Dick Pountain/14:04/29 April 1996/Idealog 21
I'm afraid the time has come to chuck my ten-pee's worth into the great debate about the Internet, the Web PC and Java (viz. is it the future of computing, life and the universe as we know it, or not?) I've been putting it off so far on the grounds that I don't know the answers, but that's rather a lame excuse in these days of Big Mouth punditry - if you don't know you're supposed to make it up. However I noticed that in the last episode of Bob Cringely's excellent 'Triumph of the Nerds' Bill Gates owned up without hesitation that he doesn't know where the Internet is leading us either, which makes me feel a bit better.
Here goes then :-
1) The desktop PC (and Mac) is facing a crisis of obesity; operating systems and applications are too fat and demand too much RAM and disk space. The piecemeal evolution of the PC has spawned a configuration nightmare which Windows 95 does little to remedy. We still waste far too much time debugging, debagging, defragging and generally helping fallen PC's back onto their feet.
2) Ergo, a simplified Web PC - perhaps even without any local storage - begins to look tempting. You'd run small software components that you download over the Internet when you need them, from some centrally-maintained repository. The objection that this means a return to the old mainframe-and-dumb-terminal model of computing is wide of the mark; once you've downloaded a Java applet its code gets executed on your local CPU (whatever that might be.) This is still the one-user-one-CPU model that has always motivated the PC business.
3) Java is a very attractive programming language. It's relatively simple, has clean object-oriented semantics and a syntax that borrows enough from C++ to be familiar to any professional programmer. This last point is vital. You could argue for years about whether there are better languages, like Oberon or Dylan, but these will never achieve mainstream acceptance because the toiling masses of programmers who have suffered to learn C++ are unwilling ever to learn anything else. Java got that right.
4) A diskless Web PC model might work for games, movies, news and statistics services (like stock prices) or corporate database access, but there are whole swathes of professional PC use where it is a complete non-starter.
This magazine is designed on Macintoshes, using Quark Xpress. Sean and Theresa and Tom and Lucy think it's bad enough waiting for Quark to load over our office Ethernet - I doubt they'd be at all happy running Quark off a Web server in Penge. But surely there's a 27k Java applet out there that will do everything Xpress does? Yeah, and I have these pills that you drop into a gallon of water and then run your car on it...
5) There isn't enough bandwidth on the Internet right now for the whole world to adopt a Web PC model, boring as it is to say it again. The world's telcos will eventually get their act together and build a broadband infrastructure, but that might just lead to more contradictions.
A future comms infrastructure is likely to be based on connection-oriented technologies like ATM and Fibre Channel working over point-to-point circuits that are fast enough to carry real-time data streams like video and voice. The World Wide Web was designed as a client/server system to make best use of the limited bandwidth we have today, and Java is a fix to bring some interactivity to this client/server architecture. In the future a more interactive peer-to-peer architecture might be more appropriate.
6) All the talk about CPU chips dedicated to running Java will either come to nought or worse, end in tears. The computing trail is littered with the bleached bones of chips designed only to run one language, whether that language was UCSD Pascal, Forth, Lisp or more recently C on the Hobbit. The economics of modern chip development don't favour niche products - you need to sell volume - and Sun will be asking Web PC designers to bet the farm on Java taking over the world. Lots of luck.
7) In any case dynamic or 'Just-In-Time' compilation techniques - where a Java applet's bytecode gets translated into native code once it discovers what brand of CPU it has landed on - ought to be able to deliver sufficient performance, assuming that Web PCs will actually be designed around fast general purpose RISC processors like StrongARM or the MIPS R5000.
8) I've yet to be convinced that Java applets can generate enough revenue even to support the reinvestment the software industry needs to continue (I don't care about keeping the car-parks full of Ferraris.) Herman Hauser recently convinced me that there are £5 billion pounds in freight costs to be saved annually through teleshopping - but that money will mostly accrue to Sainsbury and Tesco. To Java cultists all software should be free, so there is no problem.
9) Many advocates of the Web PC and Java have a barely-hidden agenda - their real interest is in bringing down Microsoft and Intel, rather than helping you and me to get our work done better. Many are Unix vendors who are still sore that their OS didn't take over the world, and see this as the last chance. Some are quasi-religious language cultists who possibly do bad things with live chickens (I'm a rehabilitated Forth programmer, so you can believe me on this.)
10) On 'Triumph of the Nerds' Larry Ellison (CEO of Oracle) came right out and said it, "I hate the PC". If I go the Web PC route, men like Ellison and the telephone company will control my access to software. All of a sudden Bill Gates looks rather cuddly.
--------------------------
PS. You can now read all the back-issues of Idealog on the Web at http://members.aol.com/dickp96/
No comments:
Post a Comment