From: Praveen Rao (psrao_at_windows.microsoft.com)
Date: Wed Jan 07 2004 - 17:05:39 PST
This paper provides a good overview of UNIX system without going into
depths of implementations of any specific features.
File system is cited as the most important role of the system. The paper
talks about ordinary files, directories and special files. It also
explains linking of the files. One of the features of the system
highlighted is that each I/O device supported by the system is
associated with at least one file. The mechanism to interact with I/O
devices and reading/writing ordinary data files is exactly the same.
There is uniformity in access paradigm, naming and protection scheme
across devices and ordinary files. This I believe is a very powerful
paradigm and paper does a good job of driving this point.
The paper then describes some details of implementation of the file
systems. It is made clear that in case of linked files, there is no
"master" file amongst the links and that the information about the file
is stored separately from the file. Links just increment and decrement
the reference count on the file.
The image execution model (execution/termination/synchronization) is
described in some detail and one method of inter-process communication
(pipes) is touched upon. This nicely leads the way into the discussion
of the Shell - it is discussed how commands are executed in shell, how
input and output redirection works and how the high level view of how
the shell itself is implemented.
Overall this was a very nice paper for the overview of UNIX and the key
point highlighted, IMO, was the file access model and how that extends
to devices and ties into image execution through shell. The open
question the paper leaves is what limits extending such resources usage
model (used for files and devices) to other resources like memory.
This archive was generated by hypermail 2.1.6 : Wed Jan 07 2004 - 17:05:42 PST