The Lesson of FSP

Once upon a time, there was a low bandwidth Internet link under a very wide ocean between the Center of the Internet, and the land of OZ. That being a time of Telecommunications Monopolies that were part of national governments, and being that the ocean was very, very wide, the link in question was rather expensive.

Somewhere in the land of the Center, there was an FTP server with some bags of bits (more commonly known as "files") on it that were of particular interest to some users in the land of OZ. Naturally, since the purpose of the Internet is the mutual exchange of bags of bits to the advantage of all, the users in the land of OZ FTP'd these bags of bits from the land of the Center, down the very expensive trans-oceanic link, to their computers in the land of OZ.

A Problem, or two

There arose two problems because of this activity. The first was that since these bits were of a somewhat taboo nature, the users did not generally share copies with each other in the land of OZ (they were, perhaps, just a tad embarassed), so they each individually FTP'd the bits down the expensive trans-oceanic link to their computers. This was, perhaps, a less than perfectly efficient use of this expensive trans-oceanic link. In fact, it largely consumed the bandwith of the expensive trans-oceanic link.

The second problem was that a very morally upright network administrator at the site where the expensive trans-oceanic link connected to The Center became aware of this less than perfectly efficient use of the link, and the taboo nature of the bits being transferred. After a nanosecond's reflection, he decided to block access from OZ to this FTP server with taboo bits on it in the land of the Center of the Internet at the choke point he controlled. All this to make the bandwidth of the expensive trans-oceanic link available for more moral purposes.

"... and that is all there is to that," he no doubt thought.

Uh, not quite...

Unfortunately for the morally upright network administrator, the desire of the users in the land of OZ for these taboo bits (and the determination of the administrator of the FTP server in the land of the Center to supply them) was such that they decided to invent a file transfer protocol that could not be blocked by the morally upright administrator at the choke point.

Enter FSP

They invented a protocol that did not obey the usual rules for behavoir of Internet Protocols. It didn't use the same port number throughout the life of the "connection"; it randomized them, much like Spread Spectrum radio. It didn't use TCP at all; it used UDP. It was impossible to block, using the filtering technology of the time. They called it File Service Protocol (FSP).

The first implementation of FSP was a terrible file transfer protocol. It had no ability to "back off" and share a link with other protocols trying to move bits along (TCP does this by design). This is bad because the entire Internet depends on all protocols being nice and equally sharing the bandwidth by dynamically adjusting what they use. However, FSP works, for some value of "works".

Endgame

Using FSP, the users got their taboo bits. The expensive trans-oceanic link filled up again, somewhat more persistently because of the aggressive nature of FSP. The morally upright network administrator was defeated. The expensive trans-oceanic link would just have to be upgraded to handle all the uses that the users wanted to put it to.

This could have gone another round or two, but the morally upright network administrator was smart enough to recognize that he would lose this contest of wills in the long run. So, he quit fighting it, having made his stand.

So, What's the Lesson?

The moral of this story is best summed up by a quotation:

"The Internet regards censorship as damage, and routes around it." - John Gilmore

Users will use whatever means are at their disposal to do whatever they want to do, and there is rarely a technological way to stop them. Moreover, a technological attempt to stop them will usually result in even more perverse behavoir than previously observed.


Erik Fair <fair@clock.org>
September 1, 2000