Chromoting is out

A desktop remoting solution for upcoming Chrome OS (cleverly dubbed “chromoting”) has just been outed. Is it a myth or reality? If reality, how does it work?

How chromoting works

We don’t have to speculate. The first cut of chromoting code has landed in public Chrome source code tree and you can look at it yourself.

A look at its source code shows that remoting works exactly as you would expect it to work:

• on the host side, a process periodically captures the desktop as a bitmap at some interval (e.g. capturer_win.cc on Windows host)
• it calculates the difference between current bitmap and previous bitmap as a bunch of “dirty” rectangles. This is to minimize the amount of data sent to the client
• dirty rectangles are compressed further. Currently there’s a “no-op” encoder and vp8 encoder (encoder_vp8.cc)
• this data is sent to the client (i.e. a device running Chromium OS, although there’s no reason this functionality couldn’t be built directly into Chrome browser)
• the client decodes the data and displays it
• the communication is done using Google’s Jingle protocol, which is based on XMPP and already used in Google’s IM and video chat clients

As far as “remote desktop” solutions go this is the simplest thing that can possibly work and is very similar to what VNC does.

On Windows more sophisticated solutions exist that hook GDI calls, transmit the calls over the wire and reply them on the client. This is usually faster (you need to send less data) but much more complicated to implement, not cross-platform and possibly not as useful those days when Windows apps increasingly use non-GDI technologies like WPF.

The source code also has a beginning of standard browser NPP plugin) so it’s possible they’re planning a plugin for other browsers (like IE, FireFox and Safari) and native clients for each major operating system.

A side note on code reuse

Chromoting is a good example of code reuse. A communication layer and a good compression for images are hard problems, so it’s good that the project chose to re-use existing, mature technologies (Jingle and vp8). It also shows that even if code is open-source and liberally licensed, it’s beneficial for a corporation to own it (both jingle and vp8 code is (now) developed and evolved mainly by Google)

The future

The code is clearly in its early stages. Only Windows host is working (Mac and Linux parts are just stubs). There doesn’t seem to be support for capturing windows of a specific application (which might be a useful functionality) etc. But I’m sure it’ll evolve quickly.