard which is supported by several big players.
These are very weak Pros IMO. Spring is not vendor specific either. I also think there are many big players supporting spring as well.
From a dev point of view I would much rather work with spring than JEE, and drawing good devs and keeping them working efficiently and happily will determine your project’s success or failure much more than the JEE vs Spring decision.
I recently worked on a project whose architect choose JEE just for the reasons you mentioned. The project failured under the bloat and dissatisfaction of devs wanting to work efficiently but just couldnt with JEE being the bloat it is.
Spring is not just one library, it’s a full software stack, so what are the real alternative stacks on the JEE side you are suggesting?
IBM WebSphere, Oracle Weblogic, SAP NetWeaver – not free.
GlassFish? Oracle killed most of it’s ecosystem: OpenSSO, OpenMQ, OpenESB, SailFin. And after merging with WebLogic, it’s quite obvious it will be more like Oracle’s “Express Editon”.
Geronimo? More like a mix of other Apache projects. Not really popular in enterprise or elsewhere.
JBoss? Yes. Real alternative. The only one. So isn’t it more about JBoss 7 vs Spring for ones without IBM/Oracle/SAP(usually stuck on old versions with EJB2)?
I was just thinking about “When do you need a JEE container” … and came across this blog post.
I know the answer is, you want to write a JEE app … but that leads to when do you want to write a JEE app .. and then to this blog post … my question came from a dev manager who wants to switch his unit from Weblogic to JBoss for cost savings, but then asked me why they can’t just use tomcat (like we had done at a previous company) … and I think the author is right, corporates are going to pay for JEE servers, and if you have a JEE server all the for/against arguments about spring/jee are close to valid … if a little biased
This kind of begs the question: just what is an _enterprise_ java project? i.e. when JEE is required. In my experience, the motivation seems to be mostly related to (transactional) messaging/JMS integration and JTA (or perhaps JTA transactions across multiple data stores).
Thought I’d mention that if you end up using Groovy and Grails comes along for the ride, then you’ve already adopted Spring.
I also wonder if you’re targeting PaaS would you still use JEE?
Hey Steven,
IMO, an “Enterprise Java Project” is a medium to large project which solves core business problems. It is in production for several years, versions cannot be changed instantly due to high complexity and large code basis. On the contrary, internal projects (e.g. for evaluation, proof of concept, some cool but not required core functionality) are no enterprise projects.
Of course, you do not always have to use Java EE (including an application server) in such projects, often Java SE or the integration of one or two Java EE APIs is sufficient.
I agree: Java EE ‘s main motivation is transactions and IMO easy thread management.
I do not have a problem using Spring libraries, even in combination with Java EE. Though, I would probably use JSF instead of Grails when using a Java EE application server due to its better integration into other Java EE APIs such as CDI or JPA (and not because Grails depends on Spring libraries).
Finally, yes, I would also prefer Java EE for PaaS! I think there is no difference… I have looked at both alternatives a little bit, e.g. VMware Cloud Foundry (Spring) and JBoss OpenShift (Java EE). Both are not production ready yet, but I would prefer a Java EE PaaS for same reasons as in an own hosted solution.
Thanks for your perspective Kai. I’ve just started evaluating Cloud/PaaS offerings (particularly for Java apps but also more broadly). Good to know that you don’t think either Cloud Foundry or OpenShift are production-ready yet. Others I’m looking into are Heroku, CloudBees, Amazon AWS, Google App Engine and Microsoft Azure.
What I understood reading your article is that you have never gone so far to use the advantages of Spring framework.
If you think about modularity, library switching (just because you don’t want to use that 4.1 crappy XX version) as well as testing (did you really used Arquilian in a large enterprise project?), Spring is THE only viable way, in my point of view, JEE is like using cathedral stones to build a castel.
So far so good, you have to spend more time on the elaboration of a Spring stack, spending your nigths reading documentation and elaborating POCs, and need more knowledge about the underlying libraries, is it really evil?
What I understood reading your article is that you have never gone so far to use the advantages of Spring framework.
If you think about modularity, library switching (just because you don’t want to use that 4.1 crappy XX version) as well as testing (did you really used Arquilian in a large enterprise project?), Spring is the only viable way. In my point of view, JEE is like using a built cathedral stones to build a castel.
So far so good, you have to spend more time on the elaboration of a Spring stack, spending your nights reading documentation and elaborating POCs, and need more knowledge about the underlying libraries, is it really evil?
Kai, please note, there is nothing named “JEE”. The correct name is “Java EE”. Thanks.
Bill Shannon
Java EE Spec Lead
JEE6 is a worthy alternative to Spring in the specific scenario cited (clarity over support if the customer has already invested in a JEE app server) but in most large corporate environments they prefer to stay 1 major release behind the latest and won’t be going with a JEE6 compatible WebSphere v8 or whatever any time soon. So, if the choice is between JEE5 and Spring as it is in reality in most cases right now then there is really no contest – Spring wins. Sooner or later though, JEE6 is going to win out, its inevitable, so start using JEE JSRs (e.g. 299) wherever supported by Spring today. BTW if it wasn’t for Spring, Java would be long dead. J2EE was awful and too many got burnt by it.
I would chose spring, and here is why.
http://www.vmware.com/company/news/releases/vmw-spring-momentum-3-14-12.html