Re: [PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-24 Thread Stefan Esser
Hello Jani, I happily repeat myself until you actually read my comments. There is a cookie 0 format defined by Netscape - no quoted strings and only ; as separator and a cookie 1 format defined by RFC 2109/2965 - with quoted_strings - and with , and ; as separator PHP understood up to today only Ne

Re: [PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-24 Thread Jani Taskinen
On Sun, 24 Apr 2005, Stefan Esser wrote: Forgive me my ignorance, but I do not see any handling of " chars. And there wasn't such before I added the , as acceptable separator. ^^ I hate when I have to repeat myself, so read the above line again.. Yes because PHP spoke cookie version 0 before

Re: [PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-24 Thread Stefan Esser
Hi Jani, 'value' can NOT contain any of these chars: ,; \t\r\n\013\014 Wrong. value is to the RFC either a token or a quoted_string and a quoted_string can contain , and ; If you want such chars in them, you have to encode them. Yes in your implementation that is not RFC conform Forgive

Re: [PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-24 Thread Jani Taskinen
On Sun, 24 Apr 2005, Stefan Esser wrote: As now it handles these strings it gets the same: name="whatever,this,might,be";name2="value2" name="whatever,this,might,be",name2="value2" These were actually invalid examples. 'name' can NOT contain any of these chars: =,; \t\r\n\0

Re: [PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-23 Thread Stefan Esser
As now it handles these strings it gets the same: name="whatever,this,might,be";name2="value2" name="whatever,this,might,be",name2="value2" How can this affect modsecurity at all Forgive me my ignorance, but I do not see any handling of " chars. your strings should now resu

[PHP-DEV] Re: [PHP-CVS] cvs: php-src /main php_variables.c

2005-04-23 Thread Jani Taskinen
This patch was supposed to make it RFC compliant. Can you please point me the part in the RFC 2965 which now isn't implemented correctly? As now it handles these strings it gets the same: name="whatever,this,might,be";name2="value2" name="whatever,this,might,be",name2="v