In short, I wouldn't limit myself to Twitter, nor its API, nor tags or structured content, Nova, I would ETL all textual ressources permanently identified and locatable on the whole Web (like tweets, and for longer texts, just machine summarize back to 140 chars). The qualified resources would require some mandatory metadata (location, authorship, for instance).

ETL as in extract, transform and load, of course. You have something for search engines, if I recall correctly? This is my shot at your lazyweb request. Because you have to believe, at least a little bit, that flattering works for me… ;)

And there is a business case around that too (not the flattering, the ETL), if only as system administrators and integrators, because it cannot be a desktop app. The computational requirements are way too big to fit on a non-shareable machine with nearly 50% of wasted idle time. The computing and storage ressource would need to be pooled. That would necessarily be a web service on the cloud. Probably on a MapReduce architecture, like Hadoop for the computations. And Cassandra for storage and querying. Most definitely hosted on a commodity cloud, like Amazon Elastic MapReduce. And with a gorgeous, CSS/HTML/JS-customizable GUI by some typography genius (the thing is one type-intensive beast) like @Patrick Pellerin (he did with me the most interesting tagcloud-based website I ever saw, and we did some other nice ultraminimalistic dynamic interfaces for CMS together too) that would be like an OLAP hypercube… for the Hyperconversation. And, oh yeah, all visualization and drill-down parallels apply…

Basically, a channel, like you defined it, is no more than the set of hits (see Lucene for definition) in a categorization system, that is configured by the user. There could also be a machine-recommended cluster system of related things… Or better categorization parameters. All of this with the same Mahout I mentioned to you on the phone, which could add a nifty consensus/dissension aspect provided with included and robust sentiment analysis.

Mahoot runs on Hadopp. And Cassandra just recently added support for Hadoop queries over it's data… Now is the time.

You do see that the project can even be open source, to some degrees, without endangering the revenue stream, right? A little bit like profitable Shopify releasing Active Merchant, Liquid Markup, Active Shipping and Vision for free. Heck, it could even be a coop, and we'd still make a living out of it.

Authorship ancestry for memetics cartography is also an important concept. We have an expert on that in Montreal as well, @Martin Lessard (one I have been having conversations with on the subject for years, notably in the blog post you seemed to enjoy). There should be even some sort of Simile Timeline viz system with underlying concepts like originality, forking, debate, etc. Credibility, reputation, probably in relation with the concept of Whuffie, invented by canadian Cory Doctory, developped by now fellow montrealer Tara Hunt and now exploited in the Whuffie Bank is also interesting…

This is only 1% of the first phase, in a release early, release often strategy, because, such a beauty should not limit itself at content consumption. It should be integrable/interoperable into existing content production platforms like Facebook, Twitter, Drupal, Wordpress, Disqus, Google Buzz, Google Wave, etc., etc. The ping, notice and social network infrastructure should comply with open and FOAF-based (and Montreal-born) OStatus. It should offer content ownership to the web's individual authors and content gatherers. They should be able to backup, restore, dump and manipulate their data in an open format, both human and machine readable. I called this a dataisland. I had an Open Source MDA project on Source Forge, another lifetime ago, by that name. You can probably find traces of that, still.

IMHO, though, we will feel soon feel cramped in the current versions of the DNS, XML and HTTP protocols. That's why I pointed two domains that I own (http2.org and xml2.org) to a wikipedia page where I will (and you can too) document all the criticism opposed to those fundamental (and fundamentally flawed) elements of the web infrastructure.

I think (I hope) I gave you enough clues to convince you that a real conversation, in person, in Montreal, is in order. Will you join me for the next Webcamp in Montreal? Because I think you should. You have my number, I believe.