[ https://issues.apache.org/jira/browse/JCRVLT-391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16977205#comment-16977205 ]
Konrad Windszus commented on JCRVLT-391: ---------------------------------------- One concrete issue is JCRVLT-393. The other is the concern about maintaining third party code. Regarding reproducible ouutput: * AttributeImpl is not using a HashSet but rather a list with a predictable order, so any serialization based on SAX keeps that order * Indentation of attributes is not necessary for reproducability but makes the diffs nicer, unfortunately there is not OOTB support for that in TraX (https://stackoverflow.com/questions/8393370/use-xslt-to-add-newlines-after-attributes) > Remove copied classes of Xerces > ------------------------------- > > Key: JCRVLT-391 > URL: https://issues.apache.org/jira/browse/JCRVLT-391 > Project: Jackrabbit FileVault > Issue Type: Task > Components: vlt > Affects Versions: 3.4.0 > Reporter: Konrad Windszus > Priority: Major > Fix For: 3.4.2 > > > Currently the classes below {{o.a.j.v.util.xml}} seem to be a copy of Xerces > classes. Instead of just relying on an old copy one should use a proper Maven > dependency together with https://maven.apache.org/plugins/maven-shade-plugin/ > or https://bnd.bndtools.org/instructions/conditionalpackage.html to make sure > update to the latest version is possible. > Also it needs to be clarified if the embedded Xerces should be listed > explicitly in > https://github.com/apache/jackrabbit-filevault/blob/trunk/LICENSE.txt. > The original intent is stated in > https://github.com/apache/jackrabbit-filevault/blob/9fa6c72bf4bdf36331b50b9370f3ed826a4622e8/vault-core/src/main/java/org/apache/jackrabbit/vault/util/xml/serialize/XMLSerializer.java#L64. > The only modifications to the original Xerces were linebreaks after > attributes and an alphabetic attribute sort order. > The alternative is to use the {{javax.xml.transform}} API (JAXP > Transformation API). There is no official way though to control the order of > attribute, therefore I would recommend to already emit the attributes in the > correct order instead of reordering them with the output. Controlling the > indentation of attributes is hard to achieve: > https://stackoverflow.com/questions/8393370/use-xslt-to-add-newlines-after-attributes -- This message was sent by Atlassian Jira (v8.3.4#803005)