Any changes to the UI should be done using a CMS (Content Management System) and should go live without requiring a deployment. This is important especially in enterprise environments where there are strict rules and timelines regarding production releases.
Of course, you may not care about this if you’re a single developer who deploys an app to Amazon EC2 anytime, with a few clicks.
Let’s see how can this decoupling be achieved. How can an arbitrary CMS provide the content, dinamically, to some webapp? One possible solution lies with the OS: symbolic links. You will need one or two such links, it depends.
First, the CMS should be able to export the content to a file system. All the work done in the CMS takes the form of a folder which could go directly on the application server or somewhere else, as long as it is on the same network.
Second, if the content is exported outside the application server, you have to create a symbolic link between the application server and the file system where the contents folder resides.
Third, once the application-server “sees” the contents, the webapp deployed on it needs access to them at runtime. To obtain the access, at start-up, the app should create a symbolic link between its root folder and the folder with the contents. When it comes to URLs, this means that the path to a file will come right after the context-root:
Now the web application will always deliver the files found under the folder contents, which in turn can be edited and exported from the CMS.
It should work fine with other architectures as well (where components
are rendered on the server), but it will require a little more implementation effort.
For instance I once had this setup for a JSF
app and all the server-side UI components had to be designed as custom XHTML tags.
Someone would make the HTML page in the CMS, I would give them the tag
and they would simply add it in the source wherever the component had to appear.
Finally, I’m aware that this might be quite a trivial topic. However, I decided to write about it because I know quite a few developers who make deployments with changes to error messages.
How do you separate the front-end from the back-end?