[ https://issues.apache.org/jira/browse/CODEC-333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18004618#comment-18004618 ]
Aleksandr Beliakov commented on CODEC-333: ------------------------------------------ Hi [~ggregory] , Considering the ticket CODEC-75 has been approved "as a feature", it is difficult to categorize it as a bug on my own, even though the behavior is not completely correct to my best understanding. I think that the implementation should contain two separate *decoding* tables, similarly as it is done for *encoding* for [standard base64|https://github.com/apache/commons-codec/blob/master/src/main/java/org/apache/commons/codec/binary/Base64.java#L121] and [url-safe base64|https://github.com/apache/commons-codec/blob/master/src/main/java/org/apache/commons/codec/binary/Base64.java#L133]. Currently there is only [one|https://github.com/apache/commons-codec/blob/master/src/main/java/org/apache/commons/codec/binary/Base64.java#L154] that merges both codings. If properly done, there should be two separate methods available for each decoding and verifying base64 validity content functions. I see two ways to implement the feature: * Modify behavior within the current _Base64.isBase64_ and _Base64.decodeBase64_ methods, to verify the content against the "standard base64" specification only, instead of a merged one; and create two new methods such as _Base64.isBase64url_ and _Base64.decodeBase64url_ that will verify the content against the Base64Url coding specification. This modification will introduce some breaking changes in the API, as users relying on a url-safe processing with the old methods will need to adapt the code; or * Keep existing methods as they are, containing a "merged" behavior, and introduce new methods for each coding separately, such as _Base64.isStandardBase64_ and _Base64.decodeStandardBase64_ and _Base64.isBase64url_ and {_}Base64.decodeBase64url{_}. This will avoid changes in the existing API, but will provide new methods for those who want to ensure a strict behavior. The only thing I'm not sure about is what behavior to apply on [Builder#setUrlSafe(final boolean urlSafe)|https://github.com/apache/commons-codec/blob/master/src/main/java/org/apache/commons/codec/binary/Base64.java#L97] method call, as we will basically have three different options. If you may share your thoughts on the best implementation for your library, I can provide you with a PR later when I have some spare time. KR, Aleksandr > Separate Base64 decoding to "real" base64 and base64Url > ------------------------------------------------------- > > Key: CODEC-333 > URL: https://issues.apache.org/jira/browse/CODEC-333 > Project: Commons Codec > Issue Type: Improvement > Affects Versions: 1.18.0 > Reporter: Aleksandr Beliakov > Priority: Minor > > The behavior applied in CODEC-75 does not distinguish between Base64 and > Base64Url implementations within > [isBase64|https://commons.apache.org/proper/commons-codec/apidocs/org/apache/commons/codec/binary/Base64.html#isBase64(java.lang.String)] > and > [decodeBase64|https://commons.apache.org/proper/commons-codec/apidocs/org/apache/commons/codec/binary/Base64.html#decodeBase64(java.lang.String)] > methods. Nor, it does change the behavior, even if instantiating a class > object using a Builder like below, explicitly stating that url-safe > processing should not be used: > {code:java} > Base64 base64 = Base64.builder().setUrlSafe(false).get(); {code} > with the method > [isInAlphabet|https://commons.apache.org/proper/commons-codec/apidocs/org/apache/commons/codec/binary/Base64.html#isInAlphabet(byte)] > still using the base64Url-safe decoding table. > > This causes some issues in the implementation, for instance inability to > distinguish between Base64 and PEM encoded certificates, with Apache Codec > accepting the latest as Base64. See a (chunked and dummy) example below (with > _Base64.isBase64()_ method returning true): > {code:java} > -----BEGIN CERTIFICATE----- > AQIDBAU= > -----END CERTIFICATE-----{code} > > According to the [RFC 4648|https://datatracker.ietf.org/doc/html/rfc4648], > handling of Base64 and Base64Url together should not be used: > {panel:title=5. Base 64 Encoding with URL and Filename Safe Alphabet} > This encoding may be referred to as "base64url". *This encoding should not > be regarded as the same as the "base64" encoding* and should not be referred > to as only "base64". Unless clarified otherwise, "base64" refers to the base > 64 in the previous section. > {panel} > While same handling may be acceptable in some particular cases, we think a > generic library like Apache Commons Codes shall make a clear separation > between the two encodings and unless implementors explicitly state the > support of url-safe characters, it should not be applied by default. > > This should also avoid potential errors within the implementations, when a > system shall treat base64 and base64url encodings as two separate ones (for > instance like in > [standards|https://www.etsi.org/deliver/etsi_ts/119100_119199/11918201/01.02.01_60/ts_11918201v010201p.pdf] > explicitly requiring one or another implementation, based on the usage). > > _FYI, Google Guava have two separate implementations > [BaseEncoding.base64()|https://guava.dev/releases/17.0/api/docs/com/google/common/io/BaseEncoding.html#base64()] > and > [BaseEncoding.base64Url()|https://guava.dev/releases/17.0/api/docs/com/google/common/io/BaseEncoding.html#base64Url()] > allowing implementors to explicitly choose one according to the project > needs._ -- This message was sent by Atlassian Jira (v8.20.10#820010)