Re: asymmetry between serialzier and parser inwriting XMLDecl

Elliotte Rusty Harold wrote:

> >in XmlSerializer writing XMLDecl (<?xml version='1.0' ...?>) is supported
> >by startDocument() method however XmlPullParser will report XMLDecl as PI.
> >and each rountrip will add second <?xml as it thinks it is just another PI
> >(unless there is special code to detect that PI is really XMLDecl ...)
> >
> >if we use isWhitespace() (or isEmptyElementTag()) to signalize that PI
> >is not a real PI and should be ignored but instead interpreted as
> >XML declaration ...
> >
>
> I don't see how this would fix it. The XML declaration is not a
> processing instruction. XmlPullParser should not report it as one.
> Overloading isWhiteSpace() or isEmptyElementTag() with a completely
> different meaning that has nothing to do with white space or
> empty-element tags just confuses the issue  further, and is totally
> non-obvious.
>
> Either XmlPullParser should not report the XML declaration at all, or
> it should have a different token type to represent this construct.

hi,

i agree with you. however we want to keep backward compatibility for time
being and then have 2.x version that will fix all problems that were identified.

so i would think about it as a gradual improvement - the other possibility
introducing new event will break compatibility and that is what we really
want to avoid. anyway this affects only nextToken() method and only
for PROCESSING_INSTRUCTION event and i think in typical
80/20 cases next() should be used.

thanks,

alek