If you never used Gnutella, you couldn't care less. If you're a Napster user and you've been waiting to switch until the service gets shut down, it may be too late. In fact, too many users may be the basic problem. A few months ago, during the time of Napster's pending demise, a rush of traffic on the Gnutella network overwhelmed the system.
Since then, traffic has dropped to pre-scare levels. We've learned a bit from that rush -- and it's not good news.
Here's the scoop: Gnutella is a peer-to-peer file-sharing program. It is designed for people to share and trade files across the Internet without using centralized servers. Essentially, it should operate the same way as Napster, but with major advantages. Gnutella has the capacity to transmit any type of file, not just music, and because no one "owns" the program (it is now open-source), it can't be legally scrutinized à la Napster.
Unfortunately, we have found that Gnutella is not as scalable as the centralized Napster network. Translation: the more users, the less efficient. In recent weeks, doing a search or query with the program yielded little or no results. And it often timed out when it looked for files. Something has gone wrong, and a few theories have emerged about what is happening.
There is a widespread belief that searches that produce no results are due to bottlenecks on the Gnutella network. These bottlenecks are being caused by a couple of reasons. First, there is bad code writing in the client software.
Now, this is hardly a fault of the programmers, as the code was still in beta version when it was released (and then subsequently banned from AOL). But essentially, if user A makes a request for a file from user B, who is offline, the software sends a "push" packet broadcast to all the other computers connected to user A instead of routing it back to where it came from. This lack of routing and pushing when the host is offline contributes to more than 50 percent of the total traffic on bad days.
Even if the code were solid and didn't waste bandwidth, 56k-modem users wouldn't be able to send multiple search requests quick enough to keep up with the amount of people online. When too many users do searches simultaneously, traffic backs up. And since 56k-modem users are also nodes (or other peers connected to the network) connecting hundreds of other users, that arm of the network is essential cut off.
One other issue not addressed with the current Gnutella protocol is how Web-based search engines are delivering a lot of traffic to the network but are not sharing any files in return. Everyone who is using these services is essentially leaching the system and hogging bandwidth.
Of course, all these issues are being addressed in some fashion or another, but interim solutions can carry us only so far. Developers like Sebastien Lambla, the team manager of the general Purpose Location Protocol, gPulp, are gathering and writing new proposals at gnutellaNG.wego.com/ for an overhauled system. The team is looking for support and resources from the W3C, IETF, and big players such as Intel and Cisco as well. Sebastien believes that the gPulp protocol will find any type of data on the network -- servers, web addresses, wireless devices, and even contact information. And while the new protocol won't transmit data, it will essentially complement the current infrastructure by connecting all programs that look for anything on the Internet.
In the meantime, people who are looking to download music from legally questionable sources will have to stick with Napster, Scour, or Napigator. We'll just have to wait and see if Gnutella's kinks get worked out or if it's time to call a dead horse dead.
