snellspace.com

If Web Browsers are going to automatically detect feeds and place helpful subscription icons in the chrome of the browser enabling single-click subscriptions, those browsers should provide a means for allowing users to launch third-party feed readers when the button is clicked. I use Liferea as my feed reader and Firefox for Web browsing. When I see the orange icon in the address bar, I want to click on it and have it launch Liferea.

Update: Sam sent me a link to “Feed Your Reader”. I don’t see Liferea listed as being supported but I’ll see if I can get it working later. Thanks Sam!

Throughout the Atom spec there are numerous indications that the spec assigns no significance to the order of elements appearing within feed and entry elements. Unfortunately, I’ve found that this doesn’t hold true for a number of implementations. Take this test for instance: http://www.snellspace.com/public/sourcetest.xml and pull it up in Thunderbird 1.5. The only relevant difference between the two entries in the feed is the position of the source element. In the first entry, the source element comes before the entries own title, id, updated and alternate link elements. When parsed, Thunderbird inadvertently picks up the source element’s title, id, updated, and alternate links for the entry. This bug has been patched for Thunderbird 2.0 but won’t likely be fixed in the 1.5.x stream. Luckily, it’s easy to work around. Simply move the source element so that it appears after the entries own main metadata elements. The second entry in the test feed shows the workaround.

Unfortunately, Thunderbird is not the only feed reader with order issues. I’ll be working up a conformance test later today that tests how feed readers handle the ordering of elements within an entry.

After a day of testing Atom 1.0 support in Feed Readers, I’ve learned a few things.

• iTunes 6.0 supports Atom 1.0 based Podcasts! Without using any of their iTunes extensions! Excellent! Unfortunately, iTunes 6.0 doesn’t understand xml:base. Bogus.
• There’s a significant bug in Mozilla Thunderbird’s Atom support affecting feeds containing atom:source elements.
• Lots of FeedReaders seem to be ignoring link rel attribute values… selecting the last link in the entry as the alternate regardless of what the rel attribute says. Hint: that’s a bug
• RSSBandit has a problem properly displaying RFC3339 dates and doesn’t support xml:base at all.
• RSSOwl doesn’t support xml:base but is otherwise pretty nice
• Snarfer is really promising! Very simple and slick but it definitely needs to mature.
• Sage has really come a long way. It supports xml:base. If you use Firefox Live Bookmarks, consider Sage as a good replacement
• Lots of FeedReaders are hiding lots of information from their users. Very few, for instance, expose atom:category values. None expose the atom:contributor element. Most only expose “alternate” and “enclosure” links, leaving “related”, “via”, “self”, etc out in the cold.
• On Linux, there doesn’t seem to be a better Feed Reader than Liferea. Hands down beats everything else I’ve seen… except that it’s very slow

A few notes to Feed Reader developers:

• Focus on spec conformance first, features second. Many Feed Readers seemed to sacrifice on basic spec compliance but offered all kinds of bells and whistles to users.
• Pay attention to link element rel values! If it doesn’t say rel=”alternate”, don’t use it as the alternate link (aka permalink)
• Expose Categorization information. It’s extremely useful information that you’re hiding from your users! (hint: “categorization information” == both the term and scheme attributes on the atom:category element)
• Expose Links. Atom is not RSS. There is a lot more to look at than the permalink and enclosure. Related links, Via links, etc are all very useful.
• Support xml:base! It’s not that hard to do.
• Support xml:lang! It’s not that hard to do.

Add Links to the growing list of Atom Conformance Tests.

It’s amazing how many applications out there have a really hard time following even the most basic requirements of a standards spec. Silly really.