2.1.1 Conditional Source Text Processing
When converting source text into input elements, Internet Explorer ECMAScript first does the processing necessary to remove or replace any conditional text spans and then does the input element conversion using the results of that processing as the actual source text input to the identification of lexical input elements.
Each Program (see [ECMA-262/5] section 14), whether presented as either a discrete source text or as the argument to the eval built-in function, and each FunctionBody (see [ECMA-262/5] section 13) processed by the standard built-in Function constructor ([ECMA-262/5] section 18.104.22.168) has conditional source text processing performed independently upon it.
This specification defines conditional source text processing as if it were performed over an entire source text prior to any input element identification. It is an unobservable implementation detail whether this processing is actually performed in that manner or whether it is performed incrementally interweaved with input element identification. Additionally, conditional source text processing is only supported in IE10 mode and below.