Hi,
earlz wrote:I do not see the point of a Browser OS. Your already going to have to have a lot of drivers and other support to get a browser and such even running, why not just make it a general purpose OS? Plus who wants to buy $200 worth of hardware that is useless without the internet?
Once upon a time lots of companies used dumb terminals (e.g. a main computer, with clients using something like VT52 over serial cables). The main reason for this is that computers were extremely expensive. This changed - computers became cheap, and now it's cheaper to have many "modest" desktop computers and a "modest" file server than it is to have a large server that's powerful enough to do all the processing.
However, lots of companies do still use the modern equivalent - "thin clients" (e.g. a main computer, with clients using RDP, X or VNC over ethernet). The main reasons for this are security (e.g. someone can steal the "thin client" without getting any of the company's data) and it's easier to maintain one computer than to maintain lots of separate computers. Basically, the cost of maintenance and administration became far more important than the cost of the hardware.
The main problem I see here is that HTML isn't suitable. For HTML the server tells the client what content to display and the client decides how to present that content. This means that the server (and applications running on the server) have very little control over what the screen actually looks like (which can complicate lots of things for lots of reasons). Another problem is that the client may need all information for the page rather than just the information that's visible at the time. For example, imagine a word processor, where you open a 25 MiB file and the application loads the entire file, converts the entire file into HTML and sends about 50 MiB of data to the client; then resends that 50 MiB of data to the client each time the user presses a key.
There are protocols that are designed specifically for the purpose. Windows uses RDP (Remote Desktop Protocol), most Unix systems use X. The other popular protocol is VNC. This approach has an extra benefit: you wouldn't need to write software for the clients - people could buy standard thin clients (that come with support for the protocol) and use them, or (for desktop machines) could download a suitable client for any other OS and use it.
Of course even with these protocols it's still not perfect. Most thin clients now are reasonably powerful (e.g. 1 GHz or faster CPU with 512 MiB or more RAM) and are powerful enough to run (some or all of) applications themselves and caching frequently used files locally. If done right, the client could do a lot more and you'd end up with less latency, less network bandwidth usage, and reduced processing power needed in the server (and less hardware costs); just by using the resources in the client more efficiently. Basically, IMHO the best possible approach is "distributed computing" with diskless clients.
Cheers,
Brendan
