Provided by: elektra-doc_0.7.1-1_all
elektra - A framework to store configuration atoms hierarchically
This section is provided for the sake of the openness of Elektra.
You should not access the Elektra's key files directly. You should
use the API or the kdb(1) command for that.
Elektra Key Storage Strategy.PP Elektra implements a very simple way to
store the key-value pairs. The value (and some metainfo) for each key
is stored in a single file. The key name (and some of its context) is
sufficient to find the file name that stores the value.
The system/* keys are stored under /etc/kdb/, and the user/* keys can
be found under each user's $HOME/.kdb/.
Here are some examples of key names, and where Elektra goes to look for
them in the disk.
Maps to: /etc/kdb/system/sw/XFree86/screen0/driver
Maps to: ~$USER/.kdb/user/env/PATH
Maps to: ~luciana/.kdb/user/env/PATH
Maps to: /etc/kdb/system/mime.types/some.mime
Some may think that one file per key will consume many filesystem
i-nodes. Actually, when not using Reiser4 filesystem, it may consume
some more disk space, and it may also be not so efficient than reading
one single text file, as KConfig does. But Elektra's nature lets
applications load their keys on demand; so it is possible to avoid the
read-all-use-some approach. Writing updated keys back to disk is also
more robust, because unchanged keys won't be touched, different from a
single file approach, that must be entirelly rewritten.
Besides that, big applications (like Mozilla, Konqueror, KDE, Gnome)
key gathering time is a very small part of what they have to do to
start up. And the benefits of an homogeneous key database to the entire
system are much bigger then these issues. Think about a common
integration between everything, flexibility, security granularity and
openness. XML, Storage Backends and Elektra.PP This document you are
just reading was written in DocBook XML. XML is a wonderfull
technology, but brings no value to this software. Two main goals of the
Elektra Project are to be lightweight, and to be accessible by early
boot stage programs like /sbin/init and the /etc/rc.d scripts.
XML parsing libraries are memory eaters, not so fast as we can expect,
and they are usually located under /usr/lib, which may be unavailable
to these early boot stage needs.
Some discussions asked for a sort of plugin architecture to let user
decide the format to store keys content. Well, the info that goes into
the key file is not big deal as you'll see, and sometimes, too many
options is bad business, and not the path for the Elektra Project.
So no XML, no plugin architecture, no sophistication. Lets keep it
simple and generic. A very straight forward text based file format was
defined to store a single key value. Key Files Format.PP Inside
Elektra key database, each key is a file, and every file is a key. So
most of the key's metainformation are actually its file attributes, as
you can see in a regular ls(1) command output.
So what needs to be stored inside the key file is the data type (binary
or text), key comment and the actual data. The format of each key file
File Format Version
As many lines of
comments as we want (UTF-8 encoded)
The data encoded as text
So if we set the key system/hw/eth0/driver as type String and value
"3com", and comment "The driver for my network interface", we'll find
the file /etc/kdb/system/hw/eth0/driver containing:
The driver for my network interface
Other example: setting user/tmp/proprietary as Binary, and value "A
groovy data I want to hide", and comment "Stay away from here", you'll
get in ~$USER/.kdb/user/tmp/proprietary the following:
Stay away from here
41206772 6f6f7679 20646174 61204920 77616e74 20746f20 68696465
The current data types are:
Between 20 and 39
Binary value. The data will be encoded into a text format. Today
only type 20 is used, and means a raw stream of bytes with no
special semantics to Elektra. The other values are reserved for
future use; being treated still as binary values but possibly with
some semantics to Elektra or a higher level application.
40 up to 254
Text, UTF-8 encoded. Values higher then 40 are reserved for future
or application specific implementations of more abstract data
types, like time/date, color, font, etc. But always represented as
pure text that can be edited in any text editor like vi(1).
Types between 0 and 19 are used only internaly in the API, and will
never appear into a key file. They are used to define meta keys as
directory, link, etc.
Avi Alkalay <avi at unix.sh>
Linux Market Developer, Senior IT and Software Architect, IBM Linux
Impact Team :: ibm.com/linux
Copyright (C) 2004 Avi Alkalay