On Wireless Data

It's easy to look around at all of the "smart phones," iPads, wireless modems, and think that the future is here, or even that we're living on the cusp of a new technological moment. While wireless data is amazing particularly with respect to where it was a few years ago--enhanced by a better understanding of how to make use of wireless data--it is also true that we're not there yet.

And maybe, given a few years, we'll get there. But it'll be a while. The problem is that too much of the way we use the Internet these days assumes high quality connections to the network. Wireless connections are low quality regardless of speed, in that latency is high and dropped packets are common. While some measures can be taken to speed up the transmission of data once connections are established, and this can give the illusion of better quality, the effect is mostly illusory.

Indeed in a lot of ways the largest recent advancements in wireless technology have been with how applications and platforms are designed in the wireless context rather than anything to do with the wireless transmission technology. Much of the development in the wireless space in the last two or three years has revolved around making a little bit of data go a long way, in using the (remarkably powerful) devices for more of the application's work, and in figuring out how to cache some data for "offline use," when it's difficult to use the radio. These are problems that can be addressed and largely solved in software, although there are limitations and inconsistencies in approach that continue to affect user experience.

We, as a result, have a couple of conditions. First that we can transmit a lot of data over the air without much trouble, but data integrity and latency (speed) are things we may have to give up on. Second that application development paradigms that can take advantage of this will succeed. Furthermore, I think it's fairly safe to say that in the future, successful mobile technology will develop in this direction as opposed against these trends. Actual real-time mobile technology is dead in the water, although I think some simulated real-time communication works quite well in these contexts.

Practically this means, applications that tap an APO for data that is mostly processed locally. Queue-compatible message passing systems that don't require persistent connections. Software and protocols that assume you're always "on-line" and are able to store transmissions gracefully until you come out of the subway or get off of a train. Of course, this also means designing applications and systems that are efficient with regards to their use of data will be more successful.

The notion that fewer transmissions that consist of bigger "globs" of data will yield better performance than a large number of very small intermediate transmissions, is terribly foreign. It shouldn't be, this stuff has been around for a while, but nevertheless here we are.

Isn't the future grand?

comments powered by Disqus