On the question about how a search engine would work. The Dat team have thought (and implemented) quite a bit about building performant indexes for distributed file systems. I've wondered if we relaxed the time requirements for search whether there are other processes that we could use that would e more robust. "Who has 10.19.20.13?" seems like it might be a little bit too low level, but what if it was "Who has the most reliable data about the publishing industry between 1500 and 1600?" expiration in 30 minutes with an extension of up to 90 minutes for each preliminary result with more than two nodes responding.
Could it work? What kinds of use cases could it support? How would a bot that had a good model of which forums to post things to do in comparison if it had 2 hours, 24 hours, 1 week? What about if we put _lower_ bounds on the amount of time that had to be spent looking (this seems like a fun challenge for proof of 'real' work).
While this is at a much higher level than the current dat protocol, it is in a sense quite similar to the query "Who has 353904391670d2803b34990e37f4d2e96f49351998e162d0e335b16812daf592e0f71470af7bee31f6a1da03744d03bcde659d73a0ebf56fd4a9fc6ef67edf60 that is 5 bytes long?"
> "Who has the most reliable data about the publishing industry between 1500 and 1600?"
So, to verify I understand: the backend of this would be handled by grad students (working with some kind of 30/90 minute timer alarm), and would experience distributed consistency failures caused by differences in which departmental tradition they did their major area exams?
Pretty much. Though for that question you might have to go all the way up to tenured faculty, which could incite a full on Byzantine war over which particular source was more reliable, however I think at that point it is clear that we have lucked out and found two really great sources of information instead of just one and potentially triggered the creation of a third :)
Part of the 0.9 release of beaker will be the https://unwalled.garden spec, which is a bunch of file formats geared toward improving discovery and search (and general applications). Beaker 0.9 has some internal FTS indexes that let you query content published by people you follow. It's a kind of social search.
Could it work? What kinds of use cases could it support? How would a bot that had a good model of which forums to post things to do in comparison if it had 2 hours, 24 hours, 1 week? What about if we put _lower_ bounds on the amount of time that had to be spent looking (this seems like a fun challenge for proof of 'real' work).
While this is at a much higher level than the current dat protocol, it is in a sense quite similar to the query "Who has 353904391670d2803b34990e37f4d2e96f49351998e162d0e335b16812daf592e0f71470af7bee31f6a1da03744d03bcde659d73a0ebf56fd4a9fc6ef67edf60 that is 5 bytes long?"