You are viewing a plain text version of this content. The canonical link for it is here.
Posted to cvs@avalon.apache.org by do...@apache.org on 2003/06/06 04:46:33 UTC
cvs commit: avalon-logkit/src/xdocs whitepaper.xml
donaldp 2003/06/05 19:46:33
Modified: src/xdocs whitepaper.xml
Log:
Fix spelling
Submiteted By: Anton Tagunov <at...@mail.cnt.ru>
Revision Changes Path
1.10 +84 -84 avalon-logkit/src/xdocs/whitepaper.xml
Index: whitepaper.xml
===================================================================
RCS file: /home/cvs/avalon-logkit/src/xdocs/whitepaper.xml,v
retrieving revision 1.9
retrieving revision 1.10
diff -u -r1.9 -r1.10
--- whitepaper.xml 19 Aug 2002 09:30:53 -0000 1.9
+++ whitepaper.xml 6 Jun 2003 02:46:33 -0000 1.10
@@ -2,17 +2,17 @@
<document>
<properties>
<title>LogKit - Whitepaper</title>
-
+
<author email="peter at apache.org">Peter Donald</author>
-
+
</properties>
<body>
<section name="Abstract">
<p>
- Logging is an integral component to any software development project.
+ Logging is an integral component to any software development project.
During the development stages it offers a valuable source of debugging
information for the developer. During deployment it can provide valuable
- operational data that allows administrators to diagnose problems as they
+ operational data that allows administrators to diagnose problems as they
arise. This whitepaper describes the design and implementation of LogKit.
</p>
</section>
@@ -20,20 +20,20 @@
<p>
LogKit, began life as a facade for a more complex logging toolkit. During
- development it was found that existing toolkits were complex and difficult
- to use. A malformed logger configuration file could cause the whole system
+ development it was found that existing toolkits were complex and difficult
+ to use. A malformed logger configuration file could cause the whole system
to fail or the logging information to go into the void.
</p>
<p>
- Over time it was discovered that the facade, while simple to use, inherited
+ Over time it was discovered that the facade, while simple to use, inherited
many of the faults of the underlying logging toolkit. The logging was slow
- and was impossible to dynamically reconfigure.
- </p>
-
+ and was impossible to dynamically reconfigure.
+ </p>
+
<p>
At this time the whole system was overhauled. Existing logging toolkits
- were surveyed and user feedback was gathered from both developers and
+ were surveyed and user feedback was gathered from both developers and
administrators. The new toolkit focused on speed, reliability and dynamic
reconfiguration. It then added, at that time unique features, for filtering
log events, multiple listeners (aka LogTargets) per category and delayed
@@ -47,7 +47,7 @@
<p>
The most basic components of LogKit are the Logger, the LogEvent and the
LogTarget. The Logger represents the client interface to the logging system.
- The developer interacts with Logger to generate LogEvents. The LogEvents
+ The developer interacts with Logger to generate LogEvents. The LogEvents
are routed to a LogTarget. The LogTarget decides what to do with the LogEvent,
usually it is recorded to a file, a database or transmitted over the network.
</p>
@@ -56,8 +56,8 @@
<section name="Priorities">
<p>
One of the advantages of a logging toolkit is fine grain control over which
- statements get printed. At some times during development you may wish to enable
- all logging statements and at other times they may wish to disable debug
+ statements get printed. At some times during development you may wish to enable
+ all logging statements and at other times they may wish to disable debug
messages. It was from this need that the notion of of Priorities were born.
A Priority describes the urgency of a LogEvent. Below is a list of priorities
that are usable within the LogKit system (extracted from the constants of class
@@ -71,12 +71,12 @@
</li>
<li>
<strong>INFO:</strong><br/>
- Useful information messages such as state changes, client
+ Useful information messages such as state changes, client
connection, user login etc.
</li>
<li>
<strong>WARN:</strong><br/>
- A problem or conflict has occurred but it may be recoverable, then
+ A problem or conflict has occurred but it may be recoverable, then
again it could be the start of the system failing.
</li>
<li>
@@ -91,8 +91,8 @@
</ul>
<p>
- The range of priorities is limited by design as past experience has
- indicated more levels do not offer any significant advantage to the
+ The range of priorities is limited by design as past experience has
+ indicated more levels do not offer any significant advantage to the
user.
</p>
@@ -107,7 +107,7 @@
A user can log at a certain priority by calling a method with matching name and
the message as a string. For instance to log at debug level you would call
logger.debug("My Message"). Similar methods exist for info(), warn() and error().
- There is also a set of similarly named methods that take both a message string
+ There is also a set of similarly named methods that take both a message string
and an exception.
</p>
@@ -115,8 +115,8 @@
Where performance is critical it is often useful to check if a priority is
enabled before constructing the message. In many cases the construction of
the message is an expensive string operation and conversion operation. In this
- case it is useful to know before you create the message whether it will be
- logged. The pattern of the method to check if a priority is enabled is
+ case it is useful to know before you create the message whether it will be
+ logged. The pattern of the method to check if a priority is enabled is
"is<em>Priority</em>Enabled()". An example use case is displayed below.
</p>
@@ -139,13 +139,13 @@
For instance you may wish to log the network subsystem with DEBUG priority while
the simulator subsystem with WARN priority. To accomplish this LogKit uses a concept
termed Categories. Categories, often called Channels, Subjects or Facilities, are
- a subdivision of the logging namespace.
+ a subdivision of the logging namespace.
</p>
<p>
Each category is a name, made up of name components separated by a ".". So
- a category named "network.interceptor.connected" is made up of three name
- components "network", "interceptor" and "connected", ordered from left to
+ a category named "network.interceptor.connected" is made up of three name
+ components "network", "interceptor" and "connected", ordered from left to
right. Every logger is associated with a category at creation.
</p>
@@ -155,21 +155,21 @@
name component is the most specific. So "network.interceptor.connected" is a child
category of "network.interceptor", which is in turn a child category of "network".
There is also a root category "" that is hidden inside the org.apache.log.Hierarchy
- class.
+ class.
</p>
<p>
The main reason for structuring logging namespace in a hierarchical manner is to
allow inheritance. A logger will inherit it's parent priority if it has not
- been explicitly set. This allows you to set the "network" logger to have INFO
- priority and unless the "network.interceptor" has had it's priority set it will
- inherit the INFO priority.
+ been explicitly set. This allows you to set the "network" logger to have INFO
+ priority and unless the "network.interceptor" has had it's priority set it will
+ inherit the INFO priority.
</p>
<p>
- Unlike other logging toolkits, there is no performance penalty for having deep
+ Unlike other logging toolkits, there is no performance penalty for having deep
hierarchies. Each logger caches a Priority to check against. When a logger
- has it's logger set or unset, it updates the cached version of it's child
+ has it's logger set or unset, it updates the cached version of it's child
loggers.
</p>
@@ -178,45 +178,45 @@
<section name="Log Targets">
<p>
- In LogKit, LogTargets are the destination of LogEvents. Decoupling LogEvent
+ In LogKit, LogTargets are the destination of LogEvents. Decoupling LogEvent
generation from handling allows developers to change destinations of LogEvents
dynamically or via configuration files. Possible destinations include writing
- to a database, a file, an IRC channel, a syslog server, an instant messaging
+ to a database, a file, an IRC channel, a syslog server, an instant messaging
client etc.
</p>
<p>
- Like Priorities, it is often useful to allow LogTargets to be inherited between
- loggers. Like Priority inheritance, LogTarget inheritance does not suffer any
+ Like Priorities, it is often useful to allow LogTargets to be inherited between
+ loggers. Like Priority inheritance, LogTarget inheritance does not suffer any
runtime performance penalty.
</p>
<subsection name="Filters">
<p>
- Filters are a special kind of LogTarget. Instead of writing to an output
- destination they are used to filter out LogEvents or modify the LogEvents
- details and pass it on to another LogTarget.
+ Filters are a special kind of LogTarget. Instead of writing to an output
+ destination they are used to filter out LogEvents or modify the LogEvents
+ details and pass it on to another LogTarget.
</p>
<p>
- This can be a useful feature when you assign multiple LogTargets to a
+ This can be a useful feature when you assign multiple LogTargets to a
logger. For instance you may add two LogTargets to a logger, one which
- writes LogEvents to a database and one which writes a message to the
- administrators pager. However you may want to record all messages to
- the database but only transmit FATAL_ERROR messages to pager. In this
+ writes LogEvents to a database and one which writes a message to the
+ administrators pager. However you may want to record all messages to
+ the database but only transmit FATAL_ERROR messages to pager. In this
case you would use a PriorityFilter to filter out non-FATAL_ERROR messages
- for pager log target.
+ for pager log target.
</p>
-
+
</subsection>
<subsection name="AsyncLogTarget">
<p>
AsyncLogTarget is another kind of special LogTarget. It takes a log event
- and copies it into a queue so that another thread can actually write the
- LogEvent to another LogTarget. This is useful if logging to a particular
+ and copies it into a queue so that another thread can actually write the
+ LogEvent to another LogTarget. This is useful if logging to a particular
LogTarget is a slow operation (such as a MailLogTarget).
</p>
@@ -236,39 +236,39 @@
</source>
</subsection>
- </section>
+ </section>
<section name="Formatters">
<p>
- LogTargets that write to a serial or unstructured store (ie filesystem or
- network based LogTargets) need some method to serialize the LogEvent
+ LogTargets that write to a serial or unstructured store (ie filesystem or
+ network based LogTargets) need some method to serialize the LogEvent
before writing to the store. The most common way to serialize the LogEvent
- is to use a Formatter.
+ is to use a Formatter.
</p>
<p>
The Formatter interface takes a LogEvent and returns a String object. The
- most commonly use LogEvent is the PatternFormatter. The pattern formatter
+ most commonly use LogEvent is the PatternFormatter. The pattern formatter
takes a format specifier that has a similar format to c's printf function.
</p>
<p>
- The format specifier consists of a string containing raw text combined with
- pattern elements. Each pattern element has the generalized form
- "%[+|-]#.#{field:subformat}". The +|- indicates whether the pattern element
+ The format specifier consists of a string containing raw text combined with
+ pattern elements. Each pattern element has the generalized form
+ "%[+|-]#.#{field:subformat}". The +|- indicates whether the pattern element
should be left or right justified (defaults to left justified if unspecified).
The #.# indicates the minimum and maximum size of output, if unspecified the
- output is neither padded nor truncated. 'field' indicates the field to be
- written and must be one of "category", "context", "message", "time",
- "rtime" (time relative to start of application), "throwable" or "priority".
- This parameter must be supplied and correlates to fields of LogEvent.
- 'subformat' is currently unused except in the case of "context" field. This is
+ output is neither padded nor truncated. 'field' indicates the field to be
+ written and must be one of "category", "context", "message", "time",
+ "rtime" (time relative to start of application), "throwable" or "priority".
+ This parameter must be supplied and correlates to fields of LogEvent.
+ 'subformat' is currently unused except in the case of "context" field. This is
further discussed below.
</p>
<p>
- Following is a number of examples for PatternFormatter's format specifier
+ Following is a number of examples for PatternFormatter's format specifier
and actual output.
</p>
@@ -287,8 +287,8 @@
<p>
There is also ExtendedPatternFormatter that allows two extra fields, namely
- "method" and "thread". The "method" field attempts to determine the method that
- called the Logger method to generate the LogEvent. The "thread" field displays
+ "method" and "thread". The "method" field attempts to determine the method that
+ called the Logger method to generate the LogEvent. The "thread" field displays
the name of the current thread.
</p>
@@ -299,25 +299,25 @@
<p>
In many systems you need to include extra information depending in logs that depends
on information not included in the LogEvent. For instance the Formatters section described
- an ExtendedPatternFormatter that included information such as calling method and calling
+ an ExtendedPatternFormatter that included information such as calling method and calling
thread. Other contextual information that you may need to include in log files include
- user executing log statement, the network interface that the client component is listening
- to (ie 127.0.0.1 vs 192.168.1.1), hostname (especially important on multihomed boxen) or
+ user executing log statement, the network interface that the client component is listening
+ to (ie 127.0.0.1 vs 192.168.1.1), hostname (especially important on multihomed boxs) or
source of LogEvent (useful when writing a centralized log server).
</p>
<p>
- There are a number of strategies to deal with application specific contextual
- information. Some logging toolkits encourage extending the Logger, LogEvent and
- LogTargets while others encourage using application specific LogTargets. The way
- that LogKit solves this problem is by using a generic ContexMap object.
+ There are a number of strategies to deal with application specific contextual
+ information. Some logging toolkits encourage extending the Logger, LogEvent and
+ LogTargets while others encourage using application specific LogTargets. The way
+ that LogKit solves this problem is by using a generic ContexMap object.
</p>
<p>
- The ContextMap allows the user to store arbitrary objects using a string key. These
+ The ContextMap allows the user to store arbitrary objects using a string key. These
objects can then be extracted by the LogTargets and used as appropriate. If you were
- using the PatternFormatter you could extract a value from ContextMap by specify the
- "context" field with it's subformat set to the appropriate key. Some examples are
+ using the PatternFormatter you could extract a value from ContextMap by specify the
+ "context" field with it's subformat set to the appropriate key. Some examples are
illustrated below.
</p>
@@ -337,24 +337,24 @@
<p>
<em>Warning:</em> Older versions of LogKit also incorporated a ContextStack
that offered hierarchical management of context. It was discovered that this
- design encouraged bad practices and thus use of this feature has been
+ design encouraged bad practices and thus use of this feature has been
deprecated in favour of using ContextMap. The ContextStack is still accessible
if users do not specify a subformat or specify the subformat "stack". However
- it is strongly advised that users do not use this feature as it will be
+ it is strongly advised that users do not use this feature as it will be
removed in a future iteration.
</p>
<p>
- When using a ContextMap you create the object and populate it using the
+ When using a ContextMap you create the object and populate it using the
set(key,value) method. After populating the ContextMap you call the method
makeReadOnly(). This makes it impossible for hostile code to modify context
in which logging occurs.
</p>
<p>
- Then you need to associate the ContextMap with a thread by using
+ Then you need to associate the ContextMap with a thread by using
ContextMap.bind( myContextMap ). The ContextMap is bound to a thread
- using java.lang.InheritableThreadLocal and thus maps are inherited between
+ using java.lang.InheritableThreadLocal and thus maps are inherited between
threads as specified by InheritableThreadLocal.
</p>
@@ -374,12 +374,12 @@
ContextMap.bind( context );
</source>
- </section>
+ </section>
<section name="Examples">
<p>
- One of the best ways to learn how to use a toolkit is to see an
+ One of the best ways to learn how to use a toolkit is to see an
example in action. With that heres some example uses of LogKit. The first
example is in a simple application while the next example is in a servlet.
The servlet example demonstrates the usage of filters.
@@ -402,7 +402,7 @@
final String path = context.getRealPath("/") + "/WEB-INF/logs/" + logName ;
-//Create a logger to write to a file as specified
+//Create a logger to write to a file as specified
//with servlet init parameters
final String pattern = "%7.7{priority} %5.5{time} [%8.8{category}] " +
"(%{context}): %{message}\\n%{throwable}";
@@ -431,15 +431,15 @@
<section name="Conclusion">
<p>
- LogKit is a friendly, easy to use logging toolkit. It is high performing and
+ LogKit is a friendly, easy to use logging toolkit. It is high performing and
easily integrated into existing products. By design it does not specify any
- configuration format but instead encourages users to integrate it into their
- existing products. It also is designed to run in a secure environment by
- limiting client access to hierarchies.
+ configuration format but instead encourages users to integrate it into their
+ existing products. It also is designed to run in a secure environment by
+ limiting client access to hierarchies.
</p>
</section>
</body>
-
+
</document>
---------------------------------------------------------------------
To unsubscribe, e-mail: cvs-unsubscribe@avalon.apache.org
For additional commands, e-mail: cvs-help@avalon.apache.org