P

Partition
A partition is a piece of a disk.

Most modern hard disks store a large amount of data very efficiently and quickly. Often it is useful to divide up a disk into partitions. Each partition can be used for something different. Usually each partition has a volume created on it during the formatting process.

Partitioning a disk is like taking a large, empty building and installing walls. The building is more useful when it is broken up into rooms. Of course, each room is still empty and needs to be outfitted; likewise, an empty partition needs to be formatted. As with walls of a building, repartitioning is destructive and dangerous and is best done when the disk has no data on it.

You can partition your disk using a disk partitioning program. Which program you use depends on your computer and your budget. On x86 and Sparc architectures, your partitioning program may very well be called fdisk, even though there are several of them, from Microsoft, Sun, and open source, and several versions from each. All named fdisk. Ironically, on PowerPC, none of the available programs are named fdisk, although one is named pdisk. Apple has published a few partitioning programs for PowerPCs with various capabilities.

There are also commercial partion programs for x86 and Mac - sometimes you get a free version with a new external disk. Partition Magic is a favorite on x86, but another program, Acronis Partition Expert, has adherents. You should expect these programs to be more reliable. That is, if and when they erase all of your data, you can rightfully complain to the developers. Other than that, the same rules apply - experiment with partitioning only when you can tolerate losing it all.

Disk partitioning programs, especially the free ones, are traditionally unreliable, and often incompatible with each other in subtle ways. In addition, they are typically badly documented, and their message wording can be confusing, misleading or even just wrong. This is unfortunate because it's easy to completely wipe away gigabytes of data, and hours of installation struggle time, with one little mistake or misunderstood feature.

Some examples: BSD and Solaris cultures use the words "slice" and "partition" with opposite meanings. Some numbers are in k, some in bytes, some in 512-byte blocks, some in megabytes or gigabytes, some in cylinders or tracks. Some programs round off quantities to cylinder boundaries, some don't. Some will take existing partitions that are not rounded off, round off the edges and repartition - destroying all of your data. One user started up a partitioning program, which gave him a warning, asking if he wanted to continue. Upon answering No, it went and deleted all of his existing partitions.

Often it is best to use the partitioning program that comes with the OS you plan to use. When you have a new disk that you are partitioning, it is best to get all the partitioning done right then, when mistakes simply erase empty partitions. Later on, when you have useful data in some of the partitions, partitioning mistakes will be very expensive. That is, if you have a backup, mistakes will be expensive. If you don't have a backuip, mistakes will be catastrophic.

If you plan to use more than one OS, partitioning can get tricky. Experimentation should be your best guide. Again, the best time to experiment is when your computer or disk are brand new, before you have any valuable data on it. Repartition and reformat many times to see how things work. Try installing your OSs a few times so you learn what works and what doesn't work. For instance, on x86, the Windows ME installer will tend to wipe away your other partitions, so you install that first. Solaris 8 will wipe away anything on disk 0 other than Windows, so you might install that second, unless you know some other tricks. Don't start customizing your OSs until you feel confident you won't have to reinstall them again.

On x86 architectures, the overall partitioning system has four slots for four primary partitions. Because this is limited, there are ways to subdivide these into smaller pieces. Microsoft FDISK usually makes the second partition an "extended partition" which is subdivided into "logical partitions". Linux usually uses these also. BSD and Solaris use another system called Slices. Sometimes Linux systems understand these, too. Unfortunately, the Solaris and BSD slices have subtle differences and are mutually incompatible.

On MacOS/PowerPC architectures, a completely different and more rational partitioning scheme is in use which accomodates a dozen or more partitions. Usually for MacOS, some of the earlier partitions store drivers or other hidden chunks of data. Some partition programs show you these, although the ones from Apple often hide these, leading to questionable megabyte arithmetic.

The partition table is not a filesystem. A disk with more than a dozen partitions is on thin ice. Play it safe with your gigabytes. A few hundred dollars for a new disk is often money well spent, saving you much time.

Path
Path is short for two different terms, Path Name and Search Path. Basically, a search path is a list of pathnames. The term path is ambiguous and its use should be avoided. Often, though, you can figure out which meaning by context.

Path Name
A fully qualified name of a file, with all of its directories, all the way from the root. On Unix, a path name starts with a slash, and has slashes separating directory names.

For instance, the pathname /usr/bin/man means:
step       gets you to
start at the root of the machine's filesystem /
go to the usr subdirectory of / /usr/
go to the bin subdirectory of usr /usr/bin/
get the file man in bin. /usr/bin/man

See also the Filesystems Overview.

pdisk
A partitioning program for Macintosh computers. This is not supplied by Apple and it is not recommended that you use this program to actually create partitions, although it is good to see the existing partitions in your disk.

Permission
Permissions in Unix prevent people from reading or modifying data that doesn't belong to them.

Every user, and in fact every program, on Unix has a username and userid. The userid is just an integer, such as 521, that's used internally as shorthand for the username. The file /etc/passwd often lists which name goes with which ID. Also, each user is a member of one or more groups; the groups and their IDs are listed in /etc/groups.

The superuser always has userid 0, username "root", and always has permission to do anything they want.

Each file has an owner (userid), and a group (groupid), and nine permission bits. You can see this information in Interrogator by turning on Full Disclosure Mode and selecting the file. The nine permission bits are broken into three groups of three. The three owner bits tell what the owner is allowed to do. If you are not the owner, but you are in the file's group, then the three group bits tell what you are allowed to do. If you are neither owner or in a group, the last three permission bits apply to you.

The three permission bits are written as "rwx" if they are all on, or "---" if they are all off, or a mix depending. "r--" means you only have read permission on the file, "rw-" means you have write permission on the file. "x" permission on a directory, called Search permission, allows you to use the directory name in a path name. "x" permission on a script file or binary executable allows you to run it as a program.

There are other permission bits with other more obscure properties (see man ls, man chmod). On Solaris and more recent versions of Linux, there is also a more granular system called ACLs where individual users can be granted or denied permissions in precise detail. On MacOS, there are also 'flag bits'; see man chflags.

Pipe
An operating system object that represents a communications connection to a program on the same computer. Strictly speaking, a pipe does not have to have a filesystem entry; the pipes you make on the command line with the | character are an example. Pipes that do appear in the file system are called Fifos.

Primary Partition
On x86 architectures, the overall partitioning system has four slots for four primary partitions. Because this is limited, there are ways to subdivide these into smaller pieces. Unfortunately, some of these methods are incompatible with each other. Microsoft's standard is called Logical Partitions. Usually Linux understands and uses these. Rules come and go over the years, for instance, Microsoft's recent fdisk partitioning program will refuse to create primary partitions 3 or 4, although it recognizes them if you have created them with Linux's fdisk. But if you create logical partitions with Linux's fdisk, Microsoft's fdisk will ignore them and pretend they don't exist.


Documentation > Glossary >
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z   
0123456789 punctuation
       

                     

T a c t i l e   I n t e r r o g a t o r   W e b s i t e