Jun 27

Marc Hedlund

Marc Hedlund

It turns out RSS and Atom really are different

I've argued that RSS and Atom are functionally equivalent for users these days. Choose one and make it easy for people to subscribe, I said.

It turns out that, for Radar readers at least, our two feed links attract different kinds of readers. Who knew? Looking at our feed traffic, about 35% of you subscribe to our RSS 1.0 feed, while the rest subscribe to our Atom feed. 52% of the RSS 1.0 readers use Bloglines, but 77% of the Atom readers use Bloglines. Our Atom readers are significantly more likely to click on a link in one of our posts. And the posts you like are different, too. For our RSS 1.0 readers, the most popular articles in June were:

Rank Title
1 The NY Times Gets into the Blog Spirit
2 Google Maps + Yahoo! Traffic Back Online
3 Review of Gershenfeld's Fab on Slashdot
4 Good Discussion of Apple/Intel Rumors
5 CafeSpot

And then for our Atom readers:

Rank Title
1 Ruby on Rails, and the Rails Beta Book
2 The Rise of Open Source Java
3 Google Maps + Yahoo! Traffic Back Online
4 Review of Gershenfeld's Fab on Slashdot
5 OS GIS Conference Day 1

So there are more Atom readers, who are more likely to use Bloglines, click on links, and read our nitty-gritty tech posts; while the RSS 1.0 readers go for the bigger brand name topics, and for some reason don't give Nat the love he deserves. (Y'all are missing out on that score.)

Can this be explained just by the relative permeation of the terms "RSS" and "Atom"? I wonder.

tags:   | comments: 15   | Sphere It

Previous  |  Next

0 TrackBacks

TrackBack URL for this entry:

Comments: 15

  Mark [06.27.05 11:57 AM]

I think these numbers not representative since OReilly has never supported the primary RSS fork (RSS 2.0) and since OReilly employees (claimed not to be working under OReilly orders) were a primary driver for the creation of RSS 1.0 and they were active in the war in syndication formats, one of the results of which is the creation of the Atom format.
I think OReilly as an organisation should grow up and bury its hatchet against Dave Winer and support RSS 2.0.

  Phil Ringnalda [06.27.05 12:05 PM]

I think "bury its hatchet against" is probably an unfortunate way to phrase that.

Me, I subscribe to the Atom feed because it's first in the autodiscovery, and I've never bothered to teach my aggregator to force me to choose when there's more than one.

  Sean [06.27.05 12:48 PM]

The reason I use the Atom feed is because it is the one on the top of the sidebar. My RSS reader can handle both, so I chose the top one. I wonder if your numbers would change if you switched their order in the sidebar.

Not only that, but when I point my RSS reader to this site, the atom feed is on the top because it alphabetically comes before the index.rdf file.

Just food for thought.

  Marc Hedlund [06.27.05 01:46 PM]

I agree that having Atom first on the page and first in auto-discovery probably explains why there are more Atom subscribers than RSS subscribers. But why are the behaviors of the two groups so different? I'll be interested to see if, for instance, the clickthrough rates remain higher on one feed than the other.

I certainly agree that our numbers are not representative of anything in particular, other than the set of Radar subscribers. But I still think they're interesting, maybe as a way of understanding the difference in public perception between the terms "RSS" and "Atom." I don't think the RSS version number matters at all to that discussion. (There is, by the way, no corporate hatchet to be buried -- we're not nearly organized enough to all hold the same hatchet. And I mean that in a good way!)

  Matt Haughey [06.27.05 04:33 PM]

I subscribed to some RSS version up until a few weeks ago when I noticed it was never updating in Bloglines. I went back to recheck and noticed the atom one worked so I moved to that. I probably subscribed on the day of launch, so perhaps it was an early version that was abandoned.

  Ben [06.27.05 07:22 PM]

When you use bloglines and enter the url as the blog the atom feed is the first one displayed of SEVEN (why so many??) different feeds for the blog. As with the other commenters...its the first option that people will click on.

  Tomas Jogin [06.28.05 03:00 AM]

I always subscribe to Atom feeds if such are available, simply because I've experienced much less problems with Atom feeds than with RSS ones.

  Matthew Gifford [06.28.05 08:32 AM]

I doesn't surprise me that the different feed formats have different audiences. However, it would have been nice to know how the RSS 2 audience difffers from those of RSS 1 and Atom. Too bad the article was limited by childish politics.

  Matthew Gifford [06.28.05 08:33 AM]

I -> It. Sorry about the typo.

  Marc Hedlund [06.28.05 10:32 AM]

Matthew, sorry to hear that the article seems tainted by political bias. Just to restate, my contention is that the differences between RSS 0.9, 1.0, 2.0, Atom, and so on are completely irrelevant to RSS users. (They only matter to the people writing software to produce and consume these formats.) It would be my preference that NONE of these distinctions be apparent in the browser window.

As a user, I did not need to be aware whether my browser was speaking to an HTTP/1.0 server or an HTTP/1.1 server. From my point of view, it just worked. Of course, HTTP/1.1 offered many features not supported in HTTP/1.0, and made implementation more consistent for server and browser vendors. That doesn't mean HTTP/1.1 is irrelevant, nor that I am opposed to HTTP/1.0. It just means that the transition was handled successfully.

All of the proposed RSS/Atom upgrades have failed this test. How terrible would it be if I could show that HTTP/1.0 users and HTTP/1.1 users read the Web differently? I can't, but in this post I was able to show differences in end-user behavior between RSS formats. That's terrible! I am willfully not caring about the better or worse aspects of any of the RSS specs (I know other people at O'Reilly do care about these details, but as I say, one of the good things about O'Reilly as a company is that we're all allowed to speak with our own voice, not the corporate monothink). So from my point of view, that there are particular versions of RSS or Atom offered on this site or any other shows a general failure of the political process, not bias in any one direction or another. We all should have only one format, and the differences I found between subscribers should not exist.

  Matthew Gifford [06.28.05 11:05 AM]

The first time I read the post, I thought you were implying something like "9 out of 10 geeks prefer Atom." After reading it again, I see that your real point isn't the differences between the audiences.

I agree that it's bad that users are aware of the different syndication formats. It doesn't seem that O'Reilly Media does not agree, however. If they did, they would be providing only one format.

  Anthony R. Thompson [06.28.05 07:18 PM]

I've been reading about syndication for some time, and I think I'm finally ready to use it on one of my sites.

However, what I thought was an easy decision (use RSS 2.0 because 2.0 is obviously the latest which fixes all the mistakes of previous versions) turns out to be anything but.

After searching on the web and reading the RSS Wikipedia page, and then seeing the ugly politics in action here, I'm more confused (and pessimistic) than ever about syndication.

I'd really like to see an (as unbiased as possible) O'Reillynet article giving the relative merits of each of the RSS and Atom versions.

In particular, I'd like to know whether one really has to provide two or more syndication files or if one should really do the trick for most readers (RSS 1.0, Atom, or whatever).

  Matthew Gifford [06.29.05 06:27 AM]

Anthony, I'd recommend choosing either RSS 2 or Atom. There's no need to provide multiple formats.

  Seth Russell [07.24.05 09:17 AM]

I noticed that there is a trend brewing for publishers to selece one format, see ,another example is my own site. With MS supporting RRSS 2.0, let us hope that one format will start to dominate. Having multiple feed formats there to confuse the browsing public is not good for spreading syndication. I seriously doubt that anyone can give me, *as a consumer* a single reason that one feed is any better than any other feed and express that in one short paragraph that makes any sense to me.

  Danny [07.24.05 02:24 PM]

My 2.02 cents:
The difference in the subscribers and their behaviour I suspect be explained by a combination of what get's presented as the first choice (either as the icon or via autodiscovery) and brand recognition, i.e. RSS 1.0 is RSS.

At this point in time, for a regular content feed (A) from a regular blog (B) read in a regular reader (C), I don't believe there is any significant difference between what the various formats offer. For this kind of application the same kind of functionality could probably be offered equally well using Netscape's RSS 0.91 from 1999.

I think most developers also agree that at this point in time for the A+B+C scenario there's no need for a publisher to provide more than one feed. So take your pick.

However, move only inches away from the A+B+C scenario and differences between the formats do become apparent. Podcasting was initially promoted as an RSS 2.0 feature, and the majority of tools that support this kind of media delivery only do it through RSS 2.0. There are tools which now support media delivery in Atom, and there's now a good spec (mod_enclosure) for the same functionality in RSS 1.0. Core support for media delivery using RSS 2.0 is limited, and the fact that it can be extended via namespaces if anything has only led to new issues.

RSS 1.0, apart from having fairly widespread use as a general syndication format is currently favoured in several specific scenarios where RSS 2.0 falls short. A good example is in science publishing, where bibliographic and other metadata needs to be delivered in an unambiguous fashion - check the Nature material on the subject. FOAF integration is another example. The RDF base of RSS 1.0 offers opportunities that go way beyond a single-domain XML applications like blog publishing. But the development of RSS tools has to date mainly focused on providing nice UIs for the syndicated approach to publishing. Most of the server-side tools apart from aggregators have been traditional search engines ported to the different formats.

Only very recently have we seen applications that diverge from this basic model getting widespread adoption. Only now is the extension capability of syndication systems starting to get exploited, with Podcasting, the Yahoo! media material, iTunes and Microsoft's lists. Personally I believe the difficulties associated with supporting extensions to the format without any extension model beyond XML namespaces will lead developers to notice diminishing returns as syndication applications move into realms other than improved tabbed browsing, the A+B+C scenario.

There has been significant growth in adoption of RDF outside of syndication. It seems probable to me that we'll soon be seeing a lot more RDF in the syndication domain, at least under the hood, thanks to its capability for integrating heterogenous data from multiple sources. A lot of work is currently going into building individual vertical solutions for bloglike content management, media distribution, social networking, event notification etc etc. Integration across systems is happening, but much of this is happening through very narrow channels, with one-off conversion for the NxN communications. But I think it's pretty clear that such systems would benefit from some kind of shared model, exactly what RDF can offer. In other words, this is exactly the kind of capability Web 2.0 needs.

Getting back to the point, there's some irony that it looks like RDF will be coming into syndication from outside, 5 years or so after RSS 0.9 originally supported it.

I believe where Atom has a technical advantage is as an over-the-wire solution primarily covering the RSS 2.0 A+B+C space. It's an updated, bugfixed RSS. There are fixes at lots of individual points: use of decent identifiers, proper XML content handling, relatively strong notions of validity etc. This is all wrapped into a clear spec. The overall effect should be that the well-known Web technologies can be exploited far more, maximising the capability of Web 1.0. It does have an extension mechanism that promises to improve the chances of compatibility between extensions, hopefully that will help in the iTunes vs. Media RSS kind of situation. When you bring in the Atom Publishing Protocol, there's a huge leap towards Web 2.0, moving beyond outdated modes of interoperation like RPC.

There's still buzz growing around syndication, and for many people that means RSS 2.0. But looking ahead, compared to the model/integration potential of RSS/RDF and the over-the-wire precision of Atom, RSS 2.0 looks set for gradual decline. Hopefully the historical political and unresolvable technical issues surrounding RSS 2.0 will appear as a passing phase, and it won't act as an albatross around syndication's neck.

Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.