[IP] Client/Server and the BAD IDEA OF THE WEEK
------ Forwarded Message
From: Brad Templeton <btm@xxxxxxxxxxxxxx>
Organization: http://www.templetons.com/brad
Date: Sun, 24 Apr 2005 12:34:10 -0700
To: David Farber <dave@xxxxxxxxxx>
Cc: <whitney@xxxxxxxxx>, <lauren@xxxxxxxxxx>
Subject: Client/Server and the BAD IDEA OF THE WEEK
> With Google only storing history if you log in, as well as offering a
> "pause" functionality to allow users to keep certain searches out of
> their history, this strikes me as a pretty reasonable implementation
> (though I'll admit that I've only played with it a very little bit).
A frequent computer engineering debate occurs over whether it is
best to do applications on the client or on a central server. There
are various reasons to go in each direction. Server apps can benefit
from economies of scale, and central maintenance. Client apps can
provide faster response and better UIs at the endpoint, and those
familiar with the end to end principle will have read other convincing
arguments. When it comes to searching the internet, the case for a server
is overwhelming.
But there is another, non-engineering question which plays a role in the
client/server decision, namely privacy. There is some data that just
makes sense to be on your personal endpoint. Or, if that's not easy
to do from an engineering standpoint, data which should be encrypted in
the server and accessible only using keys kept on the endpoint or on
the person (or in the brain) of the user.
So the question may not necessarily be whether engines should do
a search history. (Though my own search history is indeed frightening.
I recall one point where, to see what Google would sell ads on and
what they wouldn't, I deliberatley searched for kiddie porn and all
manner of other unpleasant things.)
The question should be, "How can we put the private information into
the client?"
Your own search history might serve you well -- stored by your browswer
or a plug-in to it. This assures that you retain control, and that
attempts by others to access it have to go through you.
Storing such things (and search history is only one example of a class
of items) on the client sometimes is more work, and gives up some
advantages of server computing. But in this case it's worth going that
extra distance, and possibly sacrificing a feature here or there.
Another good example are all these "unified identity" projects such as
Passport, Liberty Alliance etc. Why have a central server store my
personal data. Here the win for server apps (namely roamability and
ease of software maintenance) is pretty minor compared to the risk. The
effort to just build tools to manage your identity on the client makes
much more sense.
------ End of Forwarded Message
-------------------------------------
You are subscribed as roessler@xxxxxxxxxxxxxxxxxx
To manage your subscription, go to
http://v2.listbox.com/member/?listname=ip
Archives at: http://www.interesting-people.org/archives/interesting-people/