01_19_02_plan_9_bell_labs.txt

# Plan9

## Specialize model
* Originally: all user logged in a time shared system, now each user has a unix
* Therefore, there is admin burden
* Now: rejects run unix on everything
* New model: Specialize
      Different nodes have different functions.
      * Terminals
      * CPU Server
      * File Server

## Unified
* Everything like a file (same as Unix, even more in some place, like control device is a file in Plan9)
* Everything becomes a file server
* 9P protocol connects all clients and servers

## Local Name Space
* Different from Unix's global name space.
* Customize the name space, no uniform one. Every process sees a different one.
* Name convention /dev/con -> console windown. Change what /dev/con is mapped to, to have a local or remote console.
* Associated with process (each process has own namespace)

## WORM
* Three levels of storage.
* The disk is a cache for the WORM and the memory is a cache for the disk.
* Every day, a dump of the file system occurs automatically, written to the WORM.

## RFORK
* rfork falls somewhere between processes and kernel threads. rfork create a new process, and specify what should be shared. BSD has rfork, and Unix has clone, which is similar.

## Union Directory
* We can have concatenation of directories.
* No $PATH needed.
* /bin is a union dir of every dir we can find runnable programs. It is built as a union of$cputype/bin, /rc/bin and perhaps more defined by the user.
* The lookup in a union directory obeys the rule that search in turn, the first one match will be taken.

## Influences
* Unicode, UTF8
* file caches
* /proc is later used in Linux.
* rfork exists in BSD, clone() in Linux.
* NFS block access, plan9 byte access