Managing Taxi Companies with Thin WARs and Java EE
Victor, please introduce yourself
Hi Adam, thank you first of all to consider me and my project for an interview :-).
My name is Victor Röder, and I'm simply a freelancing software developer and I'm living close to the geographical midpoint of the European Union (at least till 2019), that is in Germany somewhere 70 km north east of Frankfurt/Main.
You can find me on Github https://github.com/victorroeder and on Twitter @vicaware. But in both spheres I'm not that active.
You told me you are building a small Taxi App. Why Java EE?
Yes, in 2017 I actually planned to do a sabbatical and suddenly I got a request from several smaller taxi companies whether
I'm willed and able to build something that could help them with their paper-and-pencil-based taxi management. So it
would be a software that I could develop on my own, nearly without time pressure and the ability to earn some money,
so I said "ok". And - who guessed it - the backend of it is (of course) done using Spring.
Ok, kidding - Java EE :). Or how I meanwhile call it: JSYSNRTO (for 'Java Stuff You Should Not Rename That Often')
Why Java EE? Well through your workshops I could recognize how lean and easy Java EE is/became, so I wanted to use it in a real-life project as soon as possible. And this was an opportunity.
I used J2EE years before and was impressed how it developed. But of course - as every technique - you even can also make it mega complicated with Java EE. So I'm glad to use the Adam Edition of the Enterprise Edition :-).
What is the purpose of the application?
The purpose is to simplify the management of small and medium-sized taxi companies. Every taxicab gets an Android-based provisioned
tablet installed with which the drivers could record and manage their shift data and record-only the taxi tours (I'm
explicitly writing 'recording-only the tours' since some guys of the German Tax Authority might read this interview as
well... "hi guys..." :-D). The tablet itself is connected via BlueTooth to a small mobile printer to print out the tour
receipts for the passengers and daily receipts with all shift details at the end of the shift for the drivers themselves.
The tablets transfer the shift and tour information to the backend and request possible changes of the master data if there was a server-side change. The maintaining of the master data is done by the company owner via a web UI. The App for the tablets is installed via the Play Store and the tablets initialize and configure themsevles when they are registered (with a correct key) and accepted by the backend. An easy on-boarding (and usage of the software later on) was very important to me, since I did not want to become a support center :-).
Are you happy with Java EE so far? Is Java EE productive?
Happy: Very. Now that the product is in place since over one year and being now somehow in a retrospective mode while
preparing me for this interview, there are of course several things, I would not repeat in future projects. I'm using
Vaadin for the backend's UI and I'm quite sure that using JSF would have been more than sufficient and JSF would
have integrated better with CDI & stuff. Also I was playing around with the Fat Entity approach for the JPA entities,
meaning that the entities managed their persistence themselves (using the ThreadLocal hack), contain the code for
JSON de-/serializiation and other logic. Reason was, that a) I wanted to try it out and b) tried to find a way to
log (for audit purposes) changes - made by UI users - on JPA level. With hindsight, I would keep it simple. But ok
- honestly - I could not blame Java EE for that but rather you Adam, because you wrote about it in your book and
thus put the bee in my bonnet ;-P.
The last but not least point that comes into mind, is that the code for JAX-RS and the stateless session beans share the same classes. I should have split that up (and let the JAX-RS classes inject the according session beans - as you recommend it) even if this results in partially tiny classes.
Which application servers, tools or IDEs are you using?
I started development with Glassfish and in production it is still Glassfish. And since I was crazy enough to implement
a server-dependent JAAS module for login, my escape possibilities are limited... Locally and for my server's
dev environment I'm using Payara. I should align that in a timely manner.
On the other hand I was playing around with other application servers and find them more polished than Glassfish/Payara...
The IDEs: Android Studio for developing the App, IntelliJ for the backend and Eclipse for the BIRT reports. Build tools: Maven for the backend project (because I wanted to) and Gradle for the Android stuff (because I had to). Of course there are many other tools, reminding me that there are nearly no small or trivial software projects out there (especially when you are working on them alone). You have to take care of so many stuff (build & quality environment, database stuff (and integration with other systems), versioning and release process, etc.), that you should always reach your goal with the most possible minimum of complexity and dependencies. Fast deployment and fast unit tests is a key metric. And automate as much and as simple as you can...
You are using the Boundary Control Entity pattern to structure your applications. What were your experiences, challenges and findings so far?
My experiences with BCE are very good. It enforces you to structure your code from the business point of view and
not from a technical perspective. That even helps other developers to find their way through the structure more
easier. Packages named 'dto' or 'services' say nothing, but business or domain related names show (new) developers
clearly the way. Additional benefit: my scroll wheel is happy, too :).
But I used the pattern only for the business-related backend code. The server-centric presentation (and thus part of the backend project) and the Android project are structured slightly different by simply having expressive (again, from business point of view) package names and all topic-related classes are found their according package (without the deeper separation of boundary/control/entity). But classes that are cross-cutting are in the parent package (as it would be for BCE). No Util, Helper or Common packages or classes (bom chicka wah wah ;)
How big is your WAR? On which exernal dependencies are you relying?
4.8 MB. But with tricks. I put several external Jars (like those for Vaadin and BIRT) directly on the application
server. Around 850 KB of the total size is my code.
Especially bringing BIRT and the thin WAR approach together is impossible. I thought about externalizing the BIRT stuff into a separate WAR whose services (i.e. report generation) could then be addressed by HTTP/REST... Or maybe I misunderstood the integration of the BIRT runtime somehow...?
(oh boy, this interview creates or reminds me of a lot of To-Dos ;-))
You are an airhacks.com alumni. Do you had any interesting conversations with other attendees (in the extensive breaks :-)
Yes, I already joined several of your workshops and had short but interesting conversations everytime (mostly chewing
my food in parallel). And as far I remember all of them were related how to fight the technical or infrastructural
complexity beast or dealing with difficult developers, architects, managers, etc that are more interested in
creating themselves a monument within the project. None of them was about complaining "Oh, my business is so
complex and it's so hard to implement."
Yes, the conversations mirrored my experiences I made with software projects so far, that they are lacking discipline for minimalism on many levels.
And it's a pity that it's often like that...
Victor, big thanks for the interview!