Thin client apps not the future?
As my picture shows, I'm old. Thin clients come, they go. Fat clients come & go as well.
Distributed apps wax and wane, the number of tiers in the architecture grows and shrinks.
The "Network Computer", remember that ?
So crap Scott Adams devoted a good chunk of a Dilbert book about how thin clients were obviously the work of weasels.
My main computer is a fat client, really fat, so fat that the server software tripped a Bloomberg alarm that disabled their shit s/w because it thought I was running a server farm to steal their data.
That's worth knowing because both first and second line support will tell you that there's no reason BB terminal won't work on Windows Server 2003.
That's bad, of course, normal people don't have three versions of Excel, 2 of Visual Studio, VMWare, or rig the graphics cards to do fast floating point

It's entirely unsupportable, and if I cost my time properly, it would need replacing about every 6 days. But I'd have it no other way.
At any point the current trend is the result of 3 1/2 main factors.
1. Communication speed vs processor speed.
Sometimes things are best done in a fat client because shipping the data to the server is hard.
Sometimes the data is too big to ship to the client so we have thin clients, fat servers.
2. Style of development. If you want stuff that runs 24/7, you have thin clients, but you're going to have to agree with everyone else what it does. Tactical development is harder with centralised systems, that's why we have Excel doing stuff that it is absurdly unsuited for.
3. Politics
When I first wrote code there was no such thing as a personal computer, everything was centralised. In 1986 I saw a change request form. It was to change a spreadsheet. Not as you might think to install a new version, no, to change some cells. Centralisation had meant that everything was locked down. A lot of development is done by quant/front office teams because now that we are in May, any request for a change is unlikely to come about this
year.
But central systems can be more cost effective and secure.
The half factor is the perception by developers of what is cool and will look good on their CVs. I wrote an editorial on Java in the mid 1990s for PC Magazine that talked of the "career advantages" of Java. This whilst I was editing the
C++ review...
That was true for the techie audience I addressed, and is true for non-quants even
Tell that to Microsoft, which is pushing the daylights out of Silverlight.
Actually I was talking to a senior Microsoftie a few days ago about this. Silverlight is indeed being pushed hard, partly because it needs pushing. The suck is not great.
This sort of stuff is again en example of the ebb and flow as the ratios of the vectors change.
Games are becoming thin client, as is music as firms like Apple, Amazon, and Microsoft, and EA games try to protect intellectual property. The "pushing updates" sounds good until you find Adobe or Microsoft have turned off the server that permissions you to use it.
If you think your central IT would not do such a thing, look at the interview with me above, I've been a CIO and quant developer in the trenches, and it can and will happen.