1.1. Purpose of this document

This document is intended to help Linux and Internet users who are learning by doing. While this is a great way to acquire specific skills, sometimes it leaves peculiar gaps in one's knowledge of the basics -- gaps which can make it hard to think creatively or troubleshoot effectively, from lack of a good mental model of what is really going on.

I'll try to describe in clear, simple language how it all works. The presentation will be tuned for people using Unix or Linux on PC-class hardware. Nevertheless, I'll usually refer simply to `Unix' here, as most of what I will describe is constant across platforms and across Unix variants.

I'm going to assume you're using an Intel PC. The details differ slightly if you're running an Alpha or PowerPC or some other Unix box, but the basic concepts are the same.

I won't repeat things, so you'll have to pay attention, but that also means you'll learn from every word you read. It's a good idea to just skim when you first read this; you should come back and reread it a few times after you've digested what you have learned.

This is an evolving document. I intend to keep adding sections in response to user feedback, so you should come back and review it periodically.

1.2. New versions of this document

New versions of the Unix and Internet Fundamentals HOWTO will be periodically posted to comp.os.linux.help and comp.os.linux.announce and news.answers. They will also be uploaded to various Linux WWW and FTP sites, including the LDP home page.

You can view the latest version of this on the World Wide Web via the URL http://www.linuxdoc.org/HOWTO/Unix-and-Internet-Fundamentals-HOWTO/index.html.

This document has been translated into Polish.

1.3. Feedback and corrections

If you have questions or comments about this document, please feel free to mail Eric S. Raymond, at esr@thyrsus.com. I welcome any suggestions or criticisms. I especially welcome hyperlinks to more detailed explanations of individual concepts. If you find a mistake with this document, please let me know so I can correct it in the next version. Thanks.

1.4. Related resources

If you're reading this in order to learn how to hack, you should also read the How To Become A Hacker FAQ. It has links to some other useful resources.

8.1. Virtual memory: the simple version

Efficiency is important, because memory is expensive. Sometimes you don't have enough to hold the entirety of all the programs the machine is running, especially if you are using a large program like an X server. To get around this, Unix uses a technique called virtual memory. It doesn't try to hold all the code and data for a process in memory. Instead, it keeps around only a relatively small working set; the rest of the process's state is left in a special swap space area on your hard disk.

Note that in the past, that "Sometimes" last paragraph ago was "Almost always" -- the size of memory was typically small relative to the size of running programs, so swapping was frequent. Memory is far less expensive nowadays and even low-end machines have quite a lot of it. On modern single-user machines with 64MB of memory and up, it's possible to run X and a typical mix of jobs without ever swapping after they're initially loded into core.

8.2. Virtual memory: the detailed version

Actually, the last section oversimplified things a bit. Yes, programs see most of your memory as one big flat bank of addresses bigger than physical memory, and disk swapping is used to maintain that illusion. But your hardware actually has no fewer than five different kinds of memory in it, and the differences between them can matter a good deal when programs have to be tuned for maximum speed. To really understand what goes on in your machine, you should learn how all of them work.

The five kinds of memory are these: processor registers, internal (or on-chip) cache, external (or off-chip) cache, main memory, and disk. And the reason there are so many kinds is simple: speed costs money. I have listed these kinds of memory in decreasing order of access time and increasing order of cost. Register memory is the fastest and most expensive and can be random-accessed about a billion times a second, while disk is the slowest and cheapest and can do about 100 random accesses a second.

Here's a full list reflecting early-2000 speeds for a typical desktop machine. While speed and capacity will go up and prices will drop, you can expect these ratios to remain fairly constant -- and it's those ratios that shape the memory hierarchy.

We can't build everything out of the fastest kinds of memory. It would be way too expensive -- and even if it weren't, fast memory is volatile. That is, it loses its marbles when the power goes off. Thus, computers have to have hard disks or other kinds of non-volatile storage that retains data when the power goes off. And there's a huge mismatch between the speed of processors and the speed of disks. The middle three levels of the memory hierarchy (internal cache, external cache, and main memory) basically exist to bridge that gap.

Linux and other Unixes have a feature called virtual memory. What this means is that the operating system behaves as though it has much more main memory than it actually does. Your actual physical main memory behaves like a set of windows or caches on a much larger "virtual" memory space, most of which at any given time is actually stored on disk in a special zone called the swap area. Out of sight of user programs, the OS is moving blocks of data (called "pages") between memory and disk to maintain this illusion. The end result is that your virtual memory is much larger but not too much slower than real memory.

How much slower virtual memory is than physical depends on how well the operating system's swapping algorithms match the way your programs use virtual memory. Fortunately, memory reads and writes that are close together in time also tend to cluster in memory space. This tendency is called locality, or more formally locality of reference -- and it's a good thing. If memory references jumped around virtual space at random, you'd typically have to do a disk read and write for each new reference and virtual memory would be as slow as a disk. But because programs do actually exhibit strong locality, your operating system can do relatively few swaps per reference.

It's been found by experience that the most effective method for a broad class of memory-usage patterns is very simple; it's called LRU or the "least recently used" algorithm. The virtual-memory system grabs disk blocks into its working set as it needs them. When it runs out of physical memory for the working set, it dumps the least-recently-used block. All Unixes, and most other virtual-memory operating systems, use minor variations on LRU.

Virtual memory is the first link in the bridge between disk and processor speeds. It's explicitly managed by the OS. But there is still a major gap between the speed of physical main memory and the speed at which a processor can access its register memory. The external and internal caches address this, using a technique similar to virtual memory as I've described it.

Just as the physical main memory behaves like a set of windows or caches on the disk's swap area, the external cache acts as windows on main memory. External cache is faster (250M accesses per sec, rather than 100M) and smaller. The hardware (specifically, your computer's memory controller) does the LRU thing in the external cache on blocks of data fetched from the main memory. For historical reasons, the unit of cache swapping is called a "line" rather than a page.

But we're not done. The internal cache gives us the final step-up in effective speed by caching portions of the external cache. It is faster and smaller yet -- in fact, it lives right on the processor chip.

If you want to make your programs really fast, it's useful to know these details. Your programs get faster when they have stronger locality, because that makes the caching work better. The easiest way to make programs fast is therefore to make them small. If a program isn't slowed down by lots of disk I/O or waits on network events, it will usually run at the speed of the smallest cache that it will fit inside.

If you can't make your whole program small, some effort to tune the speed-critical portions so they have stronger locality can pay off. Details on techniques for doing such tuning are beyond the scope of this tutorial; by the time you need them, you'll be intimate enough with some compiler to figure out many of them yourself.

8.3. The Memory Management Unit

Even when you have enough physical core to avoid swapping, the part of the operating system called the memory manager still has important work to do. It has to make sure that programs can only alter their own data segments -- that is, prevent erroneous or malicious code in one program from garbaging the data in another. To do this, it keeps a table of data and code segments. The table is updated whenever a process either requests more memory or releases memory (the latter usually when it exits).

This table is used to pass commands to a specialized part of the underlying hardware called an MMU or memory management unit. Modern processor chips have MMUs built right onto them. The MMU has the special ability to put fences around areas of memory, so an out-of-bound reference will be refused and cause a special interrupt to be raised.

If you ever see a Unix message that says "Segmentation fault", "core dumped" or something similar, this is exactly what has happened; an attempt by the running program to access memory (core) outside its segment has raised a fatal interrupt. This indicates a bug in the program code; the core dump it leaves behind is diagnostic information intended to help a programmer track it down.

There is another aspect to protecting processes from each other besides segregating the memory they access. You also want to be able to control their file accesses so a buggy or malicious program can't corrupt critical pieces of the system. This is why Unix has file permissions which we'll discuss later.

9.1. Numbers

Integer numbers are represented as either words or pairs of words, depending on your processor's word size. One 32-bit machine word is the most common integer representation.

Integer arithmetic is close to but not actually mathematical base-two. The low-order bit is 1, next 2, then 4 and so forth as in pure binary. But signed numbers are represented in twos-complement notation. The highest-order bit is a sign bit which makes the quantity negative, and every negative number can be obtained from the corresponding positive value by inverting all the bits and adding one. This is why integers on a 32-bit machine have the range -2^31 to 2^31 - 1 1 (where ^ is the `power' operation, 2^3 = 8). That 32nd bit is being used for sign.

Some computer languages give you access to unsigned arithmetic which is straight base 2 with zero and positive numbers only.

Most processors and some languages can do operations in floating-point numbers (this capability is built into all recent processor chips). Floating-point numbers give you a much wider range of values than integers and let you express fractions. The ways in which this is done vary and are rather too complicated to discuss in detail here, but the general idea is much like so-called `scientific notation', where one might write (say) 1.234 * 10^23; the encoding of the number is split into a mantissa (1.234) and the exponent part (23) for the power-of-ten multiplier (which means the number multiplied out would have 20 zeros on it, 23 minus the three decimal places).

9.2. Characters

Characters are normally represented as strings of seven bits each in an encoding called ASCII (American Standard Code for Information Interchange). On modern machines, each of the 128 ASCII characters is the low seven bits of an octet or 8-bit byte; octets are packed into memory words so that (for example) a six-character string only takes up two memory words. For an ASCII code chart, type `man 7 ascii' at your Unix prompt.

The preceding paragraph was misleading in two ways. The minor one is that the term `octet' is formally correct but seldom actually used; most people refer to an octet as byte and expect bytes to be eight bits long. Strictly speaking, the term `byte' is more general; there used to be, for example, 36-bit machines with 9-bit bytes (though there probably never will be again).

The major one is that not all the world uses ASCII. In fact, much of the world can't -- ASCII, while fine for American English, lacks many accented and other special characters needed by users of other languages. Even British English has trouble with the lack of a pound-currency sign.

There have been several attempts to fix this problem. All use the extra high bit that ASCII doesn't, making it the low half of a 256-character set. The most widely-used of these is the so-called `Latin-1' character set (more formally called ISO 8859-1). This is the default character set for Linux, HTML, and X. Microsoft Windows uses a mutant version of Latin-1 that adds a bunch of characters such as right and left double quotes in places proper Latin-1 leaves unassigned for historical reasons (for a scathing account of the trouble this causes, see the demoroniser page).

Latin-1 handles western European languages, including English, French, German, Spanish, Italian, Dutch, Norwegian, Swedish, Danish. However, this isn't good enough either, and as a result there is a whole series of Latin-2 through -9 character sets to handle things like Greek, Arabic, Hebrew, Esperanto, and Serbo-Croatian. For details, see the ISO alphabet soup page.

The ultimate solution is a huge standard called Unicode (and its identical twin ISO/IEC 10646-1:1993). Unicode is identical to Latin-1 in its lowest 256 slots. Above these in 16-bit space it includes Greek, Cyrillic, Armenian, Hebrew, Arabic, Devanagari, Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada, Malayalam, Thai, Lao, Georgian, Tibetan, Japanese Kana, the complete set of modern Korean Hangul, and a unified set of Chinese/Japanese/Korean (CJK) ideographs. For details, see the Unicode Home Page.

10.1. Low-level disk and file system structure

The surface area of your disk, where it stores data, is divided up something like a dartboard -- into circular tracks which are then pie-sliced into sectors. Because tracks near the outer edge have more area than those close to the spindle at the center of the disk, the outer tracks have more sector slices in them than the inner ones. Each sector (or disk block) has the same size, which under modern Unixes is generally 1 binary K (1024 8-bit words). Each disk block has a unique address or disk block number.

Unix divides the disk into disk partitions. Each partition is a continuous span of blocks that's used separately from any other partition, either as a file system or as swap space. The original reasons for partitions had to do with crash recovery in a world of much slower and more error-prone disks; the boundaries between them reduce the fraction of your disk likely to become inaccessible or corrupted by a random bad spot on the disk. Nowadays, it's more important that partitions can be declared read-only (preventing an intruder from modifying critical system files) or shared over a network through various means we won't discuss here. The lowest-numbered partition on a disk is often treated specially, as a boot partition where you can put a kernel to be booted.

Each partition is either swap space (used to implement virtual memory) or a file system used to hold files. Swap-space partitions are just treated as a linear sequence of blocks. File systems, on the other hand, need a way to map file names to sequences of disk blocks. Because files grow, shrink, and change over time, a file's data blocks will not be a linear sequence but may be scattered all over its partition (from wherever the operating system can find a free block when it needs one). This scattering effect is called fragmentation.

10.2. File names and directories

Within each file system, the mapping from names to blocks is handled through a structure called an i-node. There's a pool of these things near the ``bottom'' (lowest-numbered blocks) of each file system (the very lowest ones are used for housekeeping and labeling purposes we won't describe here). Each i-node describes one file. File data blocks (including directories) live above the i-nodes (in higher-numbered blocks).

Every i-node contains a list of the disk block numbers in the file it describes. (Actually this is a half-truth, only correct for small files, but the rest of the details aren't important here.) Note that the i-node does not contain the name of the file.

Names of files live in directory structures. A directory structure just maps names to i-node numbers. This is why, in Unix, a file can have multiple true names (or hard links); they're just multiple directory entries that happen to point to the same i-node.

10.3. Mount points

In the simplest case, your entire Unix file system lives in just one disk partition. While you'll see this arrangement on some small personal Unix systems, it's unusual. More typical is for it to be spread across several disk partitions, possibly on different physical disks. So, for example, your system may have one small partition where the kernel lives, a slightly larger one where OS utilities live, and a much bigger one where user home directories live.

The only partition you'll have access to immediately after system boot is your root partition, which is (almost always) the one you booted from. It holds the root directory of the file system, the top node from which everything else hangs.

The other partitions in the system have to be attached to this root in order for your entire, multiple-partition file system to be accessible. About midway through the boot process, your Unix will make these non-root partitions accessible. It will mount each one onto a directory on the root partition.

For example, if you have a Unix directory called `/usr', it is probably a mount point to a partition that contains many programs installed with your Unix but not required during initial boot.

10.4. How a file gets looked up

Now we can look at the file system from the top down. When you open a file (such as, say, /home/esr/WWW/ldp/fundamentals.sgml) here is what happens:

Your kernel starts at the root of your Unix file system (in the root partition). It looks for a directory there called `home'. Usually `home' is a mount point to a large user partition elsewhere, so it will go there. In the top-level directory structure of that user partition, it will look for a entry called `esr' and extract an i-node number. It will go to that i-node, notice that its associated file data blocks are a directory structure, and look up `WWW'. Extracting that i-node, it will go to the corresponding subdirectory and look up `ldp'. That will take it to yet another directory i-node. Opening that one, it will find an i-node number for `fundamentals.sgml'. That i-node is not a directory, but instead holds the list of disk blocks associated with the file.

10.5. File ownership, permissions and security

To keep programs from accidentally or maliciously stepping on data they shouldn't, Unix has permission features. These were originally designed to support timesharing by protecting multiple users on the same machine from each other, back in the days when Unix ran mainly on expensive shared minicomputers.

In order to understand file permissions, you need to recall the description of users and groups in the section What happens when you log in?. Each file has an owning user and an owning group. These are initially those of the file's creator; they can be changed with the programs chown(1) and chgrp(1).

The basic permissions that can be associated with a file are `read' (permission to read data from it), `write' (permission to modify it) and `execute' (permission to run it as a program). Each file has three sets of permissions; one for its owning user, one for any user in its owning group, and one for everyone else. The `privileges' you get when you log in are just the ability to do read, write, and execute on those files for which the permission bits match your user ID or one of the groups you are in, or files that have been made accessible to the world.

To see how these may interact and how Unix displays them, let's look at some file listings on a hypothetical Unix system. Here's one:

snark:~$ ls -l notes
-rw-r--r--   1 esr      users         2993 Jun 17 11:00 notes

This is an ordinary data file. The listing tells us that it's owned by the user `esr' and was created with the owning group `users'. Probably the machine we're on puts every ordinary user in this group by default; other groups you commonly see on timesharing machines are `staff', `admin', or `wheel' (for obvious reasons, groups are not very important on single-user workstations or PCs). Your Unix may use a different default group, perhaps one named after your user ID.

The string `-rw-r--r--' represents the permission bits for the file. The very first dash is the position for the directory bit; it would show `d' if the file were a directory. After that, the first three places are user permissions, the second three group permissions, and the third are permissions for others (often called `world' permissions). On this file, the owning user `esr' may read or write the file, other people in the `users' group may read it, and everybody else in the world may read it. This is a pretty typical set of permissions for an ordinary data file.

Now let's look at a file with very different permissions. This file is GCC, the GNU C compiler.

snark:~$ ls -l /usr/bin/gcc
-rwxr-xr-x   3 root     bin         64796 Mar 21 16:41 /usr/bin/gcc

This file belongs to a user called `root' and a group called `bin'; it can be written (modified) only by root, but read or executed by anyone. This is a typical ownership and set of permissions for a pre-installed system command. The `bin' group exists on some Unixes to group together system commands (the name is a historical relic, short for `binary'). Your Unix might use a `root' group instead (not quite the same as the `root' user!).

The `root' user is the conventional name for numeric user ID 0, a special, privileged account that can override all privileges. Root access is useful but dangerous; a typing mistake while you're logged in as root can clobber critical system files that the same command executed from an ordinary user account could not touch.

Because the root account is so powerful, access to it should be guarded very carefully. Your root password is the single most critical piece of security information on your system, and it is what any crackers and intruders who ever come after you will be trying to get.

About passwords: Don't write them down -- and don't pick a passwords that can easily be guessed, like the first name of your girlfriend/boyfriend/spouse. This is an astonishingly common bad practice that helps crackers no end. In general, don't pick any word in the dictionary; there are programs called dictionary crackers that look for likely passwords by running through word lists of common choices. A good technique is to pick a combination consisting of a word, a digit, and another word, such as `shark6cider' or `jump3joy'; that will make the search space too large for a dictionary cracker. Don't use these examples, though -- crackers might expect that after reading this document and put them in their dictionaries.

Now let's look at a third case:

snark:~$ ls -ld ~
drwxr-xr-x  89 esr      users          9216 Jun 27 11:29 /home2/esr
snark:~$ 

This file is a directory (note the `d' in the first permissions slot). We see that it can be written only by esr, but read and executed by anybody else.

Read permission gives you the ability to list the directory -- that is, to see the names of files and directories it contains. Write permission gives you the ability to create and delete files in the directory. If you remember that the directory includes a list of the names of the files and subdirectories it contains, these rules will make sense.

Execute permission on a directory means you can get through the directory to open the files and directories below it. In effect, it gives you permission to access the i-nodes in the directory. A directory with execute completely turned off would be useless.

Occasionally you'll see a directory that is world-executable but not world-readable; this means a random user can get to files and directories beneath it, but only by knowing their exact names (the directory cannot be listed).

It's important to remember that read, write, or execute permission on a directory is independent of the permissions on the files and directories beneath. In particular, write access on a directory means you can create new files or delete existing files there, but does not automatically give you write access to existing files.

Finally, let's look at the permissions of the login program itself.

snark:~$ ls -l /bin/login
-rwsr-xr-x   1 root     bin         20164 Apr 17 12:57 /bin/login

This has the permissions we'd expect for a system command -- except for that 's' where the owner-execute bit ought to be. This is the visible manifestation of a special permission called the `set-user-id' or setuid bit.

The setuid bit is normally attached to programs that need to give ordinary users the privileges of root, but in a controlled way. When it is set on an executable program, you get the privileges of the owner of that program file while the program is running on your behalf, whether or not they match your own.

Like the root account itself, setuid programs are useful but dangerous. Anyone who can subvert or modify a setuid program owned by root can use it to spawn a shell with root privileges. For this reason, opening a file to write it automatically turns off its setuid bit on most Unixes. Many attacks on Unix security try to exploit bugs in setuid programs in order to subvert them. Security-conscious system administrators are therefore extra-careful about these programs and reluctant to install new ones.

There are a couple of important details we glossed over when discussing permissions above; namely, how the owning group and permissions are assigned when a file or directory is first created. The group is an issue because users can be members of multiple groups, but one of them (specified in the user's /etc/passwd entry) is the user's default group and will normally own files created by the user.

The story with initial permission bits is a little more complicated. A program that creates a file will normally specify the permissions it is to start with. But these will be modified by a variable in the user's environment called the umask. The umask specifies which permission bits to turn off when creating a file; the most common value, and the default on most systems, is -------w- or 002, which turns off the world-write bit. See the documentation of the umask command on your shell's manual page for details.

Initial directory group is also a bit complicated. On some Unixes a new directory gets the default group of the creating user (this in the System V convention); on others, it gets the owning group of the parent directory in which it's created (this is the BSD convention). On some modern Unixes, including Linux, the latter behavior can be selected by setting the set-group-ID on the directory (chmod g+s).

10.6. How things can go wrong

Earlier it was hinted that file systems can be fragile things. Now we know that to get to a file you have to hopscotch through what may be an arbitrarily long chain of directory and i-node references. Now suppose your hard disk develops a bad spot?

If you're lucky, it will only trash some file data. If you're unlucky, it could corrupt a directory structure or i-node number and leave an entire subtree of your system hanging in limbo -- or, worse, result in a corrupted structure that points multiple ways at the same disk block or i-node. Such corruption can be spread by normal file operations, trashing data that was not in the original bad spot.

Fortunately, this kind of contingency has become quite uncommon as disk hardware has become more reliable. Still, it means that your Unix will want to integrity-check the file system periodically to make sure nothing is amiss. Modern Unixes do a fast integrity check on each partition at boot time, just before mounting it. Every few reboots they'll do a much more thorough check that takes a few minutes longer.

If all of this sounds like Unix is terribly complex and failure-prone, it may be reassuring to know that these boot-time checks typically catch and correct normal problems before they become really disastrous. Other operating systems don't have these facilities, which speeds up booting a bit but can leave you much more seriously screwed when attempting to recover by hand (and that's assuming you have a copy of Norton Utilities or whatever in the first place...).

One of the trends in current Unix designs is journalling file systems. These arrange traffic to the disk so that it's guaranteed to be in a consistent state that can be recovered when the system comes back up. This will speed up the boot-time integrity check a lot.

11.1. Compiled languages

The most conventional kind of language is a compiled language. Compiled languages get translated into runnable files of binary machine code by a special program called (logically enough) a compiler. Once the binary has been generated, you can run it directly without looking at the source code again. (Most software is delivered as compiled binaries made from code you don't see.)

Compiled languages tend to give excellent performance and have the most complete access to the OS, but also to be difficult to program in.

C, the language in which Unix itself is written, is by far the most important of these (with its variant C++). FORTRAN is another compiled language still used among engineers and scientists but years older and much more primitive. In the Unix world no other compiled languages are in mainstream use. Outside it, COBOL is very widely used for financial and business software.

There used to be many other compiler languages, but most of them have either gone extinct or are strictly research tools. If you are a new Unix developer using a compiled language, it is overwhelmingly likely to be C or C++.

11.2. Interpreted languages

An interpreted language depends on an interpreter program that reads the source code and translates it on the fly into computations and system calls. The source has to be re-interpreted (and the interpreter present) each time the code is executed.

Interpreted languages tend to be slower than compiled languages, and often have limited access to the underlying operating system and hardware. On the other hand, they tend to be easier to program and more forgiving of coding errors than compiled languages.

Many Unix utilities, including the shell and bc(1) and sed(1) and awk(1), are effectively small interpreted languages. BASICs are usually interpreted. So is Tcl. Historically, the most important interpretive language has been LISP (a major improvement over most of its successors). Today, Unix shells and the Lisp that lives inside the Emacs editor are probably the most important pure interpreted languages.

11.3. P-code languages

Since 1990 a kind of hybrid language that uses both compilation and interpretation has become increasingly important. P-code languages are like compiled languages in that the source is translated to a compact binary form which is what you actually execute, but that form is not machine code. Instead it's pseudocode (or p-code), which is usually a lot simpler but more powerful than a real machine language. When you run the program, you interpret the p-code.

P-code can run nearly as fast as a compiled binary (p-code interpreters can be made quite simple, small and speedy). But p-code languages can keep the flexibility and power of a good interpreter.

Important p-code languages include Python, Perl, and Java.

12.1. Names and locations

The first thing your browser has to do is to establish a network connection to the machine where the document lives. To do that, it first has to find the network location of the host www.linuxdoc.org (`host' is short for `host machine' or `network host'; www.linuxdoc.org is a typical hostname). The corresponding location is actually a number called an IP address (we'll explain the `IP' part of this term later).

To do this, your browser queries a program called a name server. The name server may live on your machine, but it's more likely to run on a service machine that yours talks to. When you sign up with an ISP, part of your setup procedure will almost certainly involve telling your Internet software the IP address of a nameserver on the ISP's network.

The name servers on different machines talk to each other, exchanging and keeping up to date all the information needed to resolve hostnames (map them to IP addresses). Your nameserver may query three or four different sites across the network in the process of resolving www.linuxdoc.org, but this usually happens very quickly (as in less than a second). We'll look at how nameservers detail in the next section.

The nameserver will tell your browser that www.linuxdoc.org's IP address is 152.19.254.81; knowing this, your machine will be able to exchange bits with www.linuxdoc.org directly.

12.2. The Domain Name System

The whole network of programs and databases that cooperates to translate hostnames to IP addresses is called `DNS' (Domain Name System). When you see references to a `DNS server', that means what we just called a nameserver. Now I'll explain how the overall system works.

Internet hostnames are composed of parts separated by dots. A domain is a collection of machines that share a common name suffix. Domains can live inside other domains. For example, the machine www.linuxdoc.org lives in the .linuxdoc.org subdomain of the .org domain.

Each domain is defined by an authoritative name server that knows the IP addresses of the other machines in the domain. The authoritative (or `primary') name server may have backups in case it goes down; if you see references to a secondary name server or (`secondary DNS') it's talking about one of those. These secondaries typically refresh their information from their primaries every few hours, so a change made to the hostname-to-IP mapping on the primary will automatically be propagated.

Now here's the important part. The nameservers for a domain do not have to know the locations of all the machines in other domains (including their own subdomains); they only have to know the location of the nameservers. In our example, the authoritative name server for the .org domain knows the IP address of the nameserver for .linuxdoc.org, but not the address of all the other machines in linuxdoc.org.

The domains in the DNS system are arranged like a big inverted tree. At the top are the root servers. Everybody knows the IP addresses of the root servers; they're wired into your DNS software. The root servers know the IP addresses of the nameservers for the top-level domains like .com and .org, but not the addresses of machines inside those domains. Each top-level domain server knows where the nameservers for the domains directly beneath it are, and so forth.

DNS is carefully designed so that each machine can get away with the minimum amount of knowledge it needs to have about the shape of the tree, and local changes to subtrees can be made simply by changing one authoritative server's database of name-to-IP-address mappings.

When you query for the IP address of www.linuxdoc.org, what actually happens is this: First, your nameserver asks a root server to tell it where it can find a nameserver for .org. Once it knows that, it then asks the .org server to tell it the IP address of a .linuxdoc.org nameserver. Once it has that, it asks the .linuxdoc.org nameserver to tell it the address of the host www.linuxdoc.org.

Most of the time, your nameserver doesn't actually have to work that hard. Nameservers do a lot of cacheing; when yours resolves a hostname, it keeps the association with the resulting IP address around in memory for a while. This is why, when you surf to a new website, you'll usually only see a message from your browser about "Looking up" the host for the first page you fetch. Eventually the name-to-address mapping expires and your DNS has to re-query — this is important so you don't have invalid information hanging around forever when a hostname changes addresses. Your cached IP address for a site is also thrown out if the host is unreachable.

12.3. Packets and routers

What the browser wants to do is send a command to the Web server on www.linuxdoc.org that looks like this:

GET /LDP/HOWTO/Fundamentals.html HTTP/1.0

Here's how that happens. The command is made into a packet, a block of bits like a telegram that is wrapped with three important things; the source address (the IP address of your machine), the destination address (152.19.254.81), and a service number or port number (80, in this case) that indicates that it's a World Wide Web request.

Your machine then ships the packet down the wire (your connection to your ISP, or local network) until it gets to a specialized machine called a router. The router has a map of the Internet in its memory -- not always a complete one, but one that completely describes your network neighborhood and knows how to get to the routers for other neighborhoods on the Internet.

Your packet may pass through several routers on the way to its destination. Routers are smart. They watch how long it takes for other routers to acknowledge having received a packet. They also use that information to direct traffic over fast links. They use it to notice when another routers (or a cable) have dropped off the network, and compensate if possible by finding another route.

There's an urban legend that the Internet was designed to survive nuclear war. This is not true, but the Internet's design is extremely good at getting reliable performance out of flaky hardware in an uncertain world. This is directly due to the fact that its intelligence is distributed through thousands of routers rather than concentrated in a few massive and vulnerable switches (like the phone network). This means that failures tend to be well localized and the network can route around them.

Once your packet gets to its destination machine, that machine uses the service number to feed the packet to the web server. The web server can tell where to reply to by looking at the command packet's source IP address. When the web server returns this document, it will be broken up into a number of packets. The size of the packets will vary according to the transmission media in the network and the type of service.

12.4. TCP and IP

To understand how multiple-packet transmissions are handled, you need to know that the Internet actually uses two protocols, stacked one on top of the other.

The lower level, IP (Internet Protocol), is responsible for labeling individual packets with the source address and destination address of two computers exchanging information over a network. For example, when you access http://www.linuxdoc.org, the packets you send will have your computer's IP address, such as 192.168.1.101, and the IP address of the www.linuxdoc.org computer, 152.2.210.81. These addresses work in much the same way that your home address works when someone sends you a letter. The post office can read the address and determine where you are and how best to route the letter to you, much like a router does for Internet traffic.

The upper level, TCP (Transmission Control Protocol), gives you reliability. When two machines negotiate a TCP connection (which they do using IP), the receiver knows to send acknowledgements of the packets it sees back to the sender. If the sender doesn't see an acknowledgement for a packet within some timeout period, it resends that packet. Furthermore, the sender gives each TCP packet a sequence number, which the receiver can use you reassemble packets in case they show up out of order. (This can easily happen if network links go up or down during a connection.)

TCP/IP packets also contain a checksum to enable detection of data corrupted by bad links. (The checksum is computed from the rest of the packet in such a way that if the either the rest of the packet or the checksum is corrupted, redoing the computation and comparing is very likely to indicate an error.) So, from the point of view of anyone using TCP/IP and nameservers, it looks like a reliable way to pass streams of bytes between hostname/service-number pairs. People who write network protocols almost never have to think about all the packetizing, packet reassembly, error checking, checksumming, and retransmission that goes on below that level.

12.5. HTTP, an application protocol

Now let's get back to our example. Web browsers and servers speak an application protocol that runs on top of TCP/IP, using it simply as a way to pass strings of bytes back and forth. This protocol is called HTTP (Hyper-Text Transfer Protocol) and we've already seen one command in it -- the GET shown above.

When the GET command goes to www.linuxdoc.org's webserver with service number 80, it will be dispatched to a server daemon listening on port 80. Most Internet services are implemented by server daemons that do nothing but wait on ports, watching for and executing incoming commands.

If the design of the Internet has one overall rule, it's that all the parts should be as simple and human-accessible as possible. HTTP, and its relatives (like the Simple Mail Transfer Protocol, SMTP, that is used to move electronic mail between hosts) tend to use simple printable-text commands that end with a carriage-return/line feed.

This is marginally inefficient; in some circumstances you could get more speed by using a tightly-coded binary protocol. But experience has shown that the benefits of having commands be easy for human beings to describe and understand outweigh any marginal gain in efficiency that you might get at the cost of making things tricky and opaque.

Therefore, what the server daemon ships back to you via TCP/IP is also text. The beginning of the response will look something like this (a few headers have been suppressed):

HTTP/1.1 200 OK
Date: Sat, 10 Oct 1998 18:43:35 GMT
Server: Apache/1.2.6 Red Hat
Last-Modified: Thu, 27 Aug 1998 17:55:15 GMT
Content-Length: 2982
Content-Type: text/html

These headers will be followed by a blank line and the text of the web page (after which the connection is dropped). Your browser just displays that page. The headers tell it how (in particular, the Content-Type header tells it the returned data is really HTML).