| Have you heard the World Wide Web being referred
to as the World Wide Wait? Many people complain about the Web being
slow (they should have tried using the Net back when a 1200 bit
per second modem was considered a screamer!). Companies want to
put more graphics on their Web pages, more animations, more Java
applets, more audio, more video...you get the idea. But each of
these increase download times. So, I'm often asked, when will the
Web be fast?
The question "when will the Web be fast?" won't get
good answers...it will never be "fast enough" just as
computers are never fast enough and never have enough memory.
And no one is going to wait for for the Web to get fast before
they start to do interesting things: we will all try to squeeze
the best value out of our opportunities. More useful questions
are, "What is the best we can do with what we have?"
and given that technology is changing rapidly, "What should
we plan to use next?"
First consider the connection from consumers to the Net--usually
the slowest connection between seller and buyer. When will that
connection be substantially faster, able to support more data
intensive communication?
The answer to this question depends on two things: technology
and diffusion. There are lots of technologies available today
to increase transmission speeds. And there will be new ones, as
yet undefined: the 56K modems, for example, are based on new ideas
about the way people use modems today to connect to online services
rather than connecting from one modem to another, unanticipated
a few years ago. But the second question is how long will it take
for the new technologies to be adopted. ISDN has been around for
more than 20 years and is still not in widespread use. ADSL promises
incoming bit rates comparable to and higher than that needed for
VHS-quality video. How long will it take to diffuse into widespread
use? Cable modems offer households 10 Mbps connections--as fast
as a LAN--but if the cable companies force too many people to
share that bandwidth, the effective bandwidth could be no faster
than a 14.4kbps modem.
Another issue is that in many cases, transmission delays are
largely due to the speed of servers, not clients. There are few
servers which send data over my 56k ISDN line at full capacity.
Many companies make the mistake of building a nice Web site but
then host it on a server inadequate for the level of traffic or
behind a Net connection that gets choked with traffic.
I don't have a Moores Law-like graph showing how household connection
speeds change over time. In fact, I did work on one some time
back only to find that modem speeds changed in fits and starts.
But today there is high incentive to come up with faster technologies
and get them to market faster than ever before. I think we can
expect even more rapid change than we've seen.
Although we may not be able to predict with confidence that
most households will have connections of speed X by year Y, we
do know that the speed will continue to increase.
But the main thing is to engineer our sites to get the most
performance out of whatever pops up. If you've done much surfing
on the Net you've noticed sites that go for glitz: their Web site
demos probably look really impressive for a presentation to the
Board of Directors, but typical Web surfers are disappointed by
the poor engineering: too much glitz, too little regard for download
time.
Here are some examples of companies doing it right. Amazon
Books has a very simple site (too simple, perhaps). But they
offer several innovative services like alerts for newly available
books and the ability to set up micro-bookstores on one's own
site. Yahoo! does a great job
of coming up fast, again partly by simplifying the page design.
Microsoft and Netscape
each get tens of millions of hits per day, have some graphic content
on their pages yet deliver those pages very quickly. Their approach?
Lots of raw computing and communications power.
And bear in mind that fast may not always be best. Because to
get fast, you're either talking simple or expensive. "Appropriate"
is best, even if it may sometimes be a long download. If there
will be a wait, warn your users and make it worth their while.
©1997, Harry Tennant & Associates |
