unicode results needed for atom feeds

[Edwin Grubbs]

I received the following traceback when retrieving an atom feed in zope 3:

Traceback (most recent call last):
  File "/home/egrubbs/canonical/lp-branches/trunk/lib/zope/publisher/publish.py", line 141, in publish
    response.setResult(result)
  File "/home/egrubbs/canonical/lp-branches/trunk/lib/zope/publisher/http.py", line 816, in setResult
    body, headers = self._implicitResult(result)
  File "/home/egrubbs/canonical/lp-branches/trunk/lib/zope/publisher/browser.py", line 680, in _implicitResult
    body, headers = super(BrowserResponse, self)._implicitResult(body)
  File "/home/egrubbs/canonical/lp-branches/trunk/lib/zope/publisher/http.py", line 856, in _implicitResult
    raise ValueError(
ValueError: Unicode results must have a text content type.

Although most feed readers will accept a content type of "text/xml" which will get around this problem, atom feeds are supposed to have the content type of "application/atom+xml" according to FeedValidator.
http://feedvalidator.org/docs/warning/UnexpectedContentType.html

I think that zope.publisher.http.HTTPResponse._implicitResult(self, body) should be more permissive of the content-type header when the body is a unicode object.

[James Henstridge (jamesh)]

We ran into this problem ages back for the RDF feeds (which have a content type of application/rdf+xml). There the workaround was to make the viewclass explicitly encode the body as UTF-8.

The same should be done for the feeds.