Re: Dead Internet Theory

Ben Collver <bencollver@tilde.pink> writes:

From the transcript of an interview with Jason Scott:
>
What I want to get to though is the fundamental lie of the
Internet, which is that the internet is decentralized and that it
functions as a a discrete series of interrelated parts that have to
some extent some sort of temporary balance of power between equally
powerful groups that causes the miracle of this interrelation to
happen.  But if you look at any aspect of it, it's been
centralized: digital certificates, domain names, network
allocation, and other aspects more Gentile like social media
accounts or being able to talk with central communities that exceed
anywhere past 5 to 10 million people.  So anytime you get up to a
certain number it just starts centralizing [...]

It’s worth thinking about how it could be otherwise (and there are real
attempts to make it so).

If you want globally unique names then a central naming authority is an
easy way to do it. Examples are the DNS (or more precisely the DNS as
reached from the IANA root) and the Internet PKI (technically there are
several central authorities here - browsers and OS vendors - but the
number is very small.)

It’s not the only way; if you generate random ‘names’ of sufficient size
then the practical effect is global uniqueness - but they tend to look
like R+eZuTq1R5C8gniEcPmU, not much good for everyday use by normal
human beings.

Usenet as a technology has a somewhat centralized model: for the most
part if a name exists in two places, it means the same thing, but it
isn’t guaranteed to exist in any particular view of the network. As
implemented, most sites delegate to a centralized authority in the form
of a standardized.

PGP attempted truly decentralized key distribution, but adoption is
negligible, even though it’s been available for decades. The poor
quality of the software doesn’t help but the key distribution mechanism
is just plain impractical outside very niche use cases. (Signal lets you
do in-person public key distribution, but in practice it mostly
delegates to phone numbers for identity and ‘trust on first use’ for
keys.)

I think the core point here is that centralizing authority works and,
mostly, attempts at decentralizing authority do not work. So...

But if you look at any aspect of it, it's been centralized

...it was, inevitably, always going to be centralized.


(Decentralizing implementation can work relatively well, in comparison -
DNS and PKI both being examples.)

--
https://www.greenend.org.uk/rjk/