The checked
and selected
attributes are allowed only two values, which are a copy of the attribute name and (from HTML 5 onwards) an empty string. Giving any other value is an error.
If you don't want to set the attribute, then the entire attribute must be omitted.
Note that in HTML 4 you may omit everything except the value. HTML 5 changed this to omit everything except the name (which makes no practical difference).
Thus, the complete (aside from variations in cAsE) set of valid representations of the attribute are:
<input ... checked="checked"> <!-- All versions of HTML / XHTML -->
<input ... checked > <!-- Only HTML 4.01 and earlier -->
<input ... checked > <!-- Only HTML 5 and later -->
<input ... checked="" > <!-- Only HTML 5 and later -->
Documents served as text/html (HTML or XHTML) will be fed through a tag soup parser, and the presence of a checked attribute (with any value) will be treated as "This element should be checked". Thus, while invalid, checked="true"
, checked="yes"
, and checked="false"
will all trigger the checked state.
I've not had any inclination to find out what error recovery mechanisms are in place for XML parsing mode should a different value be given to the attribute, but I would expect that the legacy of HTML and/or simple error recovery would treat it in the same way: If the attribute is there then the element is checked.
(And all the above applies equally to selected
as it does to checked
.)