URL – According to Mueller, it’s best to list only working URLs in xml sitemaps and only the correct version for canonical URLs. For canonical URLs, he suggests providing the “/” version and not “index.html” in his example. He goes on to point out the importance of using the same URL found in the site’s navigation and if necessary to use 301 redirects to that same URL when necessary. The navigation issue if important especially if something other than a crawler creates your sitemap. Either way, it’s worth testing to be sure your Sitemap URLs are identical to those in the user path (I’ve actually had near knock down drag outs over this issue). JohnMu suggest only including URLs to indexable content like (X)HTML pages and other documents. In addition he points out, it’s best to only include URLs webmastes want indexed.
Last modification date – In his post Mueller points out the difficulty Google can have with determining a “Last modification date” for dynamic sites due to their dynamic nature. He suggests either using the correct time or none at all. John suggests using a “Last modification date” but not “Change frequency” unless webmasters can establish a consistent frequency.
Change frequency – Like “Last modification date”, Mueller suggests not using a date/time if the actual one isn’t available.
Priority – Mueller suggests not including “Priority” meta data in xml sitemaps unless webmasters feel they can provide accurate data.
In summary, JohnMu suggests sitemap.org XML files that contain URLs for inclusion in Google’s index and only those found in the site’s navigation. He suggests “Date or change frequency” and “Priority” as optional meta data.
UPDATE: JohnMu has posted additional information over at Search Engine Roundtable in response to Barry’s post.