An advantage of web apps

I’ve got Microsoft MapPoint installed on my system, but it’s faster for
me to get a map for a given location using Google Maps.  I’ve got
a Desktop Search tool installed, but it’s faster to find something on
the Internet (searching billions of pages) than it is to find something
on my own computer.

This is because of application launch time, and paging.  If I’ve
been busy on my computer, it has to do a lot of swapping to bring up a
new application (and I’ve got a gig of RAM).  When I want a map,
the only software my computer needs to run is a browser.  The time
it takes to download the map is less than the time it takes to launch

The same is true in almost every instance.

Remote Desktop makes it possible to use a remote “rich client”
application on my local system.  With a little tweaking to make
the remote application seem more like a window on my desktop than a
complete remote desktop (more like X Windows), you could run almost any
application remotely faster than you could run it locally, simply
because the hardware that’s running that application is dedicated to it.

This seems like a more promising future than the current AJAX fad and
trying to pretend the browser is a rich client.  Browsers still
have some serious drawbacks when it comes to creating applications.

And for the rich vs reach argument; having a rich client application
running remotely means I could also install the rich client application
locally and take it with me (assuming it’s not something that needs
data from the remote hardware it’s running on).

One of the compelling things about ZIM is that it had a thin client
runtime that would remote the GUI using very high level constructs
(like “put this button here and tell me when the user clicks
it”).  The admin could run the app server on the local box, or on
a remote box, and the user didn’t need to know the difference.

There’s still no really rich client tier for a 3 tier or n tier
architecture, as far as I can tell (admittedly I haven’t searched