A couple of years ago Intel and Oracle allied with the intention of accelerating development of Internet video, the general idea as expressed at the time by Andy Grove being that information was information, so if you were going to be looking for it via the Internet you'd be looking for video as well as words. The two companies intended to produce a sort of test service as a shop window for ISPs (we weren't getting into provisioning, no sir…), but it never happened. Oracle went off to push Network Computers instead, Intel was conspicuously not one of the NC backers, and they effectively wound up in separate camps - until last week, when Intel emerged as leader of an effort to define a standard for x86-based NCs. The Intel move has been a long time coming. The company provided support for Oracle and its NC arm, Network Computer Inc (NCI) towards the beginning of this year, but hadn't said anything official prior to the announcement of the latest effort last week. Intel has been working furiously to 'prove' that Intel chips run Java best, and NCI is entirely sold on the idea of Intel being the best base for business NCs (remember the first prototypes were ARM-based), so having Intel define the standard is a logical way forward. But what, precisely, is it that Intel is defining? You'll recall that Apple, IBM, NetScape, Oracle and Sun were devising an architecturally independent NC reference profile back in 1996, but you probably won't recall anybody announcing that they had in fact completed it (due date July 96, ten on the vapour scale). Several companies including NCI do however already have operational Intel-based NCs, so we might suspect that what Intel is actually going to be doing is producing a reference design that is not processor independent in the way that was originally intended. This isn't necessarily sinister – not entirely, anyway – because in order to make high volumes and a low price point NCs will need Intel's special skills in chipset and (no doubt) board design. Significantly, Intel says the design will require an embedded Pentium or better, so you can view that as a signal that Intel finally intends to turn its work on embedded Pentiums into mass-market sales. The other thing Intel is going to be defining is also significant – the server end. Previous efforts at NC definition didn't cover this, leaving IBM, Oracle and Sun to do their own things with their own systems, but that left a rather large hole where the Intel space should be, and obviously seriously impeded NC penetration of the PC OEM business. Intel will produce a spec based on Pentium II or Pro, and will cater for specific remote execution issues such as processor load, memory, I/O and storage. This will make it simpler for OEMs from Compaq down to put out Intel-based equivalent's of NCI's network in a box. Intel has a good number of supporters on board, not as long a list as the original NC endorsers but certainly a more credible one: Microsoft, IBM, Citrix, NCI, Novell and SCO for software; Compaq, Hewlett Packard, IBM, Fujitsu, NEC, Network Computing Devices, Packard Bell-NEC, Siemens-Nixdorf, Unisys and Wyse for hardware. Note the interesting tensions in the software list. Citrix's presence puts the initiative into the Win32 mainstream, while IBM's on the software front brings WorkSpace-on-Demand into the equation. Microsoft's contribution, meanwhile, is CE. Microsoft is in an interesting position here, as although (along with Intel) it originally dismissed NCs, it seems inescapable that what Intel is working on will turn out to look pretty much like a Windows Terminal, so not supporting the latest effort would be silly. But note the key differences between the Intel NC spec and the Intel-Microsoft NetPC spec. The latter requires Microsoft software and network management systems, whereas the former will probably (it's not actually out until next year) define the hardware for management, but will leave the software side open for a range of different approaches. Recent intelligence we've heard on the first beta of Windows NT 5.0, which is the Microsoft OS intended to make the whole of the NetPC show roll, suggests that the situation is not good, probably meaning more delays, and possibly allowing some of Intel's new buddies to grab the ball back, and get an NC-based alternative established. Intel's position is meanwhile equally interesting. Its involvement means it accepts that NCs will take off to some extent, but success for remote computing of any description, NC or NetPC, will inevitably cannibalise its desktop sales, and force it into lower margin, lower spec chip sales. We think, rather than retreating to the server, Intel is hoping vast new consumer markets will open up, but nevertheless it will be a good trick if the company manages to achieve a dominant position there too. ®