What's the difference between text/xml vs application/xml for webservice response

From the RFC (3023), under section 3, XML Media Types:

If an XML document -- that is, the unprocessed, source XML document -- is readable by casual users, text/xml is preferable to application/xml. MIME user agents (and web user agents) that do not have explicit support for text/xml will treat it as text/plain, for example, by displaying the XML MIME entity as plain text. Application/xml is preferable when the XML MIME entity is unreadable by casual users.

(emphasis mine)

This is an old question, but one that is frequently visited and clear recommendations are now available from RFC 7303 which obsoletes RFC3023. In a nutshell (section 9.2):

The registration information for text/xml is in all respects the same
as that given for application/xml above (Section 9.1), except that
the "Type name" is "text".

According to this article application/xml is preferred.

---

EDIT

I did a little follow-up on the article.

The author claims that the encoding declared in XML processing instructions, like:

<?xml version="1.0" encoding="UTF-8"?>

can be ignored when text/xml media type is used.

They support the thesis with the definition of text/* MIME type family specification in RFC 2046, specifically the following fragment:

4.1.2.  Charset Parameter

   A critical parameter that may be specified in the Content-Type field
   for "text/plain" data is the character set.  This is specified with a
   "charset" parameter, as in:

     Content-type: text/plain; charset=iso-8859-1

   Unlike some other parameter values, the values of the charset
   parameter are NOT case sensitive.  The default character set, which
   must be assumed in the absence of a charset parameter, is US-ASCII.

   The specification for any future subtypes of "text" must specify
   whether or not they will also utilize a "charset" parameter, and may
   possibly restrict its values as well.  For other subtypes of "text"
   than "text/plain", the semantics of the "charset" parameter should be
   defined to be identical to those specified here for "text/plain",
   i.e., the body consists entirely of characters in the given charset.
   In particular, definers of future "text" subtypes should pay close
   attention to the implications of multioctet character sets for their
   subtype definitions.

According to them, such difficulties can be avoided when using application/xml MIME type. Whether it's true or not, I wouldn't go as far as to avoid text/xml. IMHO, it's best just to follow the semantics of human-readability(non-readability) and always remember to specify the charset.