Comer's book "Computer Networks And Internets" is a decent introduction. If you are on a budget get one edition out-of-date; it's a college textbook so the previous edition is usually about 90% cheaper than the current edition.
Once you feel comfortable with the basics, start going through RFCs for what you are interested in. In most cases the RFCs describe not just the protocol, but the reasoning behind it as well.
For me though, I need to do more than read; I need to be "hands on." One example:
When I was in college, I wanted to learn the IRC protocol better, and I noticed it was text based, so (after reading through a couple of RFCs) I connected to an IRC server with telnet in one window and the spec in another window. about 6 hours later I was finally able to connect and send messages. Those 6 hours were both less boring and more educational than 6 more hours of reading would be.
Net results on my grades was either slightly negative, or a wash; I probably missed 2 or 3 classes during those 6 hours, but I got nearly double the next highest score on my Networking midterm 3 semesters later.
I once did some research on networking and related stuff - it was not my primary topic, but I remember a few things.
During my reading, I found one of the best (as in readable) books was Doug Comer's "Internetworking With TCP/IP vol. 1" - an excellent theoretical reference. [1] However, skip the other volumes from Doug Comer (I think there are 3 volumes).
For writing practical applications, Richard Stevens' "Unix Network Programming" [2] is usually recommended, I didn't find it an easy read though. Perhaps others can pitch in.
For both the books suggested, getting a used old copy for cheap is a good idea because the core information was already there even in the first editions.
Finally, read up on PlanetLab [3]. Its a fascinating project - a small scale internet built on top of a subset of nodes contributed by universities and research organizations across the globe - that people can contribute to, and if you actually manage to get into the developer's list and make a contribution, you can quite honestly claim to have pushed the state of the art forward.
And lastly, be prepared to spend a good amount of time - I don't think it will be a fast or easy process. For whatever reason, I have found that the community around this work to be a little small, especially in comparison to how much it permeates pretty much everyone's life.
Once you feel comfortable with the basics, start going through RFCs for what you are interested in. In most cases the RFCs describe not just the protocol, but the reasoning behind it as well.
For me though, I need to do more than read; I need to be "hands on." One example:
When I was in college, I wanted to learn the IRC protocol better, and I noticed it was text based, so (after reading through a couple of RFCs) I connected to an IRC server with telnet in one window and the spec in another window. about 6 hours later I was finally able to connect and send messages. Those 6 hours were both less boring and more educational than 6 more hours of reading would be.
Net results on my grades was either slightly negative, or a wash; I probably missed 2 or 3 classes during those 6 hours, but I got nearly double the next highest score on my Networking midterm 3 semesters later.