Comparing external applications to plug-ins, we see that plug-ins solve the problem of integration. To a user, the plug-in will appear as just a part of the browser. As will be demonstrated later, it is possible, maybe with a little hassle, to create a plug-in from an already existing application.
When comparing writing plug-ins to rewriting browser source code, it is clear that plug-ins makes it possible to easily install several extensions to the same browser, without requiring any programming knowledge from the user. As more browsers support the same plug-in API, the user may depend on the browser of choice, instead of switching to a different browser to get the wanted behavior. Rewriting the browser gives one benefit over plug-ins: The programmer may incorporate features that are not (currently) available using the plug-in API, such as directly handling protocols not supported by the browser.
Java offers platform independency compared to plug-ins. In addition, a Java applet is downloaded and compiled automatically at runtime, thus freeing the users from doing manual installations. Also, the automatic download will guarantee that users always run the latest, and hopefully better, release of the viewer. However, as Java programs at the moment are run by interpreters, Java scores somewhat low when it comes to speed of non-trivial programs, compared to plug-ins and external applications. As on-the-fly Java byte code to machine code translators, and hardware Java Virtual Machines become widely available, slowness will no longer be a reason to stay away from Java.