I think we have different usage models in mind. Mine is that most of the people
who will soon get WWW through the commercial services already download FAQs,
software, and other resources via established means such as FTP and BBS-like
menu facilities.
The "cost" of WWW, from their point of view, isn't the *entire* cost, it's
the difference in time cost between using WWW as an interface and using (say)
the menu-driven front end for the services's conventional download libraries.
That difference may not be significant. It may even be negative, if they
disable graphics and get better time efficiency at finding things from WWW's
link structure.
> The point I was trying to make was that, within a few years, the vast
> majority of the Net population will be paying by the byte. We all hear
> endless media hype about how everyone will have access to the
> Information Supercollider [*1]. Who do you think is building all those
> new "on-ramps"? Universities? Philanthropists? They're set up by
> commercial organisations, who expect to make a profit out of it. That
> means paying for *bandwidth*, i.e. pay by the byte (because the Net
> services themselves are paying the telco for *their* bandwidth).
That myth again. People have been predicting charge-for-bytes for ten years,
without noticing that it has never actually happens --- the commercial ISPs
all charge either monthly flat rate or by connect time. Reason? The
accounting overhead for charge-by-byte is so high that the game is not worth
the candle.
Go on, prove me wrong. Name *one* ISP that charges by the byte. If
it were economically viable, don't you think someone would already be trying
this price structure? As you point out, these people aren't altruists. They
know that even if the overhead weren't prohibitive, the outfits still charging
by connect time would eat their lunch.
Actually, the coming thing is monthly flat rate with prices in the telco
range. The cable companies have been salivating for this market for years,
and some of them are starting to actually do it now. First rule of a free
market -- as competition increases, consumer prices go *down*.
> True, WWW is no more expensive than getting the same information by FTP.
> So why do people bother with all the extra trouble involved in setting
> up a WWW link? Because of the hypertext (i.e. easy access to a *lot more*
> information), the pretty pictures, and the GUI -- in short, exactly the
> things that eat up bandwidth.
>
> It's Hobson's choice: If you just want a single FAQ file, you might as
> well use FTP. If you want hypertext/graphics/etc, you suffer pain in the
> hip pocket for your troubles (or will in the future). *That's* why I
> think WWW, in its present inefficient form, will never take over the Net.
I think your analysis is incorrect, and that the flaw is demonstrated by
the popularity of Lynx. The big win in WWW isn't the bandwidth-eating
pretty pictures, it's the cheap hypertext links. My FAQs have nary a graphic
in them, but are significantly improved by the link faility.
Consequence: heavy use of WWW will raise costs only in proportion to the
user's *voluntary time commitment*, not because the medium is more expensive.
Again, I urge you to think of httpd as a replacement ftpd with a massively
better interface. From an economic-utility point of view, the graphics are
just a distraction.
> So how would a more practical global-hypertext system work? First, we
> need more efficient compression techniques, especially for graphics.
> Photographs/paintings -> fractal maps, line drawings -> vector graphics,
> that sort of thing. We need *lots and lots* of mirror sites. We need
> smart local servers that can figure out the cheapest way of getting a
> given item.
>
> Above all we need to get away from the URL concept, in which every link
> has a specific address hard-coded into it! The link should specify an
> object to be fetched, and the smart server figures out where to get it
> from (and if fetching it would cost more than some user-configured
> threshold amount, it puts up a little box saying, "This download will
> cost you $1.99; OK/Cancel?"). URN looks like a promising start here.
I agree with all of this, except I think your retrieval-cost popups
will reflect not byte-transmission costs but royalty structures.
-- Eric S. Raymond <esr@locke.ccil.org> WWW: //www.thyrsus.com/~esr/home.html
[
Usenet Hypertext FAQ Archive |
Search Mail Archive |
Authors |
Usenet
]
[
1993 |
1994 |
1995 |
1996 |
1997
]
© Copyright The Landfield Group, 1997
All rights reserved