What aspects of networking should I study as a programmer?

Question

I have worked/played as a software developer for years and have focused on programming almost exclusively (mostly OO software, design and methodology). I feel quite sure on "higher-level" networking concepts (like HTTP and SSL, for instance) but I've never studied -- at least not deeply -- lower-level networking concepts. Are they relevant to me as a programmer? If so, what should I study?

Any suggestions on books, resources, whatever, would be appreciated. I would appreciate something for programmers, as basic Computer Networking resources might bore me to tears (or should I get over it and start reading?)...

Note: This is a remix of this closed question.

Answer 1

In this Webby world, you need to know how TCP/IP works as it includes HTTP. Back in the day I read TCP/IP Illustrated, Volume 1: The Protocols. Maybe it's dated. Maybe it's a classic. I certainly found it readable.

At uni (college for you American types) I read an earlier version of Internetworking Protocols and hated it (it's so dry).

The TCP/IP Guide: A Comprehensive, Illustrated Internet Protocols Reference seems to be highly rated on Amazon but I've never read it so can't comment.

Your networking education should include at least basic network programming, which should involve knowing the TCP handshake and how to write a basic TCP client and server.

Answer 2

Are you sure that link is relevant? Hehe... well...

Why would you argue that the low-level details are not relevant to you as a programmer? What kind of programmer do you want to become is really a more suitable rephrasing. I'm drawing on experience from my last place of work were we had two kinds of programmers. Great ones and good ones. The great programmers knew what was going on and the good ones often failed to recognize that the inner workings of the system did something they didn't expect.

My point is this, a deeper understanding will give you the insight you need when programming on a higher level. You should know this to realize what you sacrifice when you chose to program at a higher level.

There are some exceptions to the above statement but I'm the kind of programmer who wants to do it all and I therefore advocate knowledge is power.

I've spent a lot of time learning the inner workings of TCP, IP and integer-hackery is this what you had in mind or do you wanna go further and talk about the hardware to?

Answer 3

Whether or not networks is relevant to you is entirely dependent on your job and interests. For most programmers the answer is probably no - learning the low-level intricacies of networks is not particularly relevant to a programmer. It can also be excruciatingly boring unless you happen to be involved in it directly, so I would not recommend losing any sleep over it.

At university we studied this book: Computer Networks by Andrew Tanenbaum which I remember was reasonably enjoyable, and is supposedly a classic, but I imagine it is rapidly becoming outdated.

You should have some experience in basic networking but this is best learned hands on through solving problems. Setting up a home network is a good start, then move on to an office network if you can. Even a Dummy's guide would be fine for this sort of work. Help out with your office tech support if you have the opportunity.

Answer 4

I agree with Cleatus, TCP/IP Illustrated is a great book. Understanding TCP and IP/UDP helps you even if your only programming "higher leveL" protocols like HTTP. Interesting topics include:

• Nagel algorithm
• Port reuse policy
• Packet fragmentation
• IP Multicast
• ARP (Address Resolution Protocol) and ARP refresh

In general, most network programming toolkits are pretty "leaky abstractions". If things just worked right all the time, you could ignore TCP / IP. But stuff breaks frequently and thats when understanding things like port reuse, subnets, vlans, tcp buffer sizes, DNS propagation, obscure WinSock errors (on windows) helps.

Answer 5

feel quite sure on "higher-level" networking concepts (like HTTP and SSL, for instance)

There are plenty of higher levels of abstraction available (e.g. message-based systems like MSMQ and WebSphere*MQ) to reclassify HTTP/SSL as mid- if not low-level. Which leads to suggesting a different approach.

Work to understand why the "Fallacies of Distributed Computing" are fallacies, and how that should impact architecture and design of distributed systems.

Answer 6

My suggestion related to network programming is make sure you learn/know Pthreads. There seems to be a lot of people coming out of school these days claiming to have studied "networking" but all they've ever done is fork(). I consider threading and sockets to be intimately connected. Maybe you're looking for the more concept type suggestions, like TCP/IP books as mentioned, but that's my two cents. Lots of good pthreads and BSD sockets references on the web, probably better than any book I could suggest.

Answer 7

I found TCP/IP Clearly Explained to be an excellent overview of TCP/IP and other related network protocols. It's one of those books that you can just sit down and read without having to trudge through academia style wording.