Heh. You and I have a different understanding of silent! I meant that the base64 decoder doesn't actually complain if the base64 is malformed by either the insertion of extraneous characters, or by the omission of the ending pad characters if any.
Later, Mike At 12:22 PM 2/15/2013, Xueming Shen wrote: >On 2/15/13 8:52 AM, Michael StJohns wrote: >> >>Is the "mime" variant of Base64 the correct one for this? I ask because that >>variant ignores extraneous characters rather than throwing an error on >>decode. Also, reading the code for the Base64 implementation, it silently >>"fixes" the case where there are missing padding "=" characters. Neither of >>these seem ideal for security related processing. > >Just want to point out that the latest Base64.Decoder spec has been updated to >"explicitly" says > >"The Base64 padding character '=' is accepted and interpreted as the end of >the encoded byte data, but is not required. So if the final unit of the >encoded byte data only has two or three Base64 characters (without the >corresponding padding character(s) padded), they are decoded as if followed by >padding character(s). " > >So it's no longer "silently" anymore:-) > >-Sherman > >> >> >> >>It may be reasonable to add a PEM variant to the Base64 code that deals with >>the above. >> >>Mike >> >> >> >>At 08:24 AM 2/14/2013, Mark Sheppard wrote: >>> >>>Hi, >>> as part of a refactoring of the jdk codebase to use the base64 >>> capabilities of java.util.Base64, the following modifications, >>>as per the webrev, >>> >>><http://cr.openjdk.java.net/~chegar/8006182/webrev.00/>http://cr.openjdk.java.net/~chegar/8006182/webrev.00/ >>> >>>have been made to complete task JDK-8006182. >>> >>>Could you oblige and review these changes, please? >>> >>>Description: >>>jdk8 has java.util.Base64 to define a standard API for base64 >>>encoding/decoding. It would be good to investigate whether this API could be >>>used in the security components, providers and regression tests. >>> >>>In the main this work involved replacing the sun.misc.BASE64Encoder and >>>sun.misc.BASE64Decoder with the >>>corresponding Mime Base64 Encoder/Decoder (as per rfc2045) from the >>>java.util.Base64 class. >>>This is a like for like replacement. >>>As such, sun.misc.BASE64Encoder maps to the encoder returned by >>>java.util.Base64.getMimeEncoder() >>>sun.misc.BASE64Decoder maps to the decoder returned by >>>java.util.Base64.getMimeDecoder() >>> >>>However a couple of items worth noting: >>> >>>In the jarsigner (Main.java) the standard Base64 encoder (rfc 4648), >>>java.util.Base64.getEncoder(), has been used to replace the >>>JarBASE64Encoder, which was a package private extension of BASE64Encoder, >>>which avoids writing newline to the encoded data. >>> >>>In the keytool (Main.java), methods such as dumpCert, printCert. printCRL, >>>and so on, write a Base64 encoding to an OutputStream, typically std out. >>>This is achieved in the BASE64Encoder, by passing the OutputStream to >>>methods such as encodeBuffer(). >>> >>>A couple of options exist to do this under the new Base64 utilities, which >>>include: >>> >>>* using a Mime Encoder encodeToString() and output to the stream via >>>println() >>> >>>* use the wrap capabilities of the Base64.Encoder: >>> - define a package private class, which extends FilterOutputStream (e.g. >>> NoCloseWrapperOutputStream) and, overrides close() to do nothing >>> - inject the OutputStream, passed to the keytool method, into the >>> NoCloseWrapperOutputStreamwapper, >>> - wrap() the NoCloseWrapperOutputStreamwrapper in the Mime Encoder, which >>> will in turn return an encapsulating OutputStream; >>> - write the data buffer to be encoded to the encoder's OutputStream; >>> - close the encoder's OutputStream, which completes the base64 encoding; >>> - append a newline to the initial OutputStream. >>> >>>pragmatics and the simplest thing that works, went for the first option. >>> >>>regards >>>Mark >>> >>> >>> >>> >>> >> >> >>
