[Ur] Supporting 'style' attribute securely

[Edward Z. Yang]

> And I disagree that compiler should prevent things like phishing through
> absolute positioning etc. Compiler should be safe with respect to
> _programmer's_ errors; this also means that user data can't appear in unsafe
> place unintentionally, without explicit parsing of this data. But if
> programmer do explicit parsing, compiler can not check correctness --- for
> example, if I use plain text design of forum, for safe input I need check
> alignments (that is leading spaces), compiler can't help me in this question.

The researchers in the information flow community would disagree!

Joking aside, I agree that a lot of the cases that I'm bringing up
are specific to an adversarial context.  But I think, if anything,
it indicates that if we /decide/ to care, then a lot of thought
and care should be taken.

Of course, if we decide not to care, then there are any number
of not obviously wrong things to do.

Edward