Whitespace and the annotation module

[John McCrae <john.mccrae@insight-centre.org>]

Hi all,

One more comment for the 2nd public review. I have been thinking and I 
am still not sure about our rule for whitespace in the elements. In 
particular, in the converter I have been developing I am having problems 
because we cannot apply pretty printing (indenting) to an XML file 
without changing the content in the model. Further, I think the rules 
are unintuitive and many will add whitespace and create unfortunate 
errors. Instead I propose that we adopt the HTML methodology as 
described here:

https://infra.spec.whatwg.org/#strip-newlines

In this case, before processing the content of any text carrying 
element, we will first remove all new lines ('\n', '\r'), delete all 
trailing and leading whitespace and replace all remaining blocks of 
ASCII whitespace with a single space.

I would also make a model change, replacing all references to 'non-empty 
string' with a 'normalised string'. This means a string that contains no 
new lines, does not start or end with a whitespace, contains no block of 
ASCII whitespace more than a single space and is non-empty. This ensures 
that other serializations (JSON, RDF) cannot generate content that 
cannot be represented in XML.

I do worry that this does not really cover Chinese, Japanese (and maybe 
Thai/Lao), as the whitespace rules for HTML are more complex in Unicode, 
but I think that this can probably be worked around by lexicographers 
working in these languages. We can add a note to the spec for these 
languages.

Regards,

John

--
John P. McCrae
(he/him; #startsWithAName John (rhymes with "gone") McCrae (rhymes with "hay") /dÊÉn mÃkÉeÉ/)
Assistant Professor - SFI Insight Centre for Data Analytics, Data Science Institute & Computer Science, University of Galway