charset is a wrapper around encoding_rs that provides
(non-streaming) decoding for character encodings that occur in email by
providing decoding for UTF-7 in addition to the encodings defined by
the Encoding Standard (and provided by encoding_rs).
Note: Do not use this crate for consuming Web content. For security
reasons, consumers of Web content are prohibited from supporting
UTF-7. Use encoding_rs directly when consuming Web content.
The set of encodings consisting of UTF-7 and the encodings defined in the
Encoding Standard is believed to be appropriate for consuming email,
because that's the set of encodings supported by Thunderbird.
Furthermore, UTF-7 support is believed to be necessary based on the
experience of the Firefox OS email client. In fact, while the UTF-7
implementation in this crate is independent of Thunderbird's UTF-7
implementation, Thunderbird uses encoding_rs to decode the other
encodings. The set of labels/aliases recognized by this crate
matches those recognized by Thunderbird.
Known compatibility limitations (shared with Thunderbird and known from Thunderbird bug reports):
This crate intentionally does not support encoding content into legacy
encodings. When sending email, always use UTF-8. This is, just call
.as_bytes() on &str and label the content as UTF-8.
Please see the file named COPYRIGHT.
Generated API documentation is available online.
Again, this crate is for email. Please do NOT use it for Web content.
Never try to perform any security analysis on the undecoded data in ASCII-incompatible encodings and in UTF-7 in particular. Always decode first and analyze after. UTF-7 allows even characters that don't have to be represeted as base64 to be represented as base64. Also, for consistency with Thunderbird, the UTF-7 decoder in this crate allows e.g. ASCII controls to be represented without base64 encoding even when the spec says they should be base64-encoded.
This implementation is non-constant-time by design. An attacker who can observe input length and the time it takes to decode it can make guesses about relative proportions of characters from different ranges. Guessing the proportion of ASCII vs. non-ASCII should be particularly feasible.
This is a personal project. It has a Mozilla copyright notice, because I copied and pasted from encoding_rs. You should not try to read anything more into Mozilla's name appearing.
decode_ascii().decode_latin1().Initial release.