Off the hip morning coffee commentary on two freshest hypes.
Powerset: It is difficult to figure out why behind questions. And people often make mistakes in forming their questions. In the end, I think Powerset search results will disappoint and Powerset users will end up typing in keywords. If their target market is occasional computer users, they have to be the default search engine on new computers, but I doubt they have the financial means to make that happen.
Pipes: I had a similar idea awhile back for which got the feedcrew.com domain. I stopped working on it when I realized the core problem which FeedCrew, Pipes, and Ning shares: success will lead to failure, meaning popular pipes will be cloned elsewhere with better tailor-fitted user experience and hand-coded performance. In the end, pipes will become too much of a resource drain to continue and Yahoo will have to shut the service down. Another case of reality tormenting geeky ideals.
Don't need scaleability - do it out of the box. Need to serve millions of users - get to work on it. (BTW: Feed processing is quite suited to map-reduce, so you could certainly add that in there.)
Given all that, a local pipes should satisfy all but the most demanding feed cases.
While Pipes is cool and useful to developers, there is nothing to keep developers from escaping once they used it to prototype something useful. Unless Yahoo introduces some kind of pipe-generated ad-revenue profit sharing plan (PipeSense?), Pipes will be used then left-behind.
I ask again, Is Yahoo Pipes good for Yahoo? Beyond feeding spinsters and media hype, my answer is no.
To justify this I'd throw in terms like personalisation, loose-coupling, decentralisation, dynamism and emergence. Oh, and complexity-hiding. I'll spare you the essay.
But as an Pipes-like example that points in this direction I personally use a set of Firefox tabs for the bunch of resources I visit regularly. I've got Bloglines in there, which allows me to keep track of the posts of individual bloggers. I've also got a couple of Planets, the River of News style being suitable there, plus the logs of an irc channel. Gmail gets a tab too, as does my own blog - both of which I can post through. In effect I'm using these different tools to mash up my own i/o as it suites me. It's all very decentralised, until it gets to the browser. I can add/remove sources just by tweaking bookmarks.
For an example of something like Powerset working, imagine it meshed into Google's search. If (according to attention metrics) the user seems more successful with natural language queries, results produced by that pipeline could be transparently favoured over those produced by keyword search (or vice versa). For the end user it needn't matter if the work was being done by a single service or three different ones (Google, Powerset, the meshing engine). Poogle Ads might get distracting though.
It's too easy to forget the dusty road to the hometown when trucks are roaring down the freeway that runs right by your house.
