24 FTP, HTTP, and WebDAV Access to Repository Data

HTTP Performance is Improved

Session Pooling improves performance of HTTP by avoiding the cost of re-creating session states, especially when using HTTP 1.0, which creates new connections for each request. For example, a couple of small files can be retrieved by an existing HTTP/1.1 connection in the time necessary to create a database session. You can tune the number of sessions in the pool by setting session-pool-size in Oracle XML DB xdbconfig.xml file, or disable it by setting pool size to zero.

Java Servlets

Session pooling can affect users writing Java servlets, because other users can come along and see session state initialized by another request for a different user. Hence, servlet writers should only use session memory, such as Java static variables, to hold data for the entire application rather than for a particular user. State for each user must be stored in the database or in a look-up table, rather than assuming that a session will only exist for a single user.

Figure 24-1 illustrates the Oracle XML DB protocol server components and how they are used to access files in Oracle XML DB XML repository and other data. Only the relevant components of the repository are shown

Figure 24-1 Oracle XML DB Architecture: Protocol Server

Description of the illustration adxdb019.gif

Non-Supported FTP Features

Oracle XML DB implements FTP, as defined by RFC 959, with the exception of the following optional features:

• Record-oriented files, for example, only the FILE structure of the STRU command is supported. This is the most widely used structure for transfer of files. It is also the default specified by the specification. Structure mount is not supported.

• Append.

• Allocate. This pre-allocates space before file transfer.

• Account. This uses the insecure Telnet protocol.

• Abort.

Using FTP on Standard or Non-Standard Ports

It can be configured through the Oracle XML DB configuration file /xdbconfig.xml, to listen on an arbitrary port. FTP ships listening on a non-standard, non-protected port. To use FTP on the standard port (21), your DBA has to chown the TNS listener to setuid ROOT rather than setuid ORACLE.

FTP Server Session Management

Protocol server also provides session management for this protocol. After a short wait for a new command, FTP returns to the protocol layer and the shared server is freed up to serve other connections. The duration of this short wait is configurable by changing the call-timeOut parameter in the Oracle XML DB configuration file. For high traffic sites, the call-timeout should be shorter so that more connections can be served. When new data arrives on the connection, the FTP Server is re-invoked with fresh data. So, the long running nature of FTP does not affect the number of connections which can be made to the protocol server.

Controlling Character Sets for FTP

Oracle Database supports two FTP quote commands to control character sets for different purposes: set_nls_locale and set_charset.

• set_nls_locale

  quote set_nls_locale {<charset_name> | NULL}

  This command is used to control the encoding of the file and directory names specified by the users in the FTP commands. It also controls the encoding of the file and directory names in the response returned to the users. Only IANA character set names can be specified for this parameter. If nls_locale is set to NULL or not set then it is defaulted to the database character set.

• set_charset

  quote set_charset  {<charset_name> | NULL}

  This command is used to specify the character set of the data to be sent to the server. This parameter, if defined, overrides the Byte Order Mark (BOM) and the encoding declaration inside the document. The keyword NULL is used to unset the charset parameter. The BOM is a signature to indicate the order of the following stream of bytes defined in the Unicode Standard.

  The algorithm used to determine the character encoding of incoming data is as follows:

  • The charset parameter value, if it is not NULL, determines the character set.

  • If the charset parameter value is NULL, the MIME type of the data is evaluated.

    If the MIME type is */xml then the character set is determined by the presence of the BOM and the encoding declaration inside the XML document. Otherwise, the database character set is used.

  • Text documents are assumed to be in the database character set if the set_charset command is not set. This parameter does not apply to binary files that are not text.

Handling Error 421. Modifying the FTP Session's Default Timeout Value

If you are frequently disconnected from the server and have to reconnect and traverse the entire directory before doing the next operation, you may need to modify the default timeout value for FTP sessions. If the session is idle for more than this period, it gets disconnected. You can increase the timeout value (default = 6000 centiseconds) by modifying the configuration document as follows and then restart the database:

declare
newconfig XMLType;
begin
  select updatexml(
    dbms_xdb.cfg_get(),
'/xdbconfig/sysconfig/protocolconfig/ftpconfig/session-timeout/text()',
    123456789) into newconfig from dual;
  dbms_xdb.cfg_update(newconfig);
end;/
 
commit;

Non-Supported HTTP Features

Digest Authentication (RFC 2617) is not supported. In this release, Oracle XML DB supports Basic Authentication, where a client sends the user name and password in clear text in the Authorization header.

Using HTTP on Standard or Non-Standard Ports

HTTP ships listening on a non-standard, non-protected port (8080). To use HTTP on the standard port (80), your DBA must chown the TNS listener to setuid ROOT rather than setuid ORACLE, and configure the port number in the Oracle XML DB configuration file /xdbconfig.xml.

HTTP Server and Java Servlets

Oracle XML DB supports Java servlets. To use a servlet, it must be registered with a unique name in the Oracle XML DB configuration file, along with parameters to customize its action. It should be compiled, and loaded into the database. Finally, the servlet name must be associated with a pattern, which can be an extension such as *.jsp or a path name such as /a/b/c or /sys/*, as described in Java servlet application program interface (API) version 2.2.

While processing an HTTP request, the path name for the request is matched with the registered patterns. If there is a match, then the protocol server invokes the corresponding servlet with the appropriate initialization parameters. For Java servlets, the existing Java Virtual Machine (JVM) infrastructure is used. This starts the JVM if need be, which in turn runs a Java method to initialize the servlet, create response, and request objects, pass these on to the servlet, and run it.

Sending Multibyte Data From a Client

When a client sends multibyte data in a URL, RFC 2718 specifies that the client should send the URL using the %HH format where HH is the hexadecimal notation of the byte value in UTF-8 encoding. The following are URL examples that can be sent to XML DB in an HTTP or WebDAV context:

http://urltest/xyz%E3%81%82%E3%82%A2 
http://%E3%81%82%E3%82%A2 
http://%E3%81%82%E3%82%A2/abc%E3%81%86%E3%83%8F.xml

XML DB processes the requested URL, any URLs within an IF header, any URLs within the DESTINATION header, and any URLs in the REFERRED header that contains multibyte data.

The default-url-charset configuration parameter can be used to accept requests from some clients that use other non-conformant forms of URL of non-ASCII characters. If a request with non-ASCII characters fails, try setting this value to the native character set of the client environment. The character set used in such URL fields must be specified with an IANA charset name.

default-url-charset controls the encoding for non-conforming URLs. It is not required to be set unless a non-conforming client that does not send the Content-Type charset is used.

Non-Ascii Characters in URLs

Non-ascii characters appearing in URLs passed to an HTTP Server should be converted to UTF-8 and escaped in the %HH format, where HH is the hexadecimal notation of the byte value. For flexibility, XML DB protocol server interprets the incoming URLs by testing whether it is encoded in one of the following character sets in the order presented here:

• UTF-8

• Charset parameter of the Content-Type field of the request if specified

• Character set if specified in the default-url-charset configuration parameter

• Character set of the database server

Controlling Character Sets for HTTP

The following sections describe how character sets are controlled for data transferred using HTTP.

Non-Supported WebDAV Features

Oracle XML DB supports the contents of RFC2518, with the following exceptions:

• Lock-NULL resources create actual zero-length resources in the file system, and cannot be converted to folders.

• The COPY, MOVE and DELETE methods comply with section 2 of the Internet Draft titled 'Binding Extensions to WebDAV'.

• Depth-infinity locks

• Only Basic Authentication is supported.