Re: Feedback on XML content sink plans

Henri Sivonen wrote:
> I wrote about my plans on incrementalizing the XML content sink (bug 
> 18333):
> http://hsivonen.iki.fi/kesakoodi/content-sink/
> 
> I would appreciate a sanity check. Does it look right?

I'm not completely convinced by the "Tearing down the document in mid-layout and 
replacing it with the Yellow Screen of Death would make the user experience 
unstable" part...  Some discussion of that (sicking?  peterv?  jst?) would be nice.

What happens if the error is inside a <script> tag?  Do we "partially render" 
the script?   I feel we should not.

What happens with XSLT exactly?  That is, how is a well-formedness error in a 
document that has an XSLT stylesheet handled?

 > Don’t be incremental with innerHTML.
 > Don’t be incremental with XHR.

Not an issue; that stuff isn't being rendered anyway.  So either behavior is ok 
here; just do whatever is easiest.

 > Don’t be incremental with DOM LS.

I don't believe we implement that, do we?

 > Don’t be incremental with XUL—at least not when loading the local UI.

XUL doesn't use the XML content sink, so it's all ok.

 > Don’t be incremental with XBL when loading a binding.

Again, either behavior is fine here; do whatever is easiest.

 > Don’t be incremental with RDF when loading a graph.

RDF doesn't use the XML content sink.

 > I’d rather not factor anything out of SinkContext.

That's all the code that keeps track of what's been flushed, etc.  Is the plan 
to just fork it all, then?

-Boris