alt131 - 9:53 pm on Aug 25, 2012 (gmt 0)
Agreed - but that's why the questions about whether this is to be used in a controlled environment, the types of sites that might be accessed and competence level of end-users. Also, the selector could be made more specific to reduce the chances of conflict even more (hence using !important as a bare minimum to start).
Sure, though it's not semantic. ..Even though you may use an element that we don't see often, there's still the risk of running into one
But a named span is safer.
I think the detail of that depends on the core issues of the environment and types of sites. Dealing with text-heavy sites with code that isn't that clean or has loads of .js driven whiz-bang effects clogging bandwidth, then the requirement to parse the DOM and insert multiple iterations of an inline style (or even a classed span) around every letter could slow things to the point of ceasing to be useful as a tool to achieve the priority outcome which is to improve/practise synaesthesia.
That said, no doubt are more ideal/less ideal ways to deal style once the text has been identified. I think the issue that remains open is options for identifying the text.
Luce, thought of <tt>. It has one extra letter and I know you use it. I figured that if suggested Murphy's law would immediately apply and your sites would be the first place to be visited :)