Should internal DOI's follow the Mandatory and recommended encoding?

[michaelstoner]

When converting JATS to CrossRef a DOI such as 10.7717/peerj.2114/table-1]includes a character [ that's recommend to be encode.

So when converting DOI's should we obey the Mandatory and recommended encoding for DOI deposit and URLs? This is relevant to "declaring" new DOI's as apposed to referencing existing DOIs.

The encoding is listed here http://www.doi.org/doi_handbook/2_Numbering.html#2.5.2.4

<table-wrap id="table-1">
               <object-id pub-id-type="doi">10.7717/peerj.2114/table-1]</object-id><label>Table 1</label><caption>
                  <title>Pre-feedback descriptives of participants.</title>
               </caption>

I'm going to branch and commit a proposed solution note this will allow the % symbol with must be escaped, as it's the escaped character, and without updating the project to XSLT 2.0 and therefore allowing regex I think it's a little difficult to detect that all occurrences of '% are related to encoded characters.

[michaelstoner]

The consensus is that this is really much better suited to a schematron check rather than an XSLT rule. Similar issues will occur in many places, and even for this small issue the current solution is not a complete fix.