> For what it's worth, the *original* Internet (okay, ARPANET) was
> quite "centralized" and, in fact, had "surveillance" (albeit of a
> very small group of researchers who had grown reluctant to travel
> to "brain-storm") as (one of) its primary goals.
This is a strange thing to say. Centralised in its funding, in the
early days, sure. Centralised architecturally? Not so. Small enough to
easily hold in your head? Sure but that didn't last long.
> By the time I brought AOL public in 1992, its entire profits were
> the result of HOT CHAT, which was superceded by AOL becoming the
> primary site for accessing PORN sites, since they had the largest
> server-farm and, therefore, the most room to cache "pictures."
So porn figures importantly in the demise of usenet. Here I disagree
with Dmytri's implication that usenet died because of its
decentralised nature. As I remember it, being a sysadmin and network
engineer, it was more a simple question of expense of running a full
feed. alt.binaries was big. The news server took a lot of disk space
and consumed a lot of bandwidth and accounted for a small portion of
revenue. The decisions at ISPs to stop running news servers were taken
pretty much on those terms and had little to do with thinking about
centralisation or lack thereof.
The reasoning that said that running a web site and trying to keep the
eyeballs to sell to investors was a better model than things like
usenet was after the fact. It was a line of thinking that happened at
the financier level, after all it wasn't an either-or question for the
ISPs who were, in general, in the business of infrastructure not
running web sites.
> So, there's "surveillance" (like the don't pass go, directly to
> jail type -- for instance) and the "I've got all your clicks but
> don't know what to do with them" type -- which is exactly where
> Google and Facebook are today and will likely be 10 years from now.
These two kinds of surveillance fit neatly into the ISP and the web
site categories. The former is getting ridiculous. Here (UK) we seem
to have taken some bizarre and dangerous position that Orwell's novel
is more of an instruction manual than a cautionary tale. For example a
recent report from a parliamentary enquiry  is seriously suggesting
an Opt-In content filtering scheme to be implemented by the
providers. Think about it. We already have passive surveillance and
DPI, but now we have to raise our hand and ask permission to look at
anything that some civil servant thinks might be objectionable.
This drives centralisation in the infrastructure. These measures are
expensive computationally and administratively. A network operator
might rightly think it is easier to manage such things centrally.
The latter, "I've got all your clicks but don't know what to do with
them although I'm pretty sure they're useful or valuable in some way"
is what drives the centralisation of the web services.
The previous iteration, where users were expected to consume but not
produce, and we had traffic flows that went from the big web sites to
the eyeballs but not in the opposite direction also drove
centralisation, aggregation of demand in the infrarstructure.
Thankfully this is beginning to change as people start posting more
pictures and movies and stuff on web sites (however centralised they
may be). Because it starts to mean that the asymmetry that was built
into the DSL networks when ADSL was chosen over SDSL starts to be less
practical. The structure of the network starts to become more
balanced. It starts to become practical to actually serve stuff from
your home computer and keep backups on your friend's because you trust
them more than some giant overvalued company that has little regard
for your interests. Same with passing messages to friends and
collegues, where the path that a message takes mirrors more closely
the path that it would likely take with old-fasioned word of
mouth. The "cloud" starts to diffuse. "social networks" like facebook,
linkedin and google+ become redundant and wither away...