Questions about the metadata (get-meta.xsl) stylesheet in the HTML5 transformation

[chrispy-snps]

I have a few questions about the get-meta.xsl stylesheet in the HTML5 transformation:

1. In most calls to evaluate metadata, the match is against the current element and a self::dita/*[1] element:

   <xsl:apply-templates select="*[contains(@class,' topic/title ')] |
                self::dita/*[1]/*[contains(@class,' topic/title ')]" mode="gen-metadata"/>
   

   What is the self::dita/*[1] variant for, and where does it get introduced in the processing?

2. What type of data is the mode="gen-type-metadata" mode intended for? It's an evaluation of the current element, stubbed out to return nothing:

   <xsl:apply-templates select="." mode="gen-type-metadata"/>
   ...
   <xsl:template match="*" mode="gen-type-metadata">
   </xsl:template>
   

3. What type of data is the mode="gen-format-metadata" mode intended for? It is similar to the gen-type-metadata mode:

   <xsl:apply-templates select="." mode="gen-format-metadata"/>
   ...
   <xsl:template match="@* | node()" mode="gen-metadata gen-format-metadata" priority="0"/>
   

4. Is there a bug in <abstract> processing? <shortdesc> is processed with mode="gen-metadata", which generates no output because the default action is to ignore it. <abstract> is processed with mode="gen-shortdesc-metadata". That moded template looks for <shortdesc> elements inside the <abstract> element, which is impossible per the DITA schema. I think there might be something wrong here. If someone can confirm, I'll file an issue for it.