test

Sun Java System Web Proxy Server 4.0.2 2005Q4 Configuration File Reference

Chapter 4 Syntax and Use of obj.conf

The obj.conf configuration file contains directives that instruct the SunTM Java SystemWeb Proxy Server how to handle HTTP and HTTPS requests from clients. You can modify and extend the request-handling process by adding or changing the instructions in obj.conf.

All obj.conf files are located in the instance_dir/config directory, where instance_dir is the path to the installation directory of the server instance.

By default, the obj.conf file for the server is named obj.conf.

This chapter discusses server instructions in obj.conf, the use of OBJECT tags, the use of variables, the flow of control in obj.conf, the syntax rules for editing obj.conf, and a note about example directives.

Note –

For detailed information about the standard directives and predefined Server Application Functions (SAFs) that are used in the obj.conf file, see Chapter 5, Predefined SAFs in obj.conf.

This chapter has the following sections:

How the Proxy Server Functions

”Proxy’ is a general term that means ”to act on behalf of a user in an authorized capacity’. A web proxy server intercepts client connections and obtains the requested content from an origin server (the owner of the content) on behalf of the client.

Typical web proxies accept connections from clients, make decisions as to whether or not the clients are permitted to use the proxy (or access the requested resources), and then completes connections on behalf of the clients to the various origin servers. In this manner, the web proxy acts as both a server as well as a client of the requested resource.

There are two basic types of web proxy servers: a forward proxy and a reverse proxy. While they share much of the same functionality, there are some definite differences between the two.

Forward Proxy Scenario

A forward proxy provides internal clients access through a firewall to resources on the Internet. This service is often provided as part of a larger intranet security strategy. Forward proxying allows clients to access resources outside of the firewall without compromising the integrity of the private network.

A forward proxy can be configured to keep copies of content within their local cache. Subsequent requests for that content can then be serviced from the local cache rather than obtaining it from the origin server. Caching increases performance by decreasing the time it takes to traverse the network.

Most proxy servers have the capability to filter requests from users. Administrators can choose to limit access to certain resources that may not be appropriate for the workplace and therefore deny such access.

In a forward proxy scenario, the client is aware of the proxy server and is configured to use it for various requests. The firewall can then be configured to only allow certain traffic from the proxy server rather than permitting such access to all internal clients.

Reverse Proxy Scenario

A proxy server can also provide external clients with access to internal resources the reside behind the corporate firewall. When a proxy server is used to handle connections into a private network, the process is called reverse proxying. The term, reverse, refers to the fact that traffic flows in the opposite direction than normal proxy traffic flow.

While forward proxies are best used to filter content, increase performance, and log user accesses, reverse proxies provide these benefits and more. A reverse proxy may be used to load balance across multiple servers, provide fail-over capabilities, and provide access to corporate resources in a safe and secure manner.

In a reverse proxy scenario, the client is not even aware that it is using a proxy server. This is one of the key differences between a forward and reverse proxy server scenario.

NSAPI Filters

The NSAPI API allows multiple Server Application Functions (SAFs) to interact in request processing. For example, one SAF could be used to authenticate the client after which a second SAF would generate the content.

Steps in the Request-handling Process

When the server first starts up it performs some initialization and then waits for a request from a client (such as a browser).

The obj.conf file for the server specifies how the request is handled in the following steps:

  1. Init

    The Init functions load and initialize server modules and plugins, and initialize log files.

  2. AuthTrans (authorization translation)

    Verify any authorization information (such as name and password) sent in the request.

  3. NameTrans (name translation)

    Translate the logical URI into a local file system path.

  4. PathCheck (path checking)

    Check the local file system path for validity and check that the requestor has access privileges to the requested resource on the file system.

  5. ObjectType (object typing)

    Determine the MIME-type (Multi-purpose Internet Mail Encoding) of the requested resource (for example, text/html, image/gif, and so on).

  6. Input (prepare to read input)

    Select filters that will process incoming request data read by the Service step.

  7. Output (prepare to send output)

    Select filters that will process outgoing response data generated by the Service step.

  8. Service (generate the response)

    Generate and return the response to the client.

  9. AddLog (adding log entries)

    Add entries to log file(s).

  10. Error (service)

    This step is executed only if an error occurs in the previous steps. If an error occurs, the server logs an error message and aborts the process.

  11. Connect

    Call the connect function you specify.

  12. DNS

    Call either the dns-config built-in function or a DNS function that you specify.

  13. Filter

    Run an external command and then pipe the data through the external command before processing that data in the proxy.

  14. Route

    Specify information about where the proxy server should route requests.

Directives for Handling Requests

The file obj.conf contains a series of instructions, known as directives, that tell the Sun Java System Web Proxy Server what to do at each stage in the request-handling process. Each directive invokes a Server Application Function (SAF) with one or more arguments. Each directive applies to a specific stage in the request-handling process. The stages are Init, AuthTrans, NameTrans, PathCheck, ObjectType, Input, Output, Service, AddLog, Connect, DNS, Filter, and Route..

Dynamic Reconfiguration

You do not need to restart the server for changes to certain configuration files to take effect (for example, obj.conf, mime.types, and server.xml,). All you need to do is apply the changes by clicking the Apply link and then clicking the Load Configuration Files button on the Apply Changes screen. If there are errors in installing the new configuration, the previous configuration is restored.

When you edit obj.conf and apply the changes, a new configuration is loaded into memory that contains all of the information from the dynamically configurable files.

Every new connection references the newest configuration. Once the last session referencing a configuration ends, the now unused old configuration is deleted.

Server Instructions in obj.conf

The obj.conf file contains directives that instruct the server how to handle requests received from clients such as browsers. These directives appear inside OBJECT tags.

Each directive calls a function, indicating when to call it and specifying arguments for it.

The syntax of each directive is:

Directive fn=func-name name1="value1"...nameN="valueN"

For example:

Init fn="flex-init" access="$accesslog" format.access="%Ses->client.ip% 
	- %Req->vars.auth-user% [%SYSDATE%] ’%Req->reqpb.clf-request%’ 
	%Req->srvhdrs.clf-status% %Req->srvhdrs.content-length%”

Directive indicates when this instruction is executed during the request-handling process. The value is one of Init, AuthTrans, NameTrans, PathCheck, ObjectType, Service, AddLog, Error, Connect, DNS, Filter, and Route.

The value of the fn argument is the name of the SAF to execute. All directives must supply a value for the fn parameter; if there’s no function, the instruction won’t do anything.

The remaining parameters are the arguments needed by the function, and they vary from function to function.

Sun Java System Web Proxy Server is shipped with a set of built-in Server Application Functions (SAFs) that you can use to create and modify directives in obj.conf.

Summary of the Directives

Following are the categories of server directives and a description of what each does. Each category corresponds to a stage in the request-handling process. The section Flow of Control in obj.confexplains exactly how the server decides which directive or directives to execute in each stage.

Configuring HTTP Compression

When compression is enabled in the server, an entry gets added to the obj.conf file. A sample entry is shown below:

Output fn="insert-filter" filter="http-compression" type="text/*"

Depending on the options specified, this line might also contain these options:

vary="on" compression-level="9"

To restrict compression to documents of only a particular type, or to exclude browsers that don’t work well with compressed content, you would need to edit the obj.conf file, as discussed below.

The option that appears as:

type="text/*"

restricts compression to documents that have a MIME type of text/* (for example, text/ascii, text/css, text/html, and so on). This can be modified to compress only certain types of documents. If you want to compress only HTML documents, for example, you would change the option to:

type="text/html"

Alternatively, you can specifically exclude browsers that are known to misbehave when they receive compressed content (but still request it anyway) by using the <Client> tag as follows:


<Client match="none"\\
 browser="*MSIE [1-3]*"\\
 browser="*MSIE [1-5]*Mac*"\\
 browser="Mozilla/[1-4]*Nav*">
Output fn="insert-filter" filter="http-compression" type="text/*"
</Client>

This restricts compression to browsers that are not any of the following:

Internet Explorer on Windows earlier than version 4 may request compressed data at times, but does not correctly support it. Internet Explorer on Macintosh earlier than version 6 does the same. Netscape Communicator version 4.x requests compression, but only correctly handles compressed HTML. It will not correctly handle linked CSS or JavaScript from the compressed HTML, so administrators often simply prevent their servers from sending any compressed content to that browser (or earlier).

For more information about the <Client> tag, see The Client Tag.

The Object and Client Tags

This section discusses the use of <Object> and <Client> tags in the file obj.conf.

<Object> tags group directives that apply to requests for particular resources, while <Client> tags group directives that apply to requests received from specific clients.

These tags are described in the following topics:

The Object Tag

Directives in the obj.conf file are grouped into objects that begin with an <Object> tag and end with an </Object> tag. The default object provides instructions to the server about how to process requests by default. Each new object modifies the default object’s behavior.

An Object tag may have a name attribute or a ppath attribute. Either parameter may be a wildcard pattern. For example:

<Object name="cgi">

- or -

<Object ppath="/usr/sun/webserver61/server1/docs/private/*">

The server always starts handling a request by processing the directives in the default object. However, the server switches to processing directives in another object after the NameTrans stage of the default object if either of the following conditions is true:

When the server has been alerted to use an object other than the default object, it processes the directives in the other object before processing the directives in the default object. For some steps in the process, the server stops processing directives in that particular stage (such as the Service stage) as soon as one is successfully executed, whereas for other stages the server processes all directives in that stage, including the ones in the default object as well as those in the additional object. For more details, see Flow of Control in obj.conf

Objects that Use the name Attribute

If a NameTrans directive in the default object specifies a name argument, the server switches to processing the directives in the object of that name before processing the remaining directives in the default object.

For example, the following NameTrans directive in the default object assigns the name cgi to any request whose URL starts with http://server_name/cgi/:


<Object name="default">
NameTrans fn="pfx2dir" from="/cgi" dir="
               <Install_Root>/
               <Instance_Directory>/mycgi" name="cgi"
...
</Object>

When that NameTrans directive is executed, the server starts processing directives in the object named cgi:


<Object name="cgi">

               more directives...</Object>

Objects that Use the ppath Attribute

When the server finishes processing the NameTrans directives in the default object, the logical URL of the request will have been converted to a physical path name. If this physical path name matches the ppath attribute of another object in obj.conf, the server switches to processing the directives in that object before processing the remaining ones in the default object.

For example, the following NameTrans directive translates the http://server_name/ part of the requested URL to <Install_Root>/<Instance_Directory>/docs (which is the document root directory):


<Object name="default">
NameTrans fn="document-root" root="<
               Install_Root>/<
               Instance_Directory>/docs"
...
</Object>

The URL http://server_name/internalplan1.html would be translated to <Install_Root>/<Instance_Directory>/docs/internalplan1.html. However, suppose that obj.conf contains the following additional object:


<Object ppath="*internal*">

               more directives...</Object>

In this case, the partial path *internal* matches the path <Install_Root>/<Instance_Directory>/docs/internalplan1.html. So now the server starts processing the directives in this object before processing the remaining directives in the default object.

The Client Tag

The <Client> tag is used to limit execution of a set of directives to requests received from specific clients. Directives listed between the <Client> and </Client> tags are executed only when information in the client request matches the parameter values specified.

Client Tag Parameters

The following table lists the <Client> tag parameters.

Table 4–1 Client Tag Parameters

Parameter  

Description  

browser

User-agent string sent by a browser to the Web Server 

chunked

Boolean value set by a client requesting chunked encoding 

code

HTTP response code 

dns

DNS name of the client 

internal

Boolean value indicating internally generated request 

ip

IP address of the client 

keep-alive

Boolean value indicating the client has requested a keep-alive connection 

keysize

Key size used in an SSL transaction 

match

Match mode for the <Client> tag; valid values are all, any, and none

method

HTTP method used by the browser 

name

Name of an object as specified in a previous NameTrans statement

odds

Sets a random value for evaluating the enclosed directive; specified as either a percentage or a ratio (for example, 20% or 1/5) 

path

Physical path to the requested resource 

ppath

Physical path of the requested resource 

query

Query string sent in the request 

reason

Text version of the HTTP response code 

restarted

Boolean value indicating a request has been restarted 

secret-keysize

Secret key size used in an SSL transaction 

security

Indicates an encrypted request 

type

Type of document requested (such as text/html or image/gif)

uri

URI section of the request from the browser 

urlhost

DNS name of the virtual server requested by the client (the value is provided in the Host header of the client request)

The <Client> tag parameters provide greater control over when and if directives are executed. In the following example, use of the odds parameter gives a request a 25% chance of being redirected:

<Client odds="25%">NameTrans fn="redirect" from="/Pogues" url-prefix="http://pogues.example.com"</Client>

One or more wildcard patterns can be used to specify Client tag parameter values.

Wildcards can also be used to exclude clients that match the parameter value specified in the <Client tag>. In the following example, the <Client> tag and the AddLog directive are combined to direct the Web Server to log access requests from all clients except those from the specified subnet:

<Client ip="~192.85.250.*">AddLog fn="flex-log" name="access"</Client>

Using the ~ wildcard negates the expression, so the Web Server excludes clients from the specified subnet.

You can also create a negative match by setting the match parameter of the Client tag to none. In the following example, access requests from the specified subnet are excluded, as are all requests to the server www.mycompany.com:

<Client match="none" ip="192.85.250.*" urlhost="www.mycompany.com">AddLog fn="flex-log" name="access"</Client>

Variables Defined in server.xml

You can define variables in the server.xml file and reference them in an obj.conf file. For example, the following server.xml code defines and uses a variable called docroot:

<!DOCTYPE SERVER SYSTEM "server.dtd" [
<!ELEMENT LS (DESCRIPTION?,SSLPARAMS?)>
<!ATTLIST LS
          id ID #REQUIRED
          ip CDATA "any"
          port CDATA #REQUIRED
          security %boolean; "false"
          acceptorthreads CDATA "1"
          family CDATA #IMPLIED
          blocking %boolean; "false"
          servername CDATA #REQUIRED
>

You can reference the variable in obj.conf as follows:

NameTrans fn=document-root root="$docroot"

Using this docroot variable saves you from having to define document roots for virtual server classes in the obj.conf files. It also allows you to define different document roots for different virtual servers within the same virtual server class.

Note –

Variable substitution is allowed only in an obj.conf file. It is not allowed in any other Sun Java System Web Proxy Server configuration files. Any variable referenced in an obj.conf file must be defined in the server.xml file.

Flow of Control in obj.conf

Before the server can process a request, it must direct the request to the correct server.

After the server is determined, the server executes the obj.conf file for the server. This section discusses how the server decides which directives to execute in obj.conf.

Init

The Init functions load and initialize server modules and plugins, and initialize log files.

AuthTrans

When the server receives a request, it executes the AuthTrans directives in the default object to check that the client is authorized to access the server.

If there is more than one AuthTrans directive, the server executes them all (unless one of them results in an error). If an error occurs, the server skips all other directives except for Error directives.

NameTrans

Next, the server executes a NameTrans directive in the default object to map the logical URL of the requested resource to a physical path name on the server’s file system. The server looks at each NameTrans directive in the default object in turn, until it finds one that can be applied.

If there is more than one NameTrans directive in the default object, the server considers each directive until one succeeds.

The NameTrans section in the default object must contain exactly one directive that invokes the document-root function. This functions translates the http://server_name/part of the requested URL to a physical directory that has been designated as the server’s document root. For example:

NameTrans fn="document-root" root="<Install_Root>/<Instance_Directory>/docs"

The directive that invokes document-root must be the last directive in the NameTrans section so that it is executed if no other NameTrans directive is applicable.

The pfx2dir (prefix to directory) function is used to set up additional mappings between URLs and directories. For example, the following directive translates the URL http://server_name/cgi/ into the directory path name <Install_Root>/<Instance_Directory>/docs/mycgi/:

NameTrans fn="pfx2dir" from="/cgi" dir="<Install_Root>/<Instance_Directory>/docs
	/mycgi"

Notice that if this directive appeared after the one that calls document-root, it would never be executed, with the result that the resultant directory path name would be <Install_Root>/<Instance_Directory>/docs/cgi/ (not mycgi). This illustrates why the directive that invokes document-root must be the last one in the NameTrans section.

How and When the Server Processes Other Objects

As a result of executing a NameTrans directive, the server might start processing directives in another object. This happens if the NameTrans directive that was successfully executed specifies a name or generates a partial path that matches the name or ppath attribute of another object.

If the successful NameTrans directive assigns a name by specifying a name argument, the server starts processing directives in the named object (defined with the OBJECT tag) before processing directives in the default object for the rest of the request-handling process.

For example, the following NameTrans directive in the default object assigns the name cgi to any request whose URL starts with http://server_name/cgi/.


<Object name="default">
...
NameTrans fn="pfx2dir" from="/cgi" dir="
               <Install_Root>/<
               Instance_Directory>/docs/mycgi" name="cgi"
...
</Object>

When that NameTrans directive is executed, the server starts processing directives in the object named cgi:


<Object name="cgi">

               more directives...</Object>

When a NameTrans directive has been successfully executed, there will be a physical path name associated with the requested resource. If the resultant path name matches the ppath (partial path) attribute of another object, the server starts processing directives in the other object before processing directives in the default object for the rest of the request-handling process.

For example, suppose obj.conf contains an object as follows:


<Object ppath="*internal*">

               more directives...</Object>

Now suppose the successful NameTrans directive translates the requested URL to the path name <Install_Root>/<Instance_Directory>/docs/internalplan1.html. In this case, the partial path *internal* matches the path <Install_Root>/<Instance_Directory>/docs/internalplan1.html. So now the server would start processing the directives in this object before processing the remaining directives in the default object.

PathCheck

After converting the logical URL of the requested resource to a physical path name in the NameTrans step, the server executes PathCheck directives to verify that the client is allowed to access the requested resource.

If there is more than one PathCheck directive, the server executes all of the directives in the order in which they appear, unless one of the directives denies access. If access is denied, the server switches to executing directives in the Error section.

If the NameTrans directive assigned a name or generated a physical path name that matches the name or ppath attribute of another object, the server first applies the PathCheck directives in the matching object before applying the directives in the default object.

ObjectType

Assuming that the PathCheck directives all approve access, the server next executes the ObjectType directives to determine the MIME type of the request. The MIME type has three attributes: type, encoding, and language. When the server sends the response to the client, the type, language, and encoding values are transmitted in the headers of the response. The type also frequently helps the server to determine which Service directive to execute to generate the response to the client.

If there is more than one ObjectType directive, the server applies all of the directives in the order in which they appear. However, once a directive sets an attribute of the MIME type, further attempts to set the same attribute are ignored. The reason that all ObjectType directives are applied is that one directive may set one attribute, for example type, while another directive sets a different attribute, such as language.

As with the PathCheck directives, if another object has been matched to the request as a result of the NameTrans step, the server executes the ObjectType directives in the matching object before executing the ObjectType directives in the default object.

Setting the Type By File Extension

Usually the default way the server figures out the MIME type is by calling the type-by-extension function. This function instructs the server to look up the MIME type according to the requested resource’s file extension in the MIME types table. This table was created during virtual server initialization by the MIME types file (which is usually called mime.types).

For example, the entry in the MIME types table for the extensions .html and.htm is usually:

type=text/html  exts=htm,html

which says that all files with the extension .htm or .html are text files formatted as HTML, and the type is text/html.

Note that if you make changes to the MIME types file, you must reconfigure the server before those changes can take effect.

Forcing the Type

If no previous ObjectType directive has set the type, and the server does not find a matching file extension in the MIME types table, the type still has no value even after type-by-expression has been executed. Usually if the server does not recognize the file extension, it is a good idea to force the type to be text/plain, so that the content of the resource is treated as plain text. There are also other situations where you might want to set the type regardless of the file extension, such as forcing all resources in the designated CGI directory to have the MIME type magnus-internal/cgi.

The function that forces the type is force-type.

For example, the following directives first instruct the server to look in the MIME types table for the MIME type, then if the type attribute has not been set (that is, the file extension was not found in the MIME types table), set the type attribute to text/plain.


ObjectType fn="type-by-extension"
ObjectType fn="force-type" type="text/plain"

If the server receives a request for a file abc.dogs, it looks in the MIME types table, does not find a mapping for the extension .dogs, and consequently does not set the type attribute. Since the type attribute has not already been set, the second directive is successful, forcing the type attribute to text/plain.

The following example illustrates another use of force-type. In this example, the type is forced to magnus-internal/cgi before the server gets a chance to look in the MIME types table. In this case, all requests for resources in http://server_name/cgi/ are translated into requests for resources in the directory <Install_Root>/<Instance_Directory>/docs/mycgi/. Since a name is assigned to the request, the server processes ObjectType directives in the object named cgi before processing the ones in the default object. This object has one ObjectType directive, which forces the type to be magnus-internal/cgi.


NameTrans fn="pfx2dir" from="/cgi" dir="<
               Install_Root>/<
               Instance_Directory>/docs/mycgi" name="cgi"
<Object name="cgi">
ObjectType fn="force-type" type="magnus-internal/cgi"
Service fn="send-cgi"
</Object>

The server continues processing all ObjectType directives including those in the default object, but since the type attribute has already been set, no other directive can set it to another value.

Input

The Input directive selects filters that will process incoming request data read by the Service step. It allows you to invoke the insert-filter SAF in order to install filters that process incoming data.

The Input directives are executed at most once per request.

You can define the appropriate position of a specific filter within the filter stack. For example, filters that translate content from XML to HTML are placed higher in the filter stack than filters that compress data for transmission. You can use the filter_create function to define the filter’s position in the filter stack, and init-filter-order to override the defined position.

When two or more filters are defined to occupy the same position in the filter stack, filters that were inserted later will appear higher than filters that were inserted earlier. That is, the order of Input fn="insert-filter" and Output fn="insert-filter" directives in obj.conf becomes important.

Output

The Output directive selects filters that will process outgoing response data generated by the Service step. The Output directive allows you to invoke the insert-filter SAF to install filters that process outgoing data. All Output directives are executed when the server or a plugin first attempts to write entity body data from the client.

The Output directives are executed at most once per request.

You can define the appropriate position of a specific filter within the filter stack. For example, filters that translate content from XML to HTML are placed higher in the filter stack than filters that compress data for transmission. You can use the filter_create function to define the filter’s position in the filter stack, init-filter-order to override the defined position.

When two or more filters are defined to occupy the same position in the filter stack, filters that were inserted later will appear higher than filters that were inserted earlier. That is, the order of Input fn="insert-filter" and Output fn="insert-filter" directives in obj.conf becomes important.

Service

Next, the server needs to execute a Service directive to generate the response to send to the client. The server looks at each Service directive in turn, to find the first one that matches the type, method and query string. If a Service directive does not specify type, method, or query string, then the unspecified attribute matches anything.

If there is more than one Service directive, the server applies the first one that matches the conditions of the request, and ignores all remaining Service directives.

As with the PathCheck and ObjectType directives, if another object has been matched to the request as a result of the NameTrans step, the server considers the Service directives in the matching object before considering the ones in the default object. If the server successfully executes a Service directive in the matching object, it will not get around to executing the Service directives in the default object, since it only executes one Service directive.

Service Examples

For an example of how Service directives work, consider what happens when the server receives a request for the URL D:/server_name/jos.html. In this case, all directives executed by the server are in the default object.


Service method="(GET|HEAD|POST)" type="*~magnus-internal/*" fn="send-file""

Here is an example that involves using another object:


NameTrans fn=assign-name name=personnel from=/personnel

            

<Object name="personnel">
 Service fn="index-simple"
 </Object>

Default Service Directive

There is usually a Service directive that does the default task (sends a file) if no other Service directive matches a request sent by a browser. This default directive should come last in the list of Service directives in the default object, to ensure it only gets called if no other Service directives have succeeded. The default Service directive is usually:


Service method="(GET|HEAD|POST)" type="*~magnus-internal/*" fn="send-file"

This directive matches requests whose method is GET, HEAD, or POST, which covers nearly virtually all requests sent by browsers. The value of the type argument uses special pattern-matching characters.

The characters “*~” mean “anything that doesn’t match the following characters,” so the expression *~magnus-internal/ means “anything that doesn’t match magnus-internal/.” An asterisk by itself matches anything, so the whole expression *~magnus-internal/* matches anything that does not begin with magnus-internal/.

So if the server has not already executed a Service directive when it reaches this directive, it executes the directive so long as the request method is GET, HEAD or POST, and the value of the type attribute does not begin with magnus-internal/. The invoked function is send-file, which simply sends the contents of the requested file to the client.

AddLog

After the server generate the response and sends it to the client, it executes AddLog directives to add entries to the log files.

All AddLog directives are executed. The server can add entries to multiple log files.

Depending on which log files are used and which format they use, the Init section in magnus.conf may need to have directives that initialize the logs. For example, if one of the AddLog directives calls flex-log, which uses the extended log format, the Init section must contain a directive that invokes flex-init to initialize the flexible logging system.

For more information about initializing logs, see the discussion of the functions flex-init and init-clf in Chapter 5, Predefined SAFs in obj.conf.

Error

If an error occurs during the request-handling process, such as if a PathCheck or AuthTrans directive denies access to the requested resource, or the requested resource does not exist, the server immediately stops executing all other directives and immediately starts executing the Error directives.

Connect

The Connect directive calls the connect function you specify.

Only the first applicable Connect function is called, starting from the most restrictive object. Occasionally it is desirable to call multiple functions (until a connection is established). The function returns REQ_NOACTION if the next function should be called. If it fails to connect, the return value is REQ_ABORT. If it connects successfully, the connected socket descriptor will be returned.

DNS

The DNS directive calls either the dns-config built-in function or a DNS function that you specify.

Filter

The Filter directive runs an external command and then pipes the data through the external command before processing that data in the proxy. This is accomplished using the pre-filter function.

Route

The Route directive specifies information about where the proxy server should route requests.

Changes in Function Flow

There are times when the function flow changes from the normal request-handling process. This happens during internal redirects, restarts, and URI translation functions.

Internal Redirects

An example of an internal redirect is a servlet include or forward. In this case, because there is no exposed NSAPI function to handle an internal redirect, when an internal redirect occurs, the request structure is copied into rq->orig_rq.

Restarts

A restart occurs when a REQ_RESTART is returned from a PathCheck or Service function. For example, when a CGI is redirected using a relative path.

On a restart, much of the request is cleared. Some elements of the HTTP request (rq->reqpb), the server’s “working” variables (rq->vars), and response headers (rq->srvhdrs) are cleared. The method, protocol, and clf-request variables from rq->reqpb are saved. The saved variables are put back into the data structure. The new URI is inserted (and if there is a query string in the new URI, that too is inserted) into rq->reqpb. The parameter rq->rq_attr.req_restarted is set to 1.

URI Translation

At times it is necessary to find the physical path for a URI without actually running a request. The function request_translate_uri does this. A new request structure is created and run through the AuthTrans and NameTrans stages to get the physical path. Thereafter, the new request is freed.

Syntax Rules for Editing obj.conf

Several rules are important in the obj.conf file. Be very careful when editing this file. Simple mistakes can make the server fail to start or operate correctly.

Caution – Caution –

Do not remove any directives from any obj.conf file that are present in the obj.conf file that exists when you first install Sun Java System Web Proxy Server. The server may not function properly.

Order of Directives

The order of directives is important, since the server executes them in the order they appear in obj.conf. The outcome of some directives affect the execution of other directives.

For PathCheck directives, the order within the PathCheck section is not so important, since the server executes all PathCheck directives. However, the order within the ObjectType section is very important, because if an ObjectType directive sets an attribute value, no other ObjectType directive can change that value. For example, if the default ObjectType directives were listed in the following order (which is the wrong way around), every request would have its type value set to text/plain, and the server would never have a chance to set the type according to the extension of the requested resource.


ObjectType fn="force-type" type="text/plain"
ObjectType fn="type-by-extension"

Similarly, the order of directives in the Service section is very important. The server executes the first Service directive that matches the current request and does not execute any others.

Parameters

The number and names of parameters depends on the function. The order of parameters on the line is not important.

Case Sensitivity

Items in the obj.conf file are case-sensitive including function names, parameter names, many parameter values, and path names.

Separators

The C language allows function names to be composed only of letters, digits, and underscores. You may use the hyphen (-) character in the configuration file in place of underscore (_) for your C code function names. This is only true for function names.

Quotes

Quotes (“) are only required around value strings when there is a space in the string. Otherwise they are optional. Each open-quote must be matched by a close-quote.

Spaces

Line Continuation

A long line may be continued on the next line by beginning the next line with a space or tab.

Path Names

Always use forward slashes (/) rather than backslashes (\\) in path names under Windows. Backslash escapes the next character.

Comments

Comments begin with a pound (#) sign. If you manually add comments to obj.conf, then use the Server Manager interface to make changes to your server, the Server Manager will wipe out your comments when it updates obj.conf.

About obj.conf Directive Examples

Every line in the obj.conf file begins with one of the following keywords:

Init
AuthTrans
 NameTrans
 PathCheck
 ObjectType
 InputOutputService
 AddLog
 Error
 Connect
DNS
Filter
Route
<Object>
 </Object>

If any line of any example begins with a different word in the manual, the line is wrapping in a way that it does not in the actual file. In some cases this is due to line length limitations imposed by the PDF and HTML formats of the manuals.

For example, the following directive is all on one line in the actual obj.conf file:

Init fn="flex-init" access="$accesslog" format.access="%Ses->client.ip% 
	- %Req->vars.auth-user% [%SYSDATE%] ’%Req->reqpb.clf-request%’ 
	%Req->srvhdrs.clf-status% %Req->srvhdrs.content-length%"