If Spring MVC is present in the classpath, use MvcRequestMatcher by default. This commit also adds a new securityMatcher method in HttpSecurity
Closes gh-11347
Closes gh-9159
OWASP recommends using "X-Xss-Protection: 0". The default is currently
"X-Xss-Protection: 1; mode=block". In 6.0, the default will be "0".
This commits adds the ability to configure the xssProtection header
value in ServerHttpSecurity.
This commit deprecates the use of "enabled" and "block" booleans to
configure XSS protection, as the state "!enabled + block" is invalid.
This impacts HttpSecurity.
Issue gh-9631
In 6.0, RequestAttributeSecurityContextRepository will be the default
implementation of SecurityContextRepository. This commit adds the
ability to configure a custom SecurityContextHolderStrategy, similar
to other components.
Issue gh-11060
Closes gh-11895
There is no whitespace between error message and IP address value `IpAddressMatcher#parseAddress()`
If IP value is wrong, then error text looks like `Failed to parse addressi.am.ip`.
There should be some separator between those two text tokens.
Also wrapped the address value with single quotes.
Will this add any confusion for the caller?
Or colon and `"Failed to parse address: $value` looks better?
Previously the CsrfToken was set on the request attribute with the name
equal to CsrfToken.getParameterName(). This didn't really make a lot of
sense because the CsrfToken.getParameterName() is intended to be used as
the HTTP parameter that the CSRF token was provided. What's more is it
meant that the CsrfToken needed to be read for every request to place it
as an HttpServletRequestAttribute. This causes unnecessary HttpSession
access which can decrease performance for applications.
This commit allows setting CsrfFilter.csrfReqeustAttributeName to
remove the dual purposing of CsrfToken.parameterName and to allow deferal
of reading the CsrfToken to prevent unnecessary HttpSession access.
Issue gh-11699
Previously LazyCsrfTokenRepository supported lazily saving the CsrfToken
which allowed for lazily saving the CsrfToken. However, it did not
support lazily reading the CsrfToken. This meant every request required
reading the CsrfToken (often the HttpSession).
This commit allows for lazily reading the CsrfToken and thus prevents
unnecessary reads to the HttpSession.
Closes gh-11700
Previously the CsrfToken was set on the request attribute with the name
equal to CsrfToken.getParameterName(). This didn't really make a lot of
sense because the CsrfToken.getParameterName() is intended to be used as
the HTTP parameter that the CSRF token was provided. What's more is it
meant that the CsrfToken needed to be read for every request to place it
as an HttpServletRequestAttribute. This causes unnecessary HttpSession
access which can decrease performance for applications.
This commit allows setting CsrfFilter.csrfReqeustAttributeName to
remove the dual purposing of CsrfToken.parameterName and to allow deferal
of reading the CsrfToken to prevent unnecessary HttpSession access.
Issue gh-11699
Previously LazyCsrfTokenRepository supported lazily saving the CsrfToken
which allowed for lazily saving the CsrfToken. However, it did not
support lazily reading the CsrfToken. This meant every request required
reading the CsrfToken (often the HttpSession).
This commit allows for lazily reading the CsrfToken and thus prevents
unnecessary reads to the HttpSession.
Closes gh-11700