Understanding IP, TCP, and HTTP

http://www.objc.io/issue-10/ip-tcp-http.html

by danieleggert • 12 years ago

237 61 12 years ago

rubiquity• 12 years ago

It's nice to see this recent increased emphasis on Web/mobile developers understanding the technologies that link it all together. The next thing I would add is a high level overview of the sockets API. While these topics aren't critical to most day-to-day lives of developers, they are certainly useful to understand.

marai2• 12 years ago

This is a very readable online book on networking and sockets: http://beej.us/guide/bgnet/output/html/multipage/index.html

Talk about understanding the sockets API ;-) here's the content section for chapter 5:

  5.1. getaddrinfo() — Prepare to launch!
  5.2. socket() — Get the File Descriptor!
  5.3. bind() — What port am I on?
  5.4. connect() — Hey, you!
  5.5. listen() — Will somebody please call me?

stusmall2• 12 years ago

I love beej's guide! Its where I learned socket programming. It is an art to make such a dry subject as entertaining as he does.

jnbiche• 12 years ago

Beej is where I learned socket programming. A great guide.

dnc• 12 years ago

Beej's guide is great! Thanks!

nasalgoat• 12 years ago

I had one developer with over 6 years of server-side experience who made a server/client setup where the server would open a connection to the client, pass connection info to the client, close the connection, and then have the client open a connection back to the server to return results.

When I explained how TCP worked, in that the client could connect to the server and maintain an open socket to pass info continuously he was blown away. He had no idea this was possible. Explaining UDP was a lot harder.

So, I welcome any education on basic TCP/IP functionality!

zAy0LfpBZLC8mAC10• 12 years ago

How I love it when people without deep knowledge of some subject write authoritative sounding articles.

Without guarantee of completeness, to avoid the spread of misinformation:

- IPv6 fragmentation has nothing to do with some "minimum payload size" (whatever that is) - there simply is no fragmentation being done by routers, the sender still can fragment however it pleases, and presumably will do so whenever it has to send a packet that doesn't fit through the path MTU.

- The end points use Packet Too big ICMP6 messages to determine _path_ MTU, which is different from just "the MTU".

- With IPv4, the sender chooses whether a router will fragment when the packet exceeds the next-hop MTU or whether the router should drop the packet and send a Fragmentation Needed ICMP message - where the latter again is used for path MTU discovery.

- Path MTU discovery is useful because it allows the sending IP implementation to push the chunking higher up the stack when the sending higher-level protocol has the capability (as is the case with TCP, but not with UDP, for example), which tends to produce lower overhead. Unfortunately, some clueless firewall administrators, such as those responsible for AWS EC2, do filter all ICMP because they for unknown reasons consider it to be bad, thus breaking PMTUD, which can lead to hanging TCP connections.

- TCP sequence numbers are for bytes, always, with the special case of SYN and FIN also counting as "bytes" in the sequence, but never for segments.

the_cat_kittles• 12 years ago

Just out of curiosity, what do you do? Is this knowledge germane to where you work? I've just recently become interested in this stuff, so I'm curious to get a lay of the land.

zAy0LfpBZLC8mAC10• 12 years ago

I do ... software development? ;-) There isn't really any particular category for what I do, though I tend to do more of the lower-level/backend stuff of projects, and knowing how the stuff that you build on works internally certainly is useful in optimizing and debugging.

As for getting an understanding of how TCP/IP works, I think Stevens' classic TCP/IP Illustrated still is a good book to get started, even if somewhat dated in some details (no IPv6, in particular), but the general principles still apply. Though maybe there are newer equally good books around that I just don't know about?

mhurron2• 12 years ago

"minimum payload size" (whatever that is)

I'll give them the benefit of the doubt and say he got his terms wrong. The IPv6 RFC states that IPv6 requires a minimum MTU of 1280 bytes. I guess that's what he meant.

https://www.ietf.org/rfc/rfc2460.txt

Packet Size Issues

IPv6 requires that every link in the internet have an MTU of 1280 octets or greater. On any link that cannot convey a 1280-octet packet in one piece, link-specific fragmentation and reassembly must be provided at a layer below IPv6.

zAy0LfpBZLC8mAC10• 12 years ago

Which wouldn't really make me any more confident in the reliability of the whole thing?! Confusing lower-level fragmentation and reassembly with IPv6 fragmentation is not exactly a mistake you'd be likely to make when you understand what that actually means, I would think.

unethical_ban• 12 years ago

I love it when people without deep knowledge of a subject try to learn about it and explain themselves to others.

stusmall2• 12 years ago

This is an important part of the way I learn. I will read something and then explain it to someone else. It makes me think deeper about the issue as I form the words and it gives me a great chance to get corrected when I am making unfair assumptions. I always preface this conversations with "as I understand it" or "from what I read" or some other disclaimer. I used to have a coworker who would give me soooo much guff about these disclaimers since I'd drop so many of them in one of these conversations. I just felt it was important to make it clear I wasn't coming from a place of authority and more from the perspective of a guy who is bumbling through it and trying to figure out what the hell is going on.

smikhanov• 12 years ago

When it comes to complicated subjects like internals of TCP/IP, they may make more harm than good. Volume 1 of "TCP/IP Illustrated" is, I think, 700 pages long, not without reason. When one writes condensed articles like this it makes sense to stay high-level, because the moment you get into discussing SYN/ACK handshake, you are in danger of leaving large gaps in your explanation or making unreasonable stretches to complete the picture.

On a side note, it's funny to see that most of the paragraphs in the original article end with a link to Wikipedia — as a reader, I can go there myself, what good those articles do to me?

cbsmith2• 12 years ago

I'm also going to cut him some slack with the bit about giving each segment a unique number. While formally the sequence number identifies each byte of data, it really is about providing heuristics to identify and correct for out of order, fragmented, missing & duplicate segments. It is important that it be about bytes, particularly for things like SACK, but if you are trying to simplify things you might describe it as being about the segments.

zAy0LfpBZLC8mAC10• 12 years ago

Sure, nothing wrong with simplifying things, but "Both ends are sending sackOK. This will enable Selective Acknowledgement. It switches the sequence numbers and acknowledgment number to use byte range instead of TCP segment numbers." is just flat-out wrong, and in particular suggests that "numbered segments" is not a simplification but an actual fact about how the thing works.

cbsmith2• 12 years ago

And regarding dropping the ICMP message about fragmentation... good firewall implementations have the firewall discover the MTU behind it and express THAT.. even better they might hide the hops behind it.

zAy0LfpBZLC8mAC10• 12 years ago

There is no such thing as an "MTU behind it", there is a separate path MTU for each and every ordered address pair, more or less (and that's not even static, obviously).

And obviously I was talking about packet filters, not about some kind of application firewall, which obviously doesn't have anything to do with filtering of packets anyhow.

shadesandcolour• 12 years ago

From the editorial page of this issue:

"We’ve created a new public repository on GitHub that contains all current and past objc.io articles. If you find any mistakes or have suggestions for improvements, please don’t hesitate to file issues, or even better: submit a pull request!"

Make a pull request so that people like me can learn about networking too.

jakejake• 12 years ago

Anybody writing routing code would be foolish to use this, or any other "simplified" article as a protocol reference. But, I'll agree that it is presented in such a way (and with enough technical detail) that any technical errors should be corrected.

If I had deep knowledge in this area, I'd probably applaud the effort and send corrections, rather than criticize.

danieleggert2• 12 years ago

Thanks for taking the time to comment on the article. I'll update the article later today to correct these. Your help is appreciated.

Myrmornis• 12 years ago

What an unpleasant attitude. The author clearly made a lot of effort and as far as I know it's all accurate. If you know better then say so pleasantly.

norswap• 12 years ago

If you want to further your understanding of network protocols, there's an excellent open textbook available here: http://cnp3book.info.ucl.ac.be/

dd3141• 12 years ago

Wow...what a great resource! I just downloaded it and will start reading it tonight. Thanks!

yeukhon2• 12 years ago

Also https://www.coursera.org/course/comnetworks is helpful. I use this to prepare interview, just so I don't have to dig my textbook.

maaaats• 12 years ago

At Uni we had a book called "Computer Networking - A top down approach". One of the best teaching books I've ever read. The amount of detail is very nice balanced, and as the title says it's a top down approach where one layer at a time is discussed. Very interesting.

greyskull• 12 years ago

We're using that same book, sixth edition, in my networks course right now. It's overly verbose like every other textbook, but the content is solid.

tshaddox• 12 years ago

I've heard that a good way to gauge a person's general technological literacy is to simply ask "what happens when I type a URL in a browser and hit Enter?" Obviously, the question is deliberately open-ended, and any step in the process can be broken down into more detailed steps (up to a point). I'd like to see an article that initially shows high-level steps (e.g. DNS request, HTTP request, server processing, HTTP response, parsing and rendering), but allows each step to be expanded progressively with increasing detail.

pbhjpbhj• 12 years ago

[Deeper]

The Pauli exclusion principle prevents electrons with the same quantum characteristics from entering the same space, this interaction occurs across the amassed copper atoms which form the majority of the metallic wires that approach one another in the internal structure of the keyboard ...

mrtbld• 12 years ago

There’s a misconception that restarting the (HTTP) request will fix the problem. That is not the case. Again, TCP will resend those packets that need resending on its own.

But that's not true if the connection is interrupted at the socket level, right?

For example, if the device switches from 3G to Wi-Fi, or from Wi-Fi to wire, then I believe, its hardware address changes, its IP address changes and the socket becomes stale. But the TCP connection, would it be closed right away or would it hang until some timeout? (And does it depend on the OS?)

zAy0LfpBZLC8mAC10• 12 years ago

The layers are conceptually independent, and in a way even the concept of "switches from 3G to WiFi" is a misconception.

The TCP socket doesn't know anything about any "interfaces" or "links" or anything like that, it only knows about its and the remote IP address (and port), and the IP stack will deliver any packets to it that it receives that are addressed to that port on that address coming from the corresponding remote address and port, no matter which link it was received through (possibly subject to reverse path filtering on end hosts as a security measure). Similarly, each outbound packet is routed independently, so if the routing table changes half-way through a TCP connection, packets simply will be routed via a different link (the end host really just does the same as any other router does, and the fundamental idea of packet switched networks is that routers to not know about connections, they simply forward each packet independently, potentially switching links as needed at any time).

It perfectly possible, for example, to bridge between WiFi and wired ethernet, and have a gateway that routes some IP network onto that Ethernet/WiFi, then, while connected to the WiFi, establish a TCP connection, disconnect from the WiFi, connect to the Ethernet via cable, using the same IP addresss on the Ethernet interface as you previously used on the WiFi interface, and the TCP connection will survive that just fine (it might take a moment for the router to time out its neighbour cache entry and re-resolve your IP address into the new hardware address, but that's just a matter of a few seconds). You could even connect to both, configure things such that the kernel only replied to ARP/ICMP6 ND on WiFi, say, and route outbound packets through the cable, then the outbound packets of the TCP connection would go through the cable while the inbound packets would go through the air.

The only thing that actually breaks a connection is when packets addressed to the address that your TCP connection is using cannot reach you anymore, or when packets you send using that address can not reach the other side anymore, for example because you send them through a link that does not allow you to use that address. The latter really is mostly what kills TCP connections on mobile phones: the default route gets changed from WiFi to G3, say, and your mobile provider won't allow you to continue sending through their network packets using the address you got assigned by the WiFi - so the connection hangs even if the WiFi interface might actually still be up and able to receive packets addressed to that address.

One important thing to notice in this: There isn't really any way how a TCP implementation could detect right away that any of this has happened, as it cannot know what the filtering policies of your provider(s)/network(s) are or whether you disconnected only temporarily or whether you will reconnect to a different access point to the same network ... - so, when some mobile platform kills TCP sockets when you "change from 3G to WiFi" that really is a dirty hack that makes a load of assumptions about some typical setups that don't necessarily hold true.

teddyh• 12 years ago

I prefer The Unix and Internet Fundamentals HOWTO:

http://en.tldp.org/HOWTO/Unix-and-Internet-Fundamentals-HOWT...

brudgers• 12 years ago

David Wetherall teaches this course @ Coursera.

https://www.coursera.org/course/comnetworks

He pretty much wrote the book.

laichzeit0• 12 years ago

If you want to learn about IP, TCP, UDP and some of the protocols below this I would highly recommend reading Richard Stevens book TCP/IP Illustrated, Volume 1: The Protocols.

For two reasons: It's probably one of the best introductions to the subject that has ever been written, and it's a model example of how a technical book should be written.

I'd be hard pressed to find a reason not to go this route at least once in your life. I know the material pretty well but I still re-read Stevens books every few years just because it is so good.

beggi• 12 years ago

TCP/IP Illustrated books are super detailed. If you don't want to dive quite that deep I recommend "Computer Networking: A Top-Down Approach Featuring the Internet".

roel_v• 12 years ago

"I'd be hard pressed to find a reason not to go this route at least once in your life. I know the material pretty well but I still re-read Stevens books every few years just because it is so good."

Then again, that's a lot of effort to spend on something that the vast majority of us don't need to know in much detail. The main reasons for knowing all the details are

- to write a new networking stack, or working on an existing one;

- to write or maintain server software or routers or caches or other software directly involved in networking;

- to break or exploit existing software.

(obviously 'because it's interesting' is a valid, but not practical reason to know)

mje__• 12 years ago

If you write anything that communicates over a network (e.g. anything using HTTP), you need to know at least some of this stuff, otherwise you're not going to be able to explain why (for example) your service call latencies have a big spike around 200ms.

jokoon• 12 years ago

I'm still curious about an explanation why do we have both TCP and UDP.

For example if you do peer to peer, you need low latency, and UDP is best for that.

I think it's because TCP is hardware optimized, but it's designed to transmit a file in a stream, so if a packet is corrupt, it just waits to send that packet. In that fashion, TCP tend to be slower, but on average it's more efficient for single files or webpages.

You don't have good granularity with TCP, but if you want to work with UDP, you need to add redundancy and other mechanisms to make sure all is good.

ENet is an example of using UDP for gaming, so the goal is to have the lowest latency possible.

zAy0LfpBZLC8mAC10• 12 years ago

Bittorrent is also peer to peer, and it doesn't need low latency. Really, it's about latency, nothing to do with peer to peer.

TCP has head-of-queue blocking, as it guarantees complete and in-order delivery, so when a packet gets lost in transit, it has to wait for a retransmit of the missing packet, whereas UDP delivers packets to the application as they arrive, including duplicates and without any guarantee that a packet arrives at all or which order they arrive (it really is essentially IP with port numbers and an (optional) payload checksum added), but that is fine for telephony, for example, where it usually simply doesn't matter when a few milliseconds of audio are missing, but delay is very annoying, so you don't bother with retransmits, you just drop any duplicates, sort reordered packets into the right order for a few hundred milliseconds of jitter buffer, and if packets don't show up in time or at all, they are simply skipped, possible interpolated where supported by the codec.

Also, a major part of TCP is flow control, to make sure you get as much througput as possible, but without overloading the network (which is kinda redundant, as an overloaded network will drop your packets, which means you'd have to do retransmits, which hurts throughput), UDP doesn't have any of that - which makes sense for applications like telephony, as telephony with a given codec needs a certain amount of bandwidth, you can not "slow it down", and additional bandwidth also doesn't make the call go faster.

In addition to realtime/low latency applications, UDP makes sense for really small transactions, such as DNS lookups, simply because it doesn't have the TCP connection establishment and teardown overhead, both in terms of latency and in terms of bandwidth use. If your request is smaller than a typical MTU and the repsonse probably is, too, you can be done in one roundtrip, with no need to keep any state at the server, and flow control als ordering and all that probably isn't particularly useful for such uses either.

And then, you can use UDP to build your own TCP replacements, of course, but it's probably not a good idea without some deep understanding of network dynamics, modern TCP algorithms are pretty sophisticated.

Also, I guess it should be mentioned that there is more than UDP and TCP, such as SCTP and DCCP. The only problem currently is that the (IPv4) internet is full of NAT gateways which make it impossible to use protocols other than UDP and TCP in end-user applications.

bachback• 12 years ago

you made some very interesting posts. what I'm missing from this is 0mq. I believe it introduces layering mechanisms, so that one can re-use patterns to build cool stuff (anything really), without knowing specific details. do you have a email where I can reach you?

sajal83• 12 years ago

The improvements of using HTTP pipelining can be quite dramatic over high-latency connections – which is what you have when your iPhone is not on Wi-Fi. In fact, there’s been some research that suggests that there’s no additional performance benefit to using SPDY over HTTP pipelining on mobile networks

Excellent summary but i think pipeline has been oversimplified. HTTP pipelining is a FIFO queue. The responses have to be delivered in the same order as the requests. So if the first(or an early) response took longer to generate, all other requests in the pipeline have to wait. Something that SPDY is not susceptible to.

notfoss• 12 years ago

There's a minor typo below the HTTPS section. It should be TLS not TSL ;)

Edit: By the way, it was a nice article. I especially liked the tcpdump explanation.

Understanding IP, TCP, and HTTP

Related Stories