Commentary: It's time for the personal server

Today we have a broad spectrum of personal computing devices. But as empowering as it is to have such computing resources at your fingertips, there's still a missing piece: personal servers to delegate tasks to.

Anyone reading this column is unlikely to require convincing about the general usefulness the personal computer. Today we have a broad spectrum of personal computing devices, from 68k-based PalmOS-based organizers all the way to multiprocessor Alpha boxes on individual desktops. But as empowering as it is to have such computing resources at your fingertips, there's still a missing piece: personal servers to delegate tasks to.

There's plenty of time-consuming stuff that I'd prefer not to tie up my main work machine with, other tasks I'd love to launch from an organizer and have run independently on more capable hardware, and yet other work that has no business running on a personal machine in the first place.

Our industry has even evolved to the point where most personal computer operating systems have all the features we need to build good servers with them. And with the rising popularity of Linux and the xBSDs, not to mention Windows XP's imminent release and Apple's brand-new OS X, the forthcoming generation of putatively consumer-oriented OSes have some pretty serious infrastructure under the hood.

There should be at least two
While it would be nifty to have a grand unified server that handles both network services and media stuff, I think having at least two different servers isn't unreasonable. Serendipitously enough, Rob Malda over at Slashdot posted a treatise on his idea of a converged media server just as I was writing this column. Evidently I'm not the only one thinking about this issue genre.

In fact, it's not just two of us, either. With the prominent peripheral ports (plural!) on Sony's PlayStation 2, it seems pretty clear that forward-thinking consumer electronics companies would love nothing more than to turn our home entertainment systems into collections of internetworked devices, with something like the PS2 serving as either a powerful node or even the media router itself. Now if only we could convince the MPAA and RIAA that supporting this sort of technology rather than trying to hobble it at every turn would increase their members' profits...

For my own purposes, I envision two clearly delineated servers: one for media devices and one for computers. The media box would be an audio & video recorder, tuner, and archive/jukebox. It could talk with other media peripherals, eg a DVD player, via FireWire, but also stream media to playback devices throughout the house via Ethernet.

The other box I have in mind is a combo router, firewall, 802.11 access point, and miscellaneous network services server. With the exception of the access point functionality (when will 802.11 interface cards for access points no longer require special and extra-expensive firmware?), this box is readily buildable once I find the right hardware. I have discovered that the form factors for in-home servers just aren't there yet.

I don't want a PC minitower (too big, too loud) and 1U rackmount servers are still priced way too high for non-corporate use. The closest thing I can find is something like a BookPC, but it's a very closed system, and, if its immediate predecessor (the BKi810) is any indication, it's also quite loud.

Remember the LC?
Which brings us to Apple. The company has invited the press to an event next Tuesday and the rumor mill hints that Apple will disclose its new product strategy for the education market. This reminds me of a previously successful machine for Apple in that very market, the Macintosh LC.

The LC (and its successor, the LC II) was well less than 1U tall and had everything except a display and input devices integrated. It had a small expansion slot for Ethernet or a modem, and was remarkably stackable, a feature I discovered while building a network testing lab.

What I'd like to see from Apple -- or any smart hardware vendors -- is an updated LC. Motorola is apparently producing enough 7450 G4 chips at the high end these days, so one would hope that they've also got the 7410 G4 yields figured out enough to use that chip in a low cost machine. Having an AltiVec-capable machine would provide a very nice baseline for media software developers.

Keep the box thin, literally. Make sure it fits in a 1U shelf if it needs to, but make it still look good if all you do is put it on, or next to, a desk. Put video, FireWire, USB, 10/100 Ethernet, and 802.11 on the motherboard, and -- this is important, so pay attention hardware designers! -- provide a single, full-size, unpopulated PCI slot for expansion. One hard drive bay with a UDMA-100 interface would suffice, though having two bays would be extra useful.

Make it quiet and as low-power as possible. No full-time fan, and laptop-grade power saving features. Tell your hard drive vendors that you care about power consumption and they will quite happily add the same power saving features to desktop drives that they already have in their laptop drive controllers.

Since Apple has the wherewithal to integrate clever touches like auto-pair-switching Ethernet on laptops, integrate the functionality of a video dongle so that if the machine boots without a monitor attached, it auto-sets itself to a predefined resolution so that Timbuktu et al work properly, and a monitor won't get confused if it's hooked up later.

This would be a killer machine for the education market, as a general purpose desktop machine, and would also make a darn fine light- to middleweight server. Swap the G4 for one of the many x86 options and it would make a spiffy non-G4 box as well.

Specialization is for insects
A basic problem with personal computer hardware designers today is that they are specializing their product lines too rapidly without taking advantage of designs that can serve multiple markets well. Today we have standard desktop designs, standard consumer designs, and standard server designs.

It's time now for evolved designs that provide a good general feature set and form-factor that can be used in applications ranging from consumer desktop to inexpensive-yet-plenty-powerful server. Let the users decide how best to use the hardware; don't force everyone into a scant few hardware categories.

ZDNet columnist Stephan Somogyi wonders if IBM will resurrect Lotus Improv and get it running on OS X.