The most obvious observation is that a lot depends on what kind of usage you do of your news reader and to what kind of sources you subscribe to.

In my case, for example, I live 12 hours a day in front of a computer. I check my news page almost once per hour. For users like me, the "all news in a page" / weblog approach is not bad, even subscribing to 160 feeds, since the number of new posts to read tends to be bearable and especially because we have developed the skill of scrolling trough pages very quickly and still spot relevant information.

This approach doesn't work very well if you are reading news only a few times per day: the amount of new stuff is too much and since you are picking up several hours of updates, you'd probably prefer to see them aggregated by channel rather than in sparse reverse chronological order, with all sources mixed.

At this point we are getting to how to organize feeds.

One popular approach is: let users group feeds how they prefer. MyRadio Tool and Kit allow this for Radio. NetNewsWire has a very nice "Combined View" feature that allows to have all posts (and not only titles) of a group visualized in a window. Very nice and fast.

Three panes display is also quite popular. It's how most email clients work. I think that it's good if you need to look back searching for some old post, it might also be good if you are not using very frequently. Personally I don't like the 3 panes approach because I'm lazy and I end up having to click too many times.

What other ways can we find to organize this content, both at authoring and reading levels?

Three approaches come to my mind:

• Channels/Categories
• Topics/Keywords
• Content-based filtering

A first approach is the use of channels when publishing. It's basically what most news sources already do: dividing contents by category (btw: this is what I think that the category element is for in the RSS 2.0 specs, but I might be wrong).

In other words, when publishing on a weblog or any other kind of site, authors could define their posts as part of a "channel", such as technology, politics, etc. Newsreaders able to parse this kind of information could provide users with additional tools to organize what they read. A shared taxonomy to define categories would make this process much more useful to the user.

Topics, or keywords, are a different approach. It's what we are working on (see ENT and K-collector).

The idea is to have a list of keywords shared among a group of users. The publishing tool will allow users to attach to their posts specific keywords, and aggregators will organize content according to these keywords. Keywords are also automatically attached to posts by different kind of filters.

I think that this approach has a significant advantage in k-logging environments, where the list of shared topics (which we are calling topicRoll) will not grow to match a full dictionary within days and where it's more important to be able to dig into the past content using a directory to navigate.

I don't know much about advanced content filtering. Bayesian Filtering seem a pretty interesting approach, also because it could be trained to find information relevant to use from other sources.

At evectors we are working on a reputation-based filtering system, where users of k-collector will be able to have their news filtered according to who is writing about some specific topic. It's still at a very early stage, but it sounds promising.

Whew... it looks like there's still a lot of stuff to invent and code to write, uh?