123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131 |
- John,
- I have been an extremely satisfied ADODB user for several years now.
- To give you something back for all your hard work, I've spent the last 3
- days rewriting the adodb-session.php code.
- ----------
- What's New
- ----------
- Here's a list of the new code's benefits:
- * Combines the functionality of the three files:
- adodb-session.php
- adodb-session-clob.php
- adodb-cryptsession.php
- each with very similar functionality, into a single file adodb-session.php.
- This will ease maintenance and support issues.
- * Supports multiple encryption and compression schemes.
- Currently, we support:
- MD5Crypt (crypt.inc.php)
- MCrypt
- Secure (Horde's emulation of MCrypt, if MCrypt module is not available.)
- GZip
- BZip2
- These can be stacked, so if you want to compress and then encrypt your
- session data, it's easy.
- Also, the built-in MCrypt functions will be *much* faster, and more secure,
- than the MD5Crypt code.
- * adodb-session.php contains a single class ADODB_Session that encapsulates
- all functionality.
- This eliminates the use of global vars and defines (though they are
- supported for backwards compatibility).
- * All user defined parameters are now static functions in the ADODB_Session
- class.
- New parameters include:
- * encryptionKey(): Define the encryption key used to encrypt the session.
- Originally, it was a hard coded string.
- * persist(): Define if the database will be opened in persistent mode.
- Originally, the user had to call adodb_sess_open().
- * dataFieldName(): Define the field name used to store the session data, as
- 'DATA' appears to be a reserved word in the following cases:
- ANSI SQL
- IBM DB2
- MS SQL Server
- Postgres
- SAP
- * filter(): Used to support multiple, simulataneous encryption/compression
- schemes.
- * Debug support is improved thru _rsdump() function, which is called after
- every database call.
- ------------
- What's Fixed
- ------------
- The new code includes several bug fixes and enhancements:
- * sesskey is compared in BINARY mode for MySQL, to avoid problems with
- session keys that differ only by case.
- Of course, the user should define the sesskey field as BINARY, to
- correctly fix this problem, otherwise performance will suffer.
- * In ADODB_Session::gc(), if $expire_notify is true, the multiple DELETES in
- the original code have been optimized to a single DELETE.
- * In ADODB_Session::destroy(), since "SELECT expireref, sesskey FROM $table
- WHERE sesskey = $qkey" will only return a single value, we don't loop on the
- result, we simply process the row, if any.
- * We close $rs after every use.
- ---------------
- What's the Same
- ---------------
- I know backwards compatibility is *very* important to you. Therefore, the
- new code is 100% backwards compatible.
- If you like my code, but don't "trust" it's backwards compatible, maybe we
- offer it as beta code, in a new directory for a release or two?
- ------------
- What's To Do
- ------------
- I've vascillated over whether to use a single function to get/set
- parameters:
- $user = ADODB_Session::user(); // get
- ADODB_Session::user($user); // set
- or to use separate functions (which is the PEAR/Java way):
- $user = ADODB_Session::getUser();
- ADODB_Session::setUser($user);
- I've chosen the former as it's makes for a simpler API, and reduces the
- amount of code, but I'd be happy to change it to the latter.
- Also, do you think the class should be a singleton class, versus a static
- class?
- Let me know if you find this code useful, and will be including it in the
- next release of ADODB.
- If so, I will modify the current documentation to detail the new
- functionality. To that end, what file(s) contain the documentation? Please
- send them to me if they are not publically available.
- Also, if there is *anything* in the code that you like to see changed, let
- me know.
- Thanks,
- Ross
|