But why is it parsed so much faster? As you said it's also xml?

It's faster because you're removing a layer of indirection.

In SOAP you have an HTTP request/response that contains your SOAP envelope which contains your data/methods.

With a REST model you just have an HTTP request. What you're doing is defined by the URI and the HTTP request.

REST is a different way of modelling your distributed applications. You break down your application into URI accessible resources, each accessible with a common set of methods (traditionally HTTPs GET, PUT, etc.).

So rather than send a SOAP request to your central server with a next-job-id message, and getting a SOAP response that contains your job-id XML, you would just GET http://example.com/next-job-id/ and get your job-id XML as the response.

All you need to write REST applications in perl is LWP. If you want to learn more about REST I'd browse the list of resources on the RESTwiki. Takes some effort to get into the mindset, but it can provide some elegant solutions.

Update: and, of course, the data doesn't have to be XML. Use whatever seems appropriate / more efficient.