I just read Mark Cuban’s blog for the first time in awhile, and I like his fast and loose style, so don’t be surprised if my posts get a little less formal.
Moment’s after catching up with Mark, Mick from Zapr blogged about the new MySpace Zapr link tool. I quickly gave it a spin. At first, it blew me away. The link you ultimately used to download some of Henning Schulzrinne’s fascinating lecture slides on my machine looked like this:
I looked at that and scratched my head. I even shot off a quick e-mail to my good buddy Jerry Charumilind to figure out what I was missing here. I assumed the 18.104.22.168 was the IP address of the Starbucks I’m sitting in here in Sheriden Square, New York City, and that they had somehow figured out how to publicly address my machine using my locally-running Zapr instance to open up some ports. UPnP? No, that just wouldn’t work in most cases. Too brittle. Were they doing an ICE-style direct transfer, the way I would? Not possible without Zapr installed on both ends.
Then I turned to my trusty nslookup and discovered “22.214.171.124” is, in fact, one of their servers. I suspect they use the raw IP address to make it look like they’re doing something fancy, but, in fact, they’re just relaying traffic. Suddenly the world made sense again!
Don’t get me wrong, it’s still a nifty service and a nice implementation. It’s getting towards the seamless integration of the “dark web” I’d like to see — the integration of edge resources, in this case content. If they were using open protocols, I’d give them a thumbs up. Unfortunately, we can add them to the long list of services ignoring interoperability. If they are using standards, they’re certainly not advertising it. Aside from that, the main problem I see is their bandwidth costs if the service gets popular. Yikes! They’re not caching it, as your file immediately becomes unavailable if you go offline. This means not only will a user’s machine get hammered if something is popular, but so will Zapr’s. The user’s machine would just never hold up (think of the most popular YouTube videos hosted on your machine at home), and the Zapr servers would have a heck of a time too.
How do you get around this? Just like Skype does, just like Gizmo does, and just like LittleShoot does. Require users on both ends to have the software installed and pierce the NATs and firewalls to connect the computers directly. That solves the problem of killing the central server. What about the user’s poor machine? Keep track of all the locations for the file and load balance it across the Internet. How you do that is another question I’ll leave for another day (hint: not with BitTorrent).