Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

experiments to test the assertions made by REST and ROA

branch: master
README.rst

REST vs. The World

Modern web applications have a number of requirements that both developers need and users need. These experiments will compare solutions presented by REST proponents versus current solutions.

Whenever the status quo is challenged, the burden of proof is on the challenger.

Each solution presented effects the entire system. I will start with small tests and create new tests based of the results

Phase One

At the very least, a web application is a data collator. The data collated may have varying degrees of latency. When fetch a certain piece of information results in an unacceptable latency, web developers rely on caching to decrease tho wait on the user.

I will start with data collation first and base caching experiments and the collation tests.

Data Collation

Traditional methods of collating data is accomplished through direct access to the data. The response of an URI may fetch some data from a database, some other data from a file, etc. Generally the data is accessed directly.

Within a RESTful web application, there is the potential to access data indirectly using URIs for Resources. The data collation experiment will explore which of these techniques are the fastest

Caching

The favored method of caching is accomplished by caching data structured from within the application.

RESTful caching favors HTTP/1.1 caching utilizing external cache proxies and/or relying on client caching.

This experiment expands on the data collation experiment and adds caching to the individual items.

The RESTful resources are cached using HTTP caching and the alternatives use internal data caching

Something went wrong with that request. Please try again.