Re:Wikipedia Will Fail in Four Years (2006)
by Steko on Saturday September 14, 2013 @01:49AM
(#6384799)
"torrent of pointless, trivial content that readers have to sift through to find the real stuff"
I don't think people page through online encyclopedias that way. I'm not notable, if I had a wikipedia page I don't think it would link from any page that actually exists today. How could that ever bother you? I don't think you'd come across it unless you googled my name and it took you to my wikipedia page in which case it would seem to be doing good.
Re:Wikipedia Will Fail in Four Years (2006)
by DanBC on Saturday September 14, 2013 @02:22AM
(#6384864)
I used to hit the [random] link and know I'd get something interesting.
Now, not so much. I'll get a tiny stub of something programmaticly dragged in from some huge database - a town name with maybe some population figure and location; an obscure politician with party affiliation and birthdate.
For people who enjoy gnoming these kind of articles are tedious - what's the point of correcting a comma if no-one is likely to see it?
Re:Wikipedia Will Fail in Four Years (2006)
by saraid216 on Saturday September 14, 2013 @04:46AM
(#6385075)
Supporting the random button isn't what I'd call anywhere near the top 10 most important functions an encyclopedia of any kind ought to provide. Serendipity is useful in a library or a bookstore, but when you open a book, the bibliomancy becomes avant garde art at best.
About 15 years ago some friends and I started a company whose technical premise boiled down to taking IRC and replacing the static tree-based "routing" system it has with a real dynamic routing protocol. No more netsplits! Arbitrary topologies! It was so simple! Why would anyone want anything other than a biconnected IRC? It turns out that the moment you lose the static tree property, the whole system goes to hell. All the sudden, messages aren't (and can't be) reliable, because the network can change out from under them while they're being forwarded. So you figure you'll just build a simple reliability scheme for messages to ride on. Oh, wait: IRC is a group messaging system, and not only is the distributed systems GMS problem not particularly easy to build in a "realistic" network, but providing reliability on it is essentially the multicast reliability problem, which is itself so annoying that it's part of why IP multicast failed.
My long-winded point is, sometimes something that seems like an obvious weakness of an existing system is actually fundamental to that system's viability.