James Avery suggests creating something new for topic-based aggregation. His idea is to allow aggregation of blogs by topic and not by person and wants new things invented for this.

I think they exist.

The item-level <category> element of RSS allows for setting a taxonomy on category items and that's sufficient to make James' idea work, if  (a) aggregators were looking for the domain attribute and allowed grouping by it and (b) blog engines would allow you to attribute your categories like that. Then, we "only" need to agree on common categories (or someone just writes some up) and were set. No need to register anywhere or with a central system. Here's an example of how the elements could look for .NET bloggers:

<category domain="urn:msdn-microsoft-com:netframework">ASP.NET</category>
<category domain="urn:msdn-microsoft-com:netframework">Enterprise Services</category>
<category domain="urn:msdn-microsoft-com:netframework">C#</category>
<category domain="urn:msdn-microsoft-com:netframework">VB.NET</category>
<category domain="urn:msdn-microsoft-com:serversystem">SQL Server</category>
<category domain="urn:msdn-microsoft-com:serversystem">BizTalk Server</category>

Dave Winer suggests an experiment:

Shall we run an experiment is to see if aggregators can work with RSS feeds that have a xmlns attribute at the top level, on the <rss> element?

... and continues with an example:

<rss version="2.0" xmlns="http://blogs.law.harvard.edu/tech/rss">

Now, the RSS spec doesn't say that this is okay, but neither does it say it's not okay.

Yes, the RSS spec may not but that doesn't matter, because it's just a vocabulary on top of existing specs that take matters a bit more seriously. The XML namespaces spec says: "If the URI reference in a default namespace declaration is empty, then unprefixed elements in the scope of the declaration are not considered to be in any namespace", which is true for all RSS elements as per RSS specification, because it ignores namespaces and is therefore subject to this default case. Therefore, setting a default document namespace like that may be permissible as per RSS spec, but recognizing such a document as valid RSS is just wrong. I would suggest to revise the spec and not to experiment.