In almost every situation, developers will always prefer to create a brand new application rather than migrate a legacy application. When you migrate an application, there is a large amount of baggage that needs to be accounted for which means that you may end up spending a large amount of time dealing with the legacy baggage.
On top of this, sometimes applications can be so large that it can be almost impossible to rewrite the entire application in one go. Another option is to slowly migrate features only a few at a time while introducing tests to make sure that you’re not breaking any front-end functionality.
In this series of articles, I’m going to document my journey of migrating an 8 year old vBulletin based application that currently runs my hobby website ozfortress. My plan is to slowly migrate features to Symfony2 while leaving the legacy vBulletin application to handle most of the sites features until I find time and motivation to rewrite each discreet piece of functionality into a Symfony2 bundle.
Having chatted with many people about methods for migrating and how to handle which part of the site is handled by which part of the application, legacy or new. Below, I’ll cover a few basic methods that can be used to cope with setting up a migration environment.
The first method that can be used is sending every request through
the Symfony kernel, letting the router decide which requests should
go to a special
LegacyController which will pass off the request
to the legacy application.
This method is great because it allows you to configure the routing of the entire application inside the Symfony router and not have it mixed throughout the new and old systems.
Be aware however that this method means loading the Kernel and its hard depenedencies, followed by the router and any other Symfony systems for every request which will add overhead to the legacy parts of your application.
Falling back to Symfony2
Another method which will use some webserver configuration to only use the Symfony2 kernel when the request doesnt match an existing file path on the filesystem.
The method needs very little custom configuration as the default
.htaccess file follows this logic already.
The benefit of this method is that the Symfony2 system will not be initialised unless it is needed or if there is a 404 response required.
Including parts using ESI
Another option in the toolkit, while not specifically intended to be a complete solution to migration is using a reverse proxy that supports ESI like Varnish that will let you include ESI tags in the output of your legacy application where you want content from the new application to be placed.
The advantage with using this method is that combined with either of the above methods that you can include partial new functions inside pages of the old application as you migrate them.
In the next article, I will cover setting up an environment that supports falling back to Symfony2 with ESI capabilities, along with a strategy for sharing user sessions between the applications.