It turns out that the Notorious K.R.I.D. had a work-supplied, Pentium-4-based Dell system with a noisy, rattling heat-sink fan that was slowly driving him nuts.

Blame for this rests solely and squarely with Dell: their standard practice since time immemorial has been to equip their machines with a large, semi-passive CPU heatsink, place an exhaust fan on the back of the machine, and connect the two with a plastic shroud. In theory, this is a better approach than the standard one involving a fan blowing air onto the heat sink: Dell’s approach draws air over the heat sink and then ejects it from the machine, rather than running the risk of simply recycling the same ever-hotter air. It would be a better approach in practice, too, if Dell would only find some way of turning the shroud into something other than a bullhorn to amplify the fan’s roar.

So we set about removing the Dell heat sink, and replacing it with the infamous PIPE101. This process was surprisingly complicated, and involved a number of unexpected discoveries.

1. The thermal tape which Dell uses as an interface between the CPU and heatsink is surprisingly sticky. Sticky enough, at least, to pull the CPU right out of the locked ZIF socket. I didn’t know that was possible. Miraculously, none of the pins were bent.
2. Dell, in keeping with its tradition of reinventing perfectly good wheels whenever possible, used a non-standard retaining bracket for the heatsink. Bastards.
3. Mercifully, they didn’t have the inclination or the opportunity to deviate from the standard spacing for the heatsink-bracket retention holes. Consequently there was hope that if we could track down a spare Socket 478 heatsink-retention bracket, we’d be in business. Thanks to the ever-helpful folks at BlueBonePC, we obtained one.
4. This was when we discovered that while the spacing on the holes is standard, the placement of certain capacitors near the CPU is not; a standard heatsink-retention bracket will not fit without modification. Dirk promptly supplied the neecessary modification by clipping out part of our bracket, in a way that didn’t seem to compromise its function. (Bastards!)
5. The rest of the installation went relatively smoothly. The PIPE101 fit into the available space with room to spare, and Thermaltake’s approach to Socket 478 retention proved far less squirrelly than their take on Socket 939. Of course, it was only after I’d installed heatsink and fan that I realized I’d forgotten the backing plate on the other side of the motherboard, but never mind. I was even able to remove the heatsink without simultaneously extracting the CPU this time.

We then attached an exhaust fan to the back of the case through means so crude you really don’t want to know about them — I have only one word to say on the matter: “wingnuts” — reconnected the machine, and fired it up. Miraculously, it booted despite the astounding amounts of abuse and vendor-unapproved handling to which we had subjected it.

At last report, it is running strong, but silent. Dirk is happy, and so am I, to have found a suitable home for the heatsink at last.

(I’m also happy in no small measure because Dirk reciprocated by expertly tuning up my bike, from out-of-true wheel to balky front derailleur to slow leak in front tire. He also replaced my despised toe-clip pedals with a pair of Speedplay Frogs which he sold to me at friend prices, and with which I have rapidly fallen in love. Better than just do all these things for me and present a fait accompli at the end, he let me observe every step, so that I have half a chance of doing for myself in future.)

The Netgear WG602 access point is a much happier piece of hardware than its accursed D-Link counterpart. It doesn’t want to reboot whenever you do… well, anything, its DHCP client implementation actually appears to work, and its built-in web-based UI is a model of elegance and simplicity.

It may, in fact, constitute one of the very few legitimate uses of HTML frames I’ve seen. Ever. The left frame chooses the configuration topic, the center frame allows you to set options, and the right frame displays help appropriate to the current context, so you don’t even have to make any additional clicks to find out just what the options you’re facing off against actually mean. I am pleased and impressed.

(Of course, I haven’t actually tried attaching any wireless clients through the thing yet, but trifles like that can wait for the time being. Also, for maximum perversity, I have acquired a Linksys high-gain antenna and antenna mount, just to add vendor mix-and-match to the equation.)

I have long harbored the suspicion that WiFi is a kind of mass hallucination rooted in collective embarassment. Fundamental security issues — which are grave, and many — aside, no one wants to be the first to admit that this stuff often doesn’t work worth a damn, and doesn’t come close to delivering the advertised throughput even when it does connect with some approximation of reliability. The fear of being mocked for technical incompetence by one’s companions in suffering is simply too great.

Recent experiences have done more to reinforce my opinion than to change it. Thinking that it was time to move at least partway up the technology curve, I decided to acquire an 802.11g-capable access point. Having been reasonably pleased with D-Link’s DGL-4100 router, I picked up the company’s DWL-2100AP access point, along with a matching omnidirectional antenna.

To say that I am not impressed would be a galloping understatement. I didn’t necessarily expect the product to end world hunger, but for the price, I expected it to at least be fundamentally usable. No dice.

For starters, every single configuration change requires that you reboot the device. Change the SSID? Reboot. Alter encryption settings? Reboot. Switch from a static to a dynamically-assigned IP address? Reboot. Every reboot requires that you and your browser twiddle your thumbs for 20 seconds while the AP regains its ability to act as a web server.

It wouldn’t be so bad if you could batch up the changes and effect them all at once, but you can’t. Maybe you’re supposed to use the included SNMP-based access-point management software if you want advanced functionality like that, but then, really, what was the point of even bothering with a web-based configuration interface in the first place? As it is, the thing makes the Windows update and installation proceess look like the very model of simplicity and elegance, and that’s saying something.

In the end, I could live with that, since I expect to configure it once and then forget about it. I could also live with the prevalence of engrish in the UI, for much the same reason. Remember my casual allusion to dynamically-assigned IP addresses above? I should clarify: you can request that the access point act as a DHCP client, provided you don’t mind its subsequently not working at all. Out of the box, the device presumes its IP address to be 192.168.0.50; switch it to DHCP mode, and it will stop listening on that address — or any other.

Watch your DHCP host’s logs, and you will see the access point requesting and receiving a lease. And requesting and receiving a lease. And requesting and receiving a lease… and so on, ad infinitum. Even though it’s assigned the same IP address every time, it’s completely unresponsive when contacted on that address. Installing the latest-available firmware (v. 2.0, at the time of this writing) does nothing to solve the problem.

So I’m taking it back. If it were a matter of just one flaw with an otherwise-impeccable design, I’d try contacting D-Link technical support and working through the issue. But this is clearly a product that is substandard in a number of significant ways, and was shipped anyway. Those responsible should hang their heads in shame. For $100, I expect and demand better. We’ll see if Netgear can do a better job of winning and keeping my loyalty.