In his this morning's rant on RSS, Dave mentions polling frequency for aggregators and the usual pull vs. push issue. My opinion as an application developer is that the beauty of RSS is that it's not instant. Current aggregator users simply don't expect a post to appear in their aggregators instantly. Having to wait on average 30 minutes (considering the traditional hourly scan) is a perfectly acceptable delay; it might be different on intranets but usually intranets don't have bandwidth limitations.

Developing an application which does not require instant feedback allows to scale it in a much nicer way: backend agents can be set up to routinely perform tasks, update pages, find relations, ping profiles and do all kind of activities without any hurry: we always have an hour to go.

For our own k-collector aggregator we took a slightly different approach to the hourly scan: all feeds are polled within one hour, but not all of them are scanned at the same time. The aggregation activity is dynamically scattered in the whole hour, granting prompt updates but preventing the usual processor and bandwidth spikes triggered by hourly scans (active users with a Radio or a MovableType client ping the server when they update, meaning that their posts get on the site quicker).

I do understand why somebody would want to make RSS as quick as email or instant messaging but it would need radical rethinking of the current approach and infrastructure, for example downloading RSS using something like bit torrent, spreading the overall bandwidth usage on clients and not only on servers.

PS: I do like the name RSS... but since I'm a nerdy user who has been using this name for years my opinion doesn't count much. Compared to HTTP, "html" or "tcp/ip", RSS seems a reasonably easy name for everybody to use and remember. Marketing wise I would probably change the orange XML icon from "XML" to \"RSS" (I understand why it reads XML, but it's probably a little hard to explain to the everyday user).

Robert Scoble wrote two interesting posts last week about blogs comments and group weblogs. I think that we are ready to move to the next step with comments and group weblogs.

Comments are a very useful feature on weblogs but I think that they should be decentralized: in other words I would like comments I write on other weblogs to be saved on my own weblog. The main reason for this is that I would like to know what my favourite webloggers write on all weblogs not only on their own, this would give us a far better understanding of what's going on.

Trackback is a first step in this direction but it has problems, the main one is that I can't read a thread of comments in a trackback window because only the first few characters of the posts are reported there (trackback seems to come from the same kind of thinking that decided that RSS feeds should not contain full contents, which is something I hate but which is an entirely different rant).

What I would like to see is a comment window which looks exactly like current ones (i.e. you can read the whole thread without having to click on any link), but where the content is actually syndicated from the weblogs of each comment's author.

In a lot of cases I might not want a comment I make on some other guy's weblogs to appear just as a regular post here, I would probably prefer to have them stored in a separate category (easy to do with Radio), but still they would have to live here, not scattered on hundreds of other sites.

Something similar applies to group weblogs: each author should publish in their own space, the group weblog should only be a front end aggregating related contents from several different sources.

I believe that the work that is underway on RSS categories and topics will be an important step in this direction, there will be soon an entirely new class of applications based on RSS and new kind of aggregators.

I think that we are ready to declare 2004 "Year of RSS".