Re: VR: Lat/Long of Internet Servers

Mark Waks (justin@dsd.camb.inmet.com)
Tue, 19 Jul 94 11:42:07 EDT


Here we go again...

To start off -- Ken, we had a *long* argument about "absolute
vs. relative space" early in this list's existence. It's not quite
precisely the debate we're having, but it's quite close; you'd do well
to dig into the archive and look at it.

>I'm glad you disagree vigorously -- it makes the argument more
>interesting. I stand by my words until convinced to withdraw.

That's only reasonable...

>You're confusing "hyperspace" with "cyberspace," but I'll forgive you
>that. Hypertext links should certainly be visible in VR cyberspace.
>But what should they look like and how do you follow them?

I'm not confusing them at all -- I'm quite deliberately using the
terms equivalently. The "hyperspace" model is the most natural one
to use for Net-space, I believe, since the topology of the Net
frequently bears little resemblance to real-space. As I said,
what's really important is the information, not where that
information is located.

> And how so
>you know where they lead before you follow down the primrose path? You
>argue that you don't care where they lead; you'll follow blindly,
>dunking your head in California and surfacing in Norway. I'd say
>that's a long way to swim if the same data is available in Oregon.

Not necessarily. Look, you're assuming that closer is better. I don't
see that at all. As you yourself admit later in your message, it
makes more sense to look at proximity in terms of bandwidth; a site is
"close" if the bandwidth between you and it is high. That often
has nothing at all to do with physical nearness -- my response time
to, say, CMU (around 600 miles) is typically rather faster than that
to my wife's machine 2 miles away.

Moreover, even that is really a non-issue. The real concern isn't
bandwidth, it's server capacity. You want to go to a relatively
unloaded server for best response, if there are multiple sites with
the same data. That will *often* not be the nearest server. If I've
got a choice between a well-known local site (say, MIT), or a less-
known and less-loaded one that's in Oregon, I'd much rather go to
Oregon.

And even *that* isn't really a big issue. Frankly, I'm *very*
unconvinced that we're going to see all that much data redundancy
in cyberspace. Yes, there might be common items that are replicated
all over the place, and there might be some merit in having a "local
library" for those. But I'd bet that the really interesting stuff is
going to be single-source, or at least look single-source to the
average user. Again, physical location is irrelevant.

>What does your VR view of the 'Net look like?

Nothing like anything you're thinking of. I want to be able to walk
into a "library", which has a number of "rooms". One of those rooms
might be an engineering library; another a fiction library. The
physical sites for those rooms might not even be on the same
continent -- I don't give a damn, what I care about is that they
are *conceptually* similar, so there should be at least one view
of cyberspace where they are "physically" close.

Or take a streetfront. I'd like to have a "street", lined with
shops; say a district that sells computer equipment. It would
have a bunch of different computer stores side-by-side. They
might not be in the same state locally, but that's irrelevant --
I'm buying from them mail-order, so I don't much care where they
are physically.

The point is that the really *useful* organizations are based on
similarity of information, not just physical proximity. Physical
proximity is sometimes *a* useful similarity of information -- I only
care about, say, pizza shops that are local. But it's only one of many
useful organizations, and often one of the least useful.

>My point is that you have your own, personal view of how cyberspace
>should look, and you should be able to build your VR view of the 'Net
>according to your favorite logical variables. I prefer a
>geographically oriented view, so I (and people like me) should be able
>to use that view. Geographic proximity can be very useful in queries
>and searches (see below).

Perfectly reasonable -- as I said, I have no problem with building
geographical applications on *top* of the system. What I regard as
entirely unreasonable is making it the *basis* of the system. You
implied (perhaps unintentionally) that every site must supply
lat/long, and be part of this "physical" net. I regard that as both
goofy and unreasonable -- if I don't want my site located into your
"space", you have no right to demand that I put it there. And if it's
not required of every site, it's not the basis of cyberspace -- it's
just another application.

This isn't a trivial issue. Take for instance the situation of the
Web. I just had a long argument last week with our Corporate folks.
They are *very* worried that unregulated employee home pages could
reflect badly on the Company; they're willing to provide some disk
space and bandwidth to us as a perk, but they don't want to be
responsible for what we do with it. We convinced them that it isn't
a big problem, because there is no connection between the company
home page and the employee ones -- they are conceptually clearly
separate. In your model, one would go to "Intermetrics", and would
find my "room" there. In that model, I am absolutely *certain* that
the company would not sanction employees using the company net for
putting up personal data -- the association between the company and
the employee's data is too obvious.

(A flaw, BTW -- what if the space needed for your application doesn't
map well to the physical space for the site? I can have an electronic
"library" that requires acres of virtual ground, to separate it out
into enough rooms to look and feel good. But that library may be in
a machine that sits in one small room, with other machines around it.
How do you arbitrate, when the "space" for two sites overlap heavily,
as will inevitably happen? The world isn't made up of lat/long points;
it's made up of volumes, and conceptual volumes don't map terribly
well to physical ones.)

The point is, your maps are *one* way of looking at cyberspace, and
sometimes a useful one. But it's riddled with problems, and it's
often far from the best way of organizing things. Making it the
*basis* of cyberspace is silly -- it's not the natural way that
the data works, so it shouldn't be the lowest level. It should be
a higher level of abstraction, *available* for them as want to be
on it and use it, but not *required* for anyone.

>One reason it's rare for you, right now, to ask for locale-specific
>information on the Internet is that it currently isn't accessible that
>way, so it's very hard to ask for it. Another reason is that there
>aren't a whole lot of servers in every local -- yet.

Unconvincing. I'm in Boston. There *are* lots of servers here. There
*are* local-oriented services here (weather, restaurant listings,
even some shops). I find them only marginally interesting; I use
the non-local services of the Web far, far more.

I also direct your attention to the studies that have been done,
several times, by companies doing locally-oriented data serving and
services. (Knight Ridder was the most conspicuous, as I recall.) All
were near-total flops. If something is that physically close, people
usually prefer to just go down and deal with it physically.

> If you consider
>the 'Net not just as a way of downloading nifty software and cute
>images but as a tool for improving everybody's standard of living, as
>integrated into the community, locale-specific queries, searches and
>postings become much more useful.

Excuse me, I don't much care for the implication here. I use the Web
extensively, for work, for education, and for play. I'm not into this
downloading stuff -- I *am* one of the people out dealing with
cyberspace for real, right now. And I find your structuring proposal
incredibly *limiting*; that's why I object to it so vigorously. Don't
structure my experience, and I won't structure yours. There's room in
cyberspace for both views, and I don't care for people telling me that
it has to be organized one particular way...

>OK, then: what is to be the fundament of cyberspace? Domain names?
>Obscure SNMP parameters? "Bandwidth" and "Popularity?"

Connections. Period. Rooms are hooked to other rooms however they
care to, and however makes sense; often a room will have many
entrances. I've written a substantial amount on this subject; please
go look into the archives.

-- Justin

Random Quote du Jour:

"God is not dead. He's just taking a nap."