> First off, nobody has ever said you should "invent some random meaning" for anything.
Well, yes, you have. The definition of an error in this context is that there is no meaning for the given syntax specified in the standard. If you interpret the document anyway, you are assigning it some meaning. As that meaning does not come from the standard, you have simply invented it.
> The existence of an error in one section of a document shouldn't have any impact on the rest of the document, and so there's no reason why the error should prevent rendering the rest of the document.
1. You have it all backwards. Whether something is a "section" is part of the semantics that are specified by the standard. When your document does not comply to the standard, the standard does not assign any semantics to it, and as such it also does not specify what a section is in your non-complying document. The moment you talk about "sections" in your non-HTML document, say, you have already started inventing semantics.
Also, even if we assume that we somehow can magically identify "sections" anyway: How do you justify the assumption that sections do not have any impact on one another? For one, if you take HTML, in particular when CSS is involved, you absolutely easily can have very global effects using rather local constructs, can't you? But also: What if the human-language content has cross-references between sections? How do you ensure that the meaning of the human-readable content is conveyed correctly when somehow processing a "broken section"? What if the result is that you leave out an important warning regarding the following instructions in an attempt to help the user?
Really, how is that any different from a compiler replacing functions with errors with a No-Op?
2. What is the actual advantage? You say it is obvious, but it's not to me, which is why I asked. What is the advantage of just displaying something, even if the standard does not assign it any meaning, over aborting with an error?
Well, yes, you have. The definition of an error in this context is that there is no meaning for the given syntax specified in the standard. If you interpret the document anyway, you are assigning it some meaning. As that meaning does not come from the standard, you have simply invented it.
> The existence of an error in one section of a document shouldn't have any impact on the rest of the document, and so there's no reason why the error should prevent rendering the rest of the document.
1. You have it all backwards. Whether something is a "section" is part of the semantics that are specified by the standard. When your document does not comply to the standard, the standard does not assign any semantics to it, and as such it also does not specify what a section is in your non-complying document. The moment you talk about "sections" in your non-HTML document, say, you have already started inventing semantics.
Also, even if we assume that we somehow can magically identify "sections" anyway: How do you justify the assumption that sections do not have any impact on one another? For one, if you take HTML, in particular when CSS is involved, you absolutely easily can have very global effects using rather local constructs, can't you? But also: What if the human-language content has cross-references between sections? How do you ensure that the meaning of the human-readable content is conveyed correctly when somehow processing a "broken section"? What if the result is that you leave out an important warning regarding the following instructions in an attempt to help the user?
Really, how is that any different from a compiler replacing functions with errors with a No-Op?
2. What is the actual advantage? You say it is obvious, but it's not to me, which is why I asked. What is the advantage of just displaying something, even if the standard does not assign it any meaning, over aborting with an error?