The specification defined unitless and percentage values for length attributes as representing a multiple or percentage of the default value for the attribute. This was changed to make such values represent a multiple or percentage of a reference value, where the reference value is usually, but not always, the default value. See the (Errata for) 220.127.116.11.1 Dictionary-based attributes for the rationale for this change.
The descriptions of the
mo elements now specifies a reference value for these attributes
which is the unstretched glyph size.
With the change (Errata for)
18.104.22.168 Length Valued Attributes, this means that unitless or percentage
values for these attributes represent a multiple of the unstretched
glyph size. This recovers compatibility with MathML2
(and is, of course, more useful given that the defaults for
maxsize are 0 and infinity, respectively).
indenttarget attribute does not say what should happen in situations where the target id is not a sensible alignment point. An example of such a situation is when the
indenttarget is in one column of an
mtable and the id it references is in another column.
The description of
indenttarget should be amended to say:
Specifies the id of another element whose horizontal position determines the position of indented lines when
indentalign="id". Note that the identified element may be outside of the current
mathelement, allowing for inter-expression alignment, or may be within invisible content such as
mphantom; it must appear before being referenced, however. This may lead to an id being unavailable to a given renderer; in such cases, the
indentalignshould revert to "auto". It may also lead to an id being in a nonsensical position that does not allow for alignment; the
indentalignshould revert to "auto" in such cases.
The description has been changed to use
mathsize rather than
mathcolor as an example of an attribute only allowed on token elements or mstyle, as
mathcolor is allowed on additional elements in MathML3.
There was an erroneous reference to
malign in the description of setting long division. This has been changed to
maligngroup. (Such an example using
maligngroup appeared in the MathML2 specification, but has been dropped from this specification in favor of
The table lists
displaystyle as inherited which was an unintended change from MathML2. The default value of
displaystyle in mtable should be false.
The non normative XSD schema is derived from the Relax NG, but unfortunately an error in the translation caused
share to be omitted from the content model of
apply and related constructs. (Reported by Fraser Goffin on www-math.)
diff -r1.8 mathml3-strict-content.xsd 13a14,17 > <xs:element ref="m:cbytes"/> > <xs:element ref="m:cerror"/> > <xs:element ref="m:cs"/> > <xs:element ref="m:share"/>
Unfortunately the Unicode Character U+2ADC (FORKING) has been classified as not being in "Unicode Normalization Form C". Uniquely for all the math characters in Unicode, this character decomposes in NFC to a base U+2ADD (NONFORKING) composed with the combining negation slash U+0338. Best practice is to serve documents in NFC form and so the operator dictionary entry has been changed from U+2ADC to U+2ADD U+0338.
The list of attributes in appendix I unfortunately missed some attributes, it should be ammended as follows: