Content changes that occur during pre-processing Pre-processing introduces changes to the DITA content that you might want to be aware of if you work with the temporary files. While some steps, such as conref resolution, change the DITA content in a way that is expected and mandated by the DITA specification, other steps result in temporary output that does not comply with the DITA standard. In addition, if you are using a restrictive specialization, the pre-processing steps might introduce content that is not permitted by the content model of the specialization. Robert D. Anderson Kristen James Eberlein DITA Open Toolkit
Changes that do not comply with the DITA standard

The DITA-OT makes certain changes to documents during the pre-processing operation, in order to simplify or optimize later stages in the processing chain. These changes result in temporary files that do not comply with the DITA standard.Robert, I don't think that the first two items are correctly phrased. I couldn't manage to generate either the image or mapclass attributes in my intermediate files ...

Image attributes
If the document contains images, the DITA-OT adds @dita-ot:image-width and @dita-ot:image-height attributes to the <image> elements. These attributes enable scaling when the images are finally rendered.
@mapclass attribute on generated links
When links are generated based on information in the DITA map, a @mapclass attribute is added to the <link> element that indicates the class of the <topicref> element used to create the map. This attribute can be used to sort links or filter how links are displayed in output.
Flagging
The flagging stage of the pre-processing operation creates pseudo-specializations of the <foreign> element which contain flagging information. Flagging some elements, such as <topic> and <row>, results in markup where the pseudo-specialization of the <foreign> element is the first and last child of the element.

Changes that might not comply with restrictive specializations

Some changes made during the pre-processing operation might not comply with restrictive specialization. Most of these changes are listed here, athough the list is not exhaustive. If you develop an output transformation for a restrictive specialization, you should be aware of these changes. If necessary, you can customize the pre-processing operation to avoid these changes.

Related links
The maplink step of the pre-processing operation generates contextual links based on the maps, for example, next and previous links; this step uses the <related-links> element to contain the contextual links. This can cause unexpected output if a specialized document type has removed the <related-links> element.
Cascade of map metadata
When metadata is specified in a map, it generally cascades to nested <topicref> elements and the referenced topics. If the map contains specialized metadata elements or attributes, the metadata might cascade to <topicref> elements or topic <prolog> elements that are not intended to have the metadata. If a topic document type has been specialized to disallow metadata, metadata specified in the map might nonetheless cascade into the topic <prolog> element.
Title text in links
By default, links or cross references without link text are updated to include link text. The link text is generally generated by using the title of the referenced topic. This can cause unexpected output if cross references or links have been specialized and the specialized content model does not allow text.
@xtrf and @xtrc elements
The pre-processing operation adds these two attributes to every element, in order to provide trace-back information in case of errors. These attributes are added even for specializations that forgot to define these attributes.

Changes or files that are not intended to comply with existing document types

Some processes in the toolkit are not designed to produce documents that comply fully with a DITA DTD or XSD.

Topic merge
The topicmerge process is used by the PDF, ODT, and RTF transformations to produce a single document that contains all of the content that is referenced by the master map. This file is not DITA and is specific to the DITA-OT. Its purpose is to simplify processing for output transformations that result in a single document.
Chunking
When the @chunk attribute is used in a map, DITA documents might be split or merged in a way that does not comply with any existing document type. For example, merging documents might result in a topic that nests concept, task, and specialized troubleshooting topics, even if no existing DTD or Schema permits that combination in a document.