The basic concept of copyright law is that an author is the only one who is allowed to make copies of her work, and only she can give others permission to do so as well. At a technical level, sending a response to a request involves telling another machine to pass a message along for you, so there is at least implied consent to copy it and send it to another user. However, what are the bounds of this consent? Can a router store the data it has been passed? For how long? Can it serve it to others besides the IP address the response was intended for?

There may be case law and/or actual laws that clarify these points, but I am not a lawyer. I presume you aren't either. If you think the concept of IP law has a straightforward and indisputable application to the internet that clears all questions about what routers can and can't do with the data they are passed, feel free to explain.

It seems like you're operating within the "lots of people do it, so it must be legal somehow" school of thought when it comes to routing. This isn't necessarily a problem as long as it's applied consistently. Several other commenters and I effectively made the same argument in saying that Google does almost the same thing Scribd does in terms of copying and redistributing content. It isn't possible for you to call that argument invalid then rely on that argument as proof of routing's obvious legality.

Scribd is legally indistinguishable from a caching proxy. Feel free to let Opera and all the other caching proxy operators know.

http://www.opera.com/business/solutions/turbo/