Survey of Large Browser Uploads

Modern requirements for uploading files in a browser are few and simple:

• File size should not be limited by the uploading mechanism.
• There should be some progress feedback to the user.
• Large uploads should be resumeable.
• It should be possible to upload multiple files.
• Uploaded files should be integrity checked.

In fullfilling these requirements natively currently all browsers fall short. This is due to two things: 1) The nature of HTML(4|5); and 2) suboptimal browser implementations.

HTML4: Uploads

HTML4 file uploads was introduced over 15 years ago in rfc1867 and is still the de facto standard of uploading files. The following shows the HTML code to upload a  file:

<form method="POST" action="/" enctype="multipart/form-data">
  <input type="file" name="file" />
  <input type="submit" value="Upload" />
</form>

Obviously, it's not possible for the user to select multiple files. The number of files the form can upload is equal to the number of HTML inputs of type file. In addition, there is nothing in the standard when it comes to integrity checking or resuming uploads. When it comes to the size of files no limit is stated. Browsers, however, most likely fall short in uploading large files. Underneath, Table 1 shows which of the requirements are satisfied nativly by browsers.

Other tools available for the job - and why they all fall short!

• HTML5: From the outset HTML5 looks promising. HTML5 includes the File API. The File API makes it possible to read a file from the user filesystem and process it at the client. Looking into the File API, however, shows that there is no support for streaming files. Everything is loaded into memory!
  For more see the FileReader spec
• Java: Advanced Java applets are a pain both to develop and from a user perspective. Applets take forever to start-up, the user needs to confirm the signature of the applets that read from the filesystem, the GUI is OS inconsistent, and the plugin is not ubiquitously available. In addition, in order to work with HTTP uploads at a sensible abstraction level, you need some external library (e.g. Apache httpcomponents) increasing the size of the applet. Advanced, JavaScript to Java interaction is also difficult and buggy, see applets-missing-information-about-liveconnect-and-deployment)
• Flash: Flash outshines Java in three ways: 1) it has the highest world-wide penetration rate; 2) faster startup time than Java; and 3) needs no signing. However, Flash suffers mayorwise when it comes to uploading large files. In Flash file access is facilitated by FileReference. For some reason the size of files is stored internally as signed ints causing integer overflow when working with +2GB files! 
  
• Silverlight: I'm omitting Silverlight because of the low penetration rate of 58 %.

Features of the available tools are summarized in Table 2.

The only widespread tool where it's not plain impossible to upload +2GB files is Java! That fact led me to implement an open source Java applet uploading runtime for plupload. It's available at github, and a demo is running at jdams.org.