According to a Sun document I came across online (
http://www.sun.com
and
http://java.sun.com
), BusinessWeek
thinks that Java is a programming language that will "turn the Internet
into a giant disk drive." I can't wait.
Of course, the whole Web phenomenon is in many ways a new kind of document-specific
file service. Each HTML page is just transferred to your screen via your
HTML browser from a big disk out on the Internet. The Web server is just
a new kind of file server. It serves up .htm inste
ad of .doc files. Now
with Java, it will serve up .jav files instead of .exe and .dll files. This
is one reason Novell might turn out to have a successful Web server; Novell
has done file service better than anyone else so far.
Unfortunately, all the parts of a single document--the HTML text file, multiple
GIF or JPEG pictures, the Real Audio clip and so on--are stored separately
on that disk. The Web desperately needs a compound document solution. I'm
tired of posting five files to a server when it only took one in Microsoft
Word.
The Web is certainly more like a file service than terminal emulation or
X Window, despite the fact that many people now view the Web as the next
greatest dumb terminal option for corporate applications. Attempts to make
the Web more interactive are still being fleshed out (hooking Web servers
to databases, adding server program process
ing and so on). My general advice
to those who think we can get back to terminals of any kind is this: Remember
client/server
. The Web isn't going to kill Windows. It'll just change the
way that some applications are deployed.
Nevertheless, it is this Web server as file server that scares me. And,
it should scare you. Particularly if you're thinking Java is the next best
thing since sliced bread. It's not. Instead, with Java the Internet will
gain weight. Because of this, the Internet will also gain wait.
Java Gets Bigger
I checked out the CADIS demo (
http://www.cadis.com
)
showing Java in client/server mode. I'll talk more about that in my next
column, when I begin to discuss the more specific middleware implications
of Java. For now, though, I have a more fundamental problem with Java: Serving
up application executables over the Internet this way just won't work.
Witness my CADIS connection. First of all, it downloads a 510-KB file to
my local machine. Now there's a low-calorie meal! Even with my 64-Kbps ISDN
connection, it took at least a minute befor
e anything other than "Krakatoa
Search Engine Starting" displayed on my browser. Over a 14.4-Kbps or
28.8-Kbps dial-up connection, will anyone wait? I personally can't even
stand home page graphics over 50 KB at typical async speeds. Frankly, anything
that takes more than five to 10 seconds to show up has lost my interest.
I've already forgotten what I asked for by that time. Studies have been
done on this human factor issue: People don't wait.
What was eventually displayed wasn't great either: The demo mentions that
there are known display problems on Windows95 and Netscape 2.0. Where's
the cross-platform support we've been told to expect? Is this just a problem
with the Java interpreter in Netscape's Windows95 browser? Let's assume
that--programmers being programmers--this will be fixed very soon. I worry
about things much harder to fix--like the entire Internet infrastructur
e.
The Internet doesn't always work, does it? Who knows how many hops are between
you and the server
you're trying to reach? Sometimes you get connections
quickly, but sometimes it takes a long time--even to the same host. Sometimes
there's a problem somewhere along the line (overloaded router, line failure
and so on) that's simply made a commonly used resource unavailable, for
a while. Maybe it'll work later. And maybe it won't.
The Internet may be many things to many people, but a reliable network it
is not. Can you run a business (or even live as an individual consumer)
with a fundamentally unreliable and ultimately unmanaged network as the
basic platform? The Internet is a playground for most. If it rains, you
can't play.
After we all get over the fact that it's so cool to see all the different
front pages, we settle down to wanting to do things regularly. And we expect
a level of service. We're no longer just browsing. We're using or even buying.
Also, many sites haven't scaled up to handle the many connections users
are throwing at them. Check out the Microsoft or ESPN
et.SportZone.com sites--they're
always slow. I don't wait in lines at stores much (except when forced to
around Christmas). I go to another store or I don't buy at all.
The Internet Makes Me Wwwwait
Now you're telling me that every site
will need to send me a large Java applet each time I connect? I'm not looking
forward to waiting longer for that fat content. Instead of energizing the
Net with a bolt of caffeine, Java will act like a depressant, and will bog
down the whole thing: not only at the servers, not only on the Internet's
backbone (or your ISP's backbone), but even at the user's own link.
Java applets are going to be too big for the real Internet. It's fine for
researchers with Sun boxes on Ethernets in the universities to get excited.
Sitting here--still waiting for the download--I'm thinking Java is not worth
the time it's going to take to experience it online.
Distributing Java Applets
Java is certainly beginning to sound like
it'll make the W
eb server into what we now comfortably know as a NetWare
file server with lots of Windows DLLs on it. But, in this case, we're expected
to have one server somewhere, a huge number of hops away, that has the only
copy of the DLL we need to retrieve--and we should retrieve it over our
28.8-Kbps dial-up line. Maybe we can call this dumb file service, as opposed
to dumb terminal emulation.
To save the Net from Java, someone's going to have to figure out how to
keep users from downloading the same Java applet every week or every day,
even when the applet hasn't changed. All this local caching and even the
new Web page-explicit local copiers can't cope with multiple 100-KB to 1-MB
"applets" needing to be downloaded. What size disk must a local
user have? Can the caching approach be handled by more local servers in
an ISP?
Today we only think of the Web as an interactive data access mechanism,
not as an interactive program distribution mechanism. Java will change this.
It makes the We
b into even more of a file server.
I think of this every time I try unsuccessfully to connect to Microsoft's
site to download the latest copy of Internet Explorer. At least Netscape
has about 20 sites from which to download code. Did you ever wonder how
to figure out which site might in fact be closest to you? When I connect
to PSI in Texas and try downloading from the Netscape Texas mirror site,
is the 2.5-MB Netscape ZIP file going through PSI's central hub in Washington,
D.C., or does the traffic stay in Texas? It matters.
Partitioning the Web
For the Internet to be able to keep traffic
localized or partitioned, data--and now programs--must be distributed closer
to the end users. ESPNet should be in lots of places, not just one. We can't
just say we will throw bandwidth at these problems. We will never have enough
bandwidth (or at least we will never have enough money to pay for enough
bandwid
th).
Moreover, a directory service that would make distribution and repl
ication
transparent to the user is a requirement. Lotus Notes has location-independent
doclinks along with replicated data stores. Clicking on a doclink takes
the user to the most local copy of the referenced data. This is useful infrastructure--infrastructure
the Web will require to scale.
Without it, URLs will always be hardcoded to specific physical locations.
Users won't realize there's a closer copy of the data they want. Therefore,
bytes will again flow needlessly across the Net. What a waste.
Moving Beyond the Buzz
There's so much about the Web and the new
fave buzz-thing Java that won't scale well. Why isn't Novell, for example,
making more noise about using NDS to manage user accounts across multiple
Web servers for single sign-on and simplified administration? Now, managing
users of multiple Web servers is like managing too many NetWare 3.x servers--all
separately. Why isn't Oracle touting its replication technology as the underpinning
of its Web offerings? Developers
are driving Web technology now; deployers
and, more importantly, users are going to suffer from their myopia.
Yeah, Java is really cool technology. But it's going to make things worse
before it makes things better. In the meantime, think of Java as just a
programming language. Don't think of it as the future of the Net. The future
of the Net will require technology with less glitz and more guts. It will
require a serious distributed infrastructure overhaul.
I can't wait.
Bruce Robertson can be reached at brobertson@nwc.com.
REPORTS
Analyize In-Line NAC strategies and products.
ANALYTICS Plan and design your enterprise blade server deployments
2009 IT Salary Survey: Meager Raises, Solid Prospects
Though raises are notably smaller than a year ago, and job security’s shrinking, IT careers are looking safer than many others in this economic downturn. Get all the findings in InformationWeek's 2009 IT Salary Survey. Available FREE for a limited time.