Perspectives

lock in customers and maintain hardwaremargins.

JERCIs believe that Java is not just anoth-er programming language, but the first almostuniversal programming environment to oper-ate on telephones, television sets, networkcomputers, and existing PCs, workstations,downsizing servers and mainframes. The basisfor our vision is that network computers haveno legacy. We say they have a tremendousadvantage over the thousands of varieties ofPCs that connect to a thousand printer types,along with as many configurations formodems and networks, mass store, graphics,audio, etc. that run several hundred thousanddifferent apps. Since they are just being intro-duced, each network computer has nooptions and the hardware is thus simpler andtrivial to maintain. PCs need to be loboto-mized (have their disks removed) so as to runonly our JAVA Environments, albeit slowly.

The prayers of Brothers Gil, Larry, Lou,Mark, and Scott provided further insight forfollowers. Brothers lose a lot if followers con-tinue to accelerate their use of PCs.

iJERCIS:

SAVE THE WORLD FROM PCS

GORDON BELL

GORDON BELL

is a senior researcher in Microsoft’s Telepresence Research Group.

He is a member of netWorker’s editorial advisory board.

PERSPECTIVES JULY/AUGUST 1997n W 9

I attended a JERCI (Java Environment for Re-engineering the Computer Industry) fol-lower prayer meeting. Our mission is to savethe world from costly to maintain distributedPCs (and by inference, costly to maintain,expensive workstations), and to replace themwith our own servers and network comput-ers that host various Java environments.

Network computers will never failbecause they are connected to all-caring,always-up, never-busy, servers through analways-up and never-congested network. Theclient (a.k.a. “network computer”) never failsbecause it maintains no programs or files. Allthe client programs and files are maintainedby batteries of servers. The by-product is torender the several hundred million PCs obso-lete so they can be replaced.

JERCIs believe in market-determined,open, de facto standards. Our leaders pio-neered the open-standard concept with theVendorIX UNIX dialects, and can now do itagain with their Java environments. De factostandards allow products to simultaneouslybe open, compatible, and yet differentiated to

Gil prayed that the NC would somehowingratiate him because his was new and lowcost. It would provide an image that he knewabout the Internet, the Web, servers, andwhere he was going.

Larry prayed the longest and loudest,describing how his Network Computer wasconceived to save America’s educational sys-tem by taking an ordinary PC and removingits hard disk and floppy. Just before he hadtold how the NC was conceived to satisfy justtwo archetypal users: him (corporate user)and his mother (naive user). In the process itwill bring computers to the entire world.

Larry’s NC is connected through anEthernet to Oracle’s new operating system,NOS (a.k.a. Larry OS Technology), his UNIXdialect. Mom is connected via a POTS line toan ISP running LOST. As he went on, thedisk slowly reappeared, along with the hard-ware cost savings. But since the disk was onlya cache, holding no permanent programs ordata, it could cause no harm or present amaintenance problem. Users could use slight-ly more complex, portable NCs without anetwork, because once reconnected all thedata was fully resynched with his new server.

It was a wonderful story because aLOST/PC should be able to support 200users, while its clients searched the web, didemail, used his new word processors andspreadsheets, made videophone calls, andinteracted with multimedia objects. It alsoruns new educational software that he andMichel Milken are funding. One follower hadalready ordered 10,000. I was moved by thevision, having recalled his last vision of themassively parallel NCube computers to revo-lutionize databases and warehouses, and thena year later the video server.

Brother Lou had the clearest picturebecause he knows that his followers of fifty

years easily get lost, unless he shows them theway. His NC would serve the faithful whoneeded a vision to tell their bosses. They’veheard of a new land called Cyberspace.Unfortunately they are stuck with legacy“industry standard” platforms (AIX, AS400,OS/2, and 3090), nets (Twinex, and Tokenrings), and terminals (3270 and 5280 a.k.a.VT100). His NC also runs Java—clearlyshowing that he knows the way toCyberspace.

By building an NC, Lou’s followers aresaved from buying PCs that run all thoseapps, including the new ones for multimedia.His followers must be shielded from moderndata types, distributed computing, etc.—andthe flock continues to wander with him in awell-controlled fashion. But above all, follow-ers relate to the message. Now they again getback to centralization and have the criticalcontrol of: servers, networks, clients, and allthe user apps… just like the good old days.

Mark’s vision followed by describing whatmarket-driven open standards are what it’s allabout. He’s providing a new environmentbased on Java to be able to obsolete, recreate,and run a whole new cadre of new wordprocessors, spreadsheets, and mail programs.We can license his environment. The visionwas to follow his open standard because hehas browser market share.

And then Scott spoke and we all listened,because he has shown us “the Java way” byremoving some of the C++ uglies. Now theworld can return to centralized, server-based,timeshared computers, a way of computinghe disavowed when he started a company tosell workstations. He assured us that hisJavastation would have no disk. Users arerelieved of having to worry about their pro-grams and data because the server ownertakes care of them. Apps are only limited by

PERSPECTIVES JULY/AUGUST 1997n W10

PERMISSION TO MAKE DIGITAL/ HARD COPY OF PART OR ALL OF THIS WORK FOR PERSONAL OR CLASSROOM USE IS GRANTED WITHOUT FEE PROVIDED THAT COPIES ARE NOT

MADE OR DISTRIBUTED FOR PROFIT OR COMMERCIAL ADVANTAGE, THE COPYRIGHT NOTICE, THE TITLE OF THE PUBLICATION AND ITS DATE APPEAR, AND NOTICE IS GIVEN THAT

COPYING IS BY PERMISSION OF ACM, INC. TO COPY OTHERWISE, TO REPUBLISH, TO POST ON SERVERS, OR TO REDISTRIBUTE TO LISTS REQUIRES PRIOR SPECIFIC PERMISSION

AND/OR A FEE © ACM 1091-3556/97/0700 $3.50

Perspectives

PERSPECTIVES JULY/AUGUST 1997n W 11

the pipes and servers that feed them. We nolonger need to worry about improvements inuser interfaces or any changes that requiremodifying the platforms. We can start allover again and create a completely newworld of word processors, spreadsheets, mail,higher level apps, and everything else a clientneeds. This time we will do it right.

Scott is so convinced of theway that he eliminated the use ofPowerpoint within his company.He said new, simple office suitesin Java will soon be available toreplace the PC ones with all theirfeatures, because all users need is“cut” and “paste.” Control is akey virtue of centralized softwaredistribution. If you do it right,then the cost of ownership isdramatically less than thosepesky PCs that users run any oldsoftware. He is able to quote really big costsavings by simply ignoring the cost of main-taining his expensive, centralized servers.Since his followers are concerned with thecost of ownership, they’ll presumably shipout his workstations too, based on high appsand ownership costs. Based on the total, totalcost, NT would seem to be the best alterna-tive for serving NCs.

Scott will Javize the world by providingJava processor chips, interpreters, and com-pilers to put everywhere from pacemakers tomainframes. He told us how Java programsare safe, secure, and reliable. Mission criticalprograms can be freely shipped about andrun anywhere without fear of viruses at amaintenance cost of zero. Somehow anycombination of apps can be dynamically con-figured on his NC and it just runs right. Thisis in contrast to the cost of statically bindingapps to PCs or workstations.

Programmer followers are glad for Java,having just spent fifteen years in the desertwith ADA, C++, Forth, Telescript and othermirages.

We are on the verge of something new...and only the oblivious PC makers stand inthe way. I doubt if they get it. PCs look justlike the same old Boca box into which usersstuffed boards to make a computer that com-municates with a LAN, public net, disk, dis-play and audio. One can understand whyLarry and Scott claim their mothers cannot

use PCs. We JERCIs need to per-severe and create a brand new PCenvironment that has no vestigeof the old, and in this way cantruly re-engineer the future.

JERCIs armed with networkcomputers offer to solve comput-ing’s ills as long as there aren’ttoo many types and they remainstateless... like 3270s, VT100s,Xterminals, and diskless worksta-tions. They reduce the cost ofownership of distributed comput-

ers (e.g. PCs and workstations) becauseservers can maintain things dynamically.Servers are not inexpensive to maintain andmanage either, but there are fewer of them.

As network computers get closer to reality,they seem to acquire more state, and in timewill probably look a lot like PCs and espe-cially workstations because they’re all likelyto be different. The big gain is that we againget the benefit of centrally controlled comput-ing a.k.a. mainframes that we lost severaltimes with minis, workstations, PCs, andVendorIX servers.

By the way: I hope you get the impression thatI’m 100 percent behind NCs for users whouse a few apps like mail. As an NT user andsystem manager with NCD remote window-ing terminals, I applaud the idea and think it’sthe best network computer server. I just wantto make sure when we finally get these newcomputers that have the benefit of competition— that I can use those made by Scott, Mark,Gil, Lou, Larry or Moe or whoever, inter-changeably. This just might be the final straw

We are on the

verge of

something new...

and only the

oblivious PC

makers stand in

the way.

to reduce our dependence on those expensiveworkstations. And by the way, we also get ridof the software industry because all thosedownloaded Java programs can be read byanyone.

What do I really think about the NetworkComputer?

Network computers (a.k.a. diskless work-stations, a.k.a. Xterminals using time-sharedservers) are flawed in three ways as they aredeveloping because they are likely to be:closed and proprietary; limited by networksand lack of memory that have so far shownthem to be ineffective; and they trade-off theubiquity, freedom, flexibility, and multimediaaspects of PCs. Furthermore, cost compar-isons are specious because they fail to takeinto account the myriad of costs including theneed to re-engineer the various UNIX dialects

a.k.a. VendorIX to cope with the demands ofa real-time, timesharing system and to main-tain complex remote apps on a dynamic basis.

Unique client platforms will take us backto proprietary VendorIX and mainframe serv-er platforms and the resulting hardware-soft-ware-app lock-ins. The network computer(a.k.a. diskless workstation of ten years ago)was a clear failure because apps usuallyexceeded local memory requirements requiringmemory paging via LANs. While networkcomputers (a.k.a. Xterminals) may work justfine for some apps we know today such asmail, web browsing, and 3270 and other ter-minal emulation, they are likely to be unsuit-able for the emerging apps such as videotelephony.

Let’s face it, network computers have madetheir biggest gain: they have forced the PCindustry to focus on total cost of ownership.

PERSPECTIVES JULY/AUGUST 1997n W12

~