Flutterby™! : distributed net 2005-10-19 21:04:20.946574+00

distributed net

2005-10-19 21:04:20.946574+00 by Dan Lyke 7 comments

Doin' a little web development today, using WWW::Mechanize to stuff data into Scarab, and in the process I started pondering a little off-track.

In a recent thread, Mars said:

yeah, it's long past time to start abandoning big central servers and start swinging back around toward mesh architectures. Let's start with the DNS....

Obviously this is something that comes up occasionally, but it got me to thinking: One of the potential solutions to some of the search engine problems is starting to find ways to decentralize search, either by actually decentralizing the indexes, or by offering up starting web pages as a way to build cultural context: Rank links by distance from these trusted sites, that sort of thing.

Since we're finding that we need ways to limit cultural spaces in general, why not give names cultural meaning? Why not have a name service that can be used hierarchically, so that more specificity can be offered (much like "bang paths"), but such that words like domain names become known and defined in a context as people start using them for a meaning?

#Comment Re: [Entry #8302] distributed net made: 2005-10-20 02:36:03.266278+00 by: Unknown, from NNTP

john. -- "If you think this is a free country, try selling something." - Andrew Kantor, Senior Editor, _Internet_World_

I like the idea, I think I heard you say something similiar about 10-11 years ago. The problem is, 98% of the online population can't get past putting www in front of a web address, and .com at the end.

Every(site|host|one|service) has an identity, generated with a public and private key. Once you find someone, it's possible to know if you've re-found that entity.

Every user has a process which is regularly checks with other such processes for change information. These processes can be queried for resolution of a name to, say, an IP address (some possibly transient way of identifying a service), and a signature on their key.

Users have the ability to manage the list of who they trust, and the top of that list might be their ISP, and then a friend or two, and so forth.

Addresses may have some sort of bang pathing. Maybe that's even hierarchical scoping as in currently, so that if I say "flutterby.com." I'm saying "Find who you trust as 'com' and check for 'flutterby' under it", but if I say "flutterby" it's an "ask in your neighborhood" thing, not strictly hierarchical. It's also the case that my neighbors might not trust the ".com " registry, or that I trust a ".com" that mirrors most of the master ".com" but alters a few key things, and as long as I've chosen to recognize the signature of that neighboring ".com" as the one to trust and not the one that, say, AOL (or their subscriber base) recognizes then I get a slightly different view of the world.

There are some propagation issues here, I'm just rambling and brainstorming and this isn't a solution yet, but I think there's something in this that could not only be useful as a new way of looking at DNS, it could also be a way of re-balkanizing a system that has been losing a lot by becoming homogenous.

I got here by starting to think that it'd be nice to have a search engine that could restrict results to sites that have been pointed to by the weblogs in my regular reads list, or even subsets of that, so that when I was doing, say, Un*x mechanics type searches I'd say "trust Genehack and Flutterby", but when I was searching for pr0n I could say "trust Weird Ass Shit and Sensible Erection.

Okay, relatedly, I think it was Jorn Barger who once made a case that it was entirely reasonable to keep a text version of everything you ever read on the web. There's a certain speed you read at, and there's a speed of growth of storage, and they'll continue to diverge strongly for quite a while.

Oh yeah: This should be tied in to "zeroconf" or whatever folks are calling that local net discovery thing, possibly in conjunction with a LID

like thingie, so that exchanging identity keys and discovery stuff can be done by geography, as in on an ad-hoc network or a local 802.11 node.

Whoah, this deserves some more attention: Outfoxed, which appears to be a browser that...

#Comment Re: [Entry #8302] Re: made: 2005-10-31 02:56:03.016214+00 by: Unknown, from NNTP

john. -- genehack.org * weblog == ( bioinfo / linux / opinion / stuff ) * daily *

Don't sacrifice clarity for small gains in "efficiency". - The Elements of Programming Style (Kernighan & Plaugher)