A slight change was made in the 1.9 series to modify the default usage of the Editor dijit to use div tags instead of a textarea tag; the latter usage has security implications, and usage of div tags is recommended by the Dojo project.

In order to still allow graceful degradation, a new degrade option was added to the view helper; this would allow developers to optionally use a textarea instead. However, this opens applications developed with that usage to XSS vectors. In 1.9.7, we have removed this option. Graceful degradation is still supported, however, via a noscript tag that embeds a textarea. This solution addressess all security concerns.

The takeaway is that if you were using the degrade flag, it will simply be ignored at this time.

In order to default to a more secure character encoding, Zend_Filter_HtmlEntities now defaults to UTF-8 instead of ISO-8859-1.

Additionally, because the actual mechanism is dealing with character encodings and not character sets, two new methods have been added, setEncoding() and getEncoding(). The previous methods setCharSet() and setCharSet() are now deprecated and proxy to the new methods. Finally, instead of using the protected members directly within the filter() method, these members are retrieved by their explicit accessors. If you were extending the filter in the past, please check your code and unit tests to ensure everything still continues to work.

Zend_Filter_StripTags contains a flag, commentsAllowed, that, in previous versions, allowed you to optionally whitelist HTML comments in HTML text filtered by the class. However, this opens code enabling the flag to XSS attacks, particularly in Internet Explorer (which allows specifying conditional functionality via HTML comments). Starting in version 1.9.7 (and backported to versions 1.8.5 and 1.7.9), the commentsAllowed flag no longer has any meaning, and all HTML comments, including those containing other HTML tags or nested commments, will be stripped from the final output of the filter.