- Atom and RSS are BOTH the wrong way to do syndication.<DIV>The debate between <A href="http://www.intertwingly.net/wiki/pie/FrontPage">Atom</A> and <A href="http://blogs.law.harvard.edu/tech/rss">RSS</A> is missing the point. Neither is a good way to do syndication.</DIV>
<DIV> </DIV>
<DIV>This seems obvious to me, but apparently either I'm missing the point or the rest of the world is - now that the <A href="http://www.w3.org/">W3C</A> is <A href="http://www.imc.org/atom-syntax/mail-archive/msg03711.html">taking an interest in Atom</A> I think it's worth saying this stuff again.</DIV>
<DIV> </DIV>
<DIV>Here's the problem, in a nutshell: They don't scale. </DIV>
<DIV> </DIV>
<DIV>Imagine if Windows came with a feed reader built-in and easy to configure. Pick a popular website - say CNN. Imagine 50 million users subscribed to the CNN feed. Imagine each of those polling every hour minutes for updates.</DIV>
<DIV> </DIV>
<DIV>That's 1.2 billion hits per day on the feed, all on the server that contains the URL that's advertised (or server farm, etc).</DIV>
<DIV> </DIV>
<DIV>Both Atom and RSS punish the publisher by increasing their bandwidth consumption as they grow more popular - the same way a website can get “slashdotted”, a popular site can start to suffer if too many people subscribe. Some would say it's a good problem to have, but it's a problem we should be working to avoid.</DIV>
<DIV> </DIV>
<DIV>It's one thing for some guys to say “here is a cool way to syndicate web sites”, but when organizations like the IETF and W3C get involved, I expect something a little more solid.</DIV>
<DIV> </DIV>
<DIV>Look at email. My mail client polls my mail server for new mail, but I'm not polling the same mail server you are. When you check for mail, the impact on the Internet is nil, because the request goes to a local server managed by your ISP. When I send you an email message, my server sends it to your server and forgets about it - you're not polling ME to see if I've sent you an email. You're asking your own local mail server.</DIV>
<DIV> </DIV>
<DIV>With Atom and RSS it's as if your computer were constantly polling all your friends asking them if they've got an email for you. If mail were implemented that way, it would have collapsed long ago.</DIV>
<DIV> </DIV>
<DIV>I know a random blog posting isn't the best venue for this; once there's an official forum for working through a web syndication protocol I'll get involved if I can. I'm afraid that the camps are so loyal to their positions that it's going to be hard to get anything else taken seriously, but at least if a third party is involved in the standardization, perhaps they can moderate a better solution.</DIV>
<DIV>
<DIV> </DIV>
<DIV>So here's a plea to the IETF, W3C, Atom guys and RSS guys - Forget the war between Atom and RSS, and implement something new that will make better use of network resources. Don't stick with what we have unless you're really convinced that there isn't a technically superior solution. Take a look at the mechanisms used by NNTP, SMTP and IRC. These are systems that we know can scale. Thanks.</DIV>
<DIV> </DIV></DIV>Thu, 01 Sep 2005 19:17:07 GMT