It looks like the datasets/white space issue needs to be properly investigated for consistency.

Previously (pre 3.0), white space was trimmed by the OLS. So leading and trailing whitespace was ignored. An XML node like this one:

<mynode>
    abc
    def
     ghi
</mynode>

Would display:

abc def ghi

After offering SOLO deployment this broke for SOLO apps, because the Flash XML parser was handling SOLO datasets, and it did not trim whitespace by default. So we added a trimwhitespace tag:

<dataset trimwhitespace="true" />

... now displays:

abc def ghi

(Note preserved whitespace in between the letters).

I think the best fix here would be to come up with a comprehensive test case for XML parsing, so that we can either make it consistent across runtimes/deployment modes, OR know what the differences are, so that we can document them (and possibly add options to datasets, such as ignoreinternalwhitespace). Presently the docs for this are very sparse:

http://www.openlaszlo.org/lps-latest/docs/guide/proxied.html#d0e11320

The test case should include all the data-binding examples that you think might break - line breaks, HTML tags, leading white space, trailing white space, non-ASCII characters etc.