cobolhacker.com

Logitech would now appear to be selling real laser mice. No mention of it on their website just yet, but I’m able to pre-order the MX1000 mice from distributors. They should retail for around $95.

So why is this cool? Regular optical mice, which are often mistakenly called laser mice, don’t actually use a laser, they use a tiny focused light. This works well, but it has limitations, particularly on shiny surfaces like lacquered desktops or plastic mouse pads. Real laser mice should go a long way to fixing these limits because the resolution of the laser is so high it can see, and therefore compute movement from, the smoothest of surfaces.

Unfortunately, I don’t think laser mice aren’t going to have that friendly glow like regular optical mice do, because of the unidirectional nature of the laser.

Every computer person knows what an IRQ is. The dumb things are a legacy of the Intel Architecture, the underlying multitasking technology of all Intel-compatible personal computers. It has been a thorn in the side of computer techs for at least the last fifteen years.

I remember (not so fondly), back in the DOS/Windows 3.1 era, trying to configure various PnP devices to “play nice” with each other and not share the same IRQs, as this would generally lock up the system (for a lark, you could have Windows 95 try to configure PnP devices for you…) Scanner cards and soundcards were particularly notorious for not playing well together. You see there are only 15 IRQs, and only 7 are generally programmable by the users (numbers 3, 4, 5, 7, 9, 10 and 11; also 12 if you don’t use a PS/2 mouse). Since the hardware couldn’t be trusted to configure itself, one would have to properly program the BIOS to ensure that the pushy ISA cards would get the correct IRQs assigned.

When the PCI bus came about the miracle of IRQ routing allowed PCI devices to share IRQs and self configure. This was made more complicated by the fact that some PCI slots always share the same IRQ (usually the first one and the last one) and many onboard devices like LAN controllers and sound cards inherently share an IRQ with one of the PCI sockets. But old OSes like Windows 95 could not do PCI routing properly either so once again you would have to hack your BIOS, turn off things, force IRQs to certain devices and certain slots or physically shuffle the cards in the PCI sockets to ensure that there was no sharing of IRQs.

Windows 98 Second Edition greatly lessened the problems and modern OSes like Windows 2000 and Windows XP have no problems at all with PCI interrupt sharing and the like. They simply create virtual IRQs, up to 255 of them, and interoperability is ensured.

Well, so you would think.

The other day I get a computer in the shop with a professional grade sound controller that’s having playback problems. The Echo-Layla is a popular system used by music artists to record multiple audio channels and digitize them on to the PC, where programs like Sonar can be used to stitch them together into songs. A lot of modern music is now made in this way.

The sound system’s playback of recorded multitrack stuff was slightly broken with clicks and pops and other unwanted junk, as if the machine is having trouble keeping up with the data output. This is not likely as 16 channels of PCM wouldn’t really give a Pentium 3 much of a workout let alone an Athlon XP 2500.

Suggestions on the various support websites suggest that this is a common problem with the nForce 2 chipset, like the one found on the customer’s Asus A7N8X-DX mainboard. It’s owner, figuring it was a PCI bus problem (common with older sound systems) had already tried unsucessfully tuning his PCI latency timers with the tweaking tool Powerstrip. He fears that the Serial-ATA controller his hard drives are hooked up to is monopolizing the system resources somehow.

We spend about an hour pissing with the PCI latency timers and following various other suggestions on the websites to no avail. I notice on one of the many reboots that 1) there are a crapload of PCI devices, as with any modern system, and 2) the SATA controller and the sound controller were sharing an interrupt.

On a modern system Interrupt sharing is generally not a problem. In fact, the sheer number of user devices kind of makes it necessary: serial port, USB root hub, USB root hub, yet another root hub, firewire contoller, video controller, SATA controller, modem, LAN controller, sound controller, etc, etc… So some sharing of those seven IRQs definately has to happen. But maybe the fact that the sound contoller and the SATA controller are shaing an IRQ is causing the problem, like in the bad old days. I turn off a bunch of stuff, and physically move the Echo card to a different PCI socket so it wasn’t sharing with anything. And low and behold — the clicks and pops went away.

Anyway, fixing this unusual problem was a serious flashback to 1995 when I was doing the same things to keep Windows 95 boxes from crashing. Kind of curious that after all these years adapter cards still can’t place nice with each other in the playground of Intel Architecture.

After long delays and whining Microsoft has released Service Pack 2 for Windows XP (at least to us OEMs, anyway). More than just a collection of updates, this service pack aims to fix some of the more serious security problems with the product. You can read more about what it does if you feel the urge.

Some initial thoughts:
SP2 adds something called the Security Center to help make the task of managing security easier. It includes three prominent status sections similar to the status windows of Norton Internet Security: one for the firewall, one for automatic updates, and one for your virus scanner. The status sections tell you what’s working and what isn’t. On a default install the Virus Scanner section will be off and red. More on that later. At the bottom, it also provides shortcuts to the Settings applets for Internet Security, the Firewall, and the Automatic Updates so you can easily change the behavior of these things.

After you install SP2 the first thing it does following the obligatory reboot is ask you to turn on the automatic updates. This is not a bad thing to do. By default, it uses BITS to quietly download the updates when you are online and installs them at 3:00am or on the next reboot, whichever comes first. Naturally, serious gamers will want to change the default install time, which could seriously slow down late night gaming, to a more reasonable time of maybe 10:00am.

The improved firewall is easier to configure, but on the surface doesn’t appear all that much different from the last one, other than it is enabled by default. It has a section for exceptions and looking at this section reveals that remote assistance is exempted from the firewall by default. Since this seems like an obvious avenue for attack, I would recommend unchecking the box so RA requests will be blocked. The firewall is supposed to tell you when it blocks something, but I haven’t yet seen the behavior. ICMP (ping) is also blocked by default. I used nmap on an XP SP2 box and it didn’t reveal too much.

The shortcut to Internet Security Options is the same as the Security Tab in the Internet Options section of the Control Panel. A quick breeze through the Custom settings for the Internet Zone shows that some stuff has been added, most notably a setting for a Pop Up Blocker, which is enabled by default. I guess the Microsoft guys finally got sick of popups too.

The virus scanner part seems broken right now, as it requires the virus scanner to support the reporting feature. Even a current copy of NAV 2004 did not make the Security Center believe that there was actually a virus scanner running. I suspect that this problem is going to be short lived, not just because major anti-virus vendors will want to make their products work with it, but because Microsoft is going to start selling Antivirus software any day now. And you just know that their version is going to work flawlessly.

There is also a long list of long overdue changes to Internet Explorer aimed to protect the user from nasty scripts and websites by changing how windows runs webpage scripts automatically. Unfortunately none of these seems to address the core problem that a regular Internet user doesn’t know well enough to not to run bad scripts when the browser asks. The default behavior is still to ask, meaning that users are still going to click yes so they can keep getting their cartoons, smileys, porno or whatever else they are after. Also, scripts continue to have too much control over of the functionality of the browser. Really, instead of using their current, complicated Zones and Digital Signatures scheme, IE should adopt the Java principle of the sandbox, and not allow anything to run outside of the current instance of the browser. And the browser should have a “Reset Me” feature that gets rid of every non-standard add-on and setting. Too many of the changes to the browser’s security seem to be comprimises aimed at accommodating the bad habits of web developers.

To make a long story short, I’m a bit underwhelmed by Service Pack 2, what with all the hype it’s got lately. The improved browser security stuff is nice, but seems like a hack for an idea that’s already way too complicated. A lot of the fixes also seem like Johnny-come-lately stuff. For example, reducing the RPC attack vectors is nice, but this vulnerability has already been exploited to death — the virus writers are now on the next big vulnerability that no one knows about.

I install the thing (this takes like an hour cause the machine is a P200) and it locks on boot. Even before it unpacks the kernel. ARRRRRGHHH!!! Since FC2 likes all my other boxes I can only conclude that the hardware in the box is either too old or too flaky. I had some suspicions before and this was part of the reason for the re-install in first place. So now I’ll have to build another firewall. In the meantime I’ll continue to be stuck with my trusty old D-Link DI-704 SOHO firewall. Reliable product sure, but unfortunately the device only supports NAT. This means that half the stuff inside my network will continue to be broken. I hate NAT. And I hate having to fix my own hardware.

I just noticed that Intel is sneaking a hybrid 32/64 bit processor on to the market. The Nancona is a 64-bit capable Prescott-based CPU that will be instruction-set compatible with AMD’s much-celebrated Athlon 64. Not a word about it on the Intel website yet.

I think the sneakiness is going down because they are a little embarrassed over their earlier attempt at this market which has been something of a flop. The fully-64 bit Itanium, which shipped a year and a half before the Athlon 64 did, is a top notch server processor for a perfect world – one where you are able to re-compile your operating system and all your software for it. However in the real world, where proprietary software is king, large companies who might benefit from 64-bit computing didn’t really feel like giving their software vendors millions of dollars to recompile their existing 32-bit software. So it hasn’t exactly sold like hotcakes. A sad and ironic case of market forces stifling a geniune attempt by the market leader to innovate.

And to add insult to injury, last year AMD came out of nowhere saying: “Well, you can have your cake and eat it too. We’ll give you 32-bit with 64-bit extensions.” At bit cynical perhaps, because large companies probably won’t ever migrate to 64-bit with the AMD systems they buy today for the same reason they won’t buy Itanium, but at least the illusion of choice is still there. And AMD got to make Intel look stupid and I think AMD needed the morale boost anyway. An equally sad and ironic case of market forces encouraging the upstart competitor to play it safe.

But alas, Intel lovers: don’t think you’ll be getting one any time soon. They only seem to be shipping to large server makers such as IBM and Dell.

Besides, an Athlon 64 will cost you half as much.

I saw a snippet of this hilarious flash movie on the morning news of all places. As always with these tales of Americana, the creators over at Jib Jab are being sued over this, not because they are making fun of Bush or Kerry (this is perfectly legitimate, as they are public figures), but because whoever owns the rights to the late Woody Guthrie’s music is pissed off.

So, for your convenience, and to protect the link in the event of legal crappiness, I have carefully preserved the movie in it’s original state. Enjoy.

Microsoft says that the minimum requirements for Windows XP are a 300Mhz processor and 128MB of RAM. Sounds a bit suspect to me. 128MB seems a somewhat low for a modern operating system. I’ve seen XP running on a machine with 128MB and it’s fairly painful and mostly unusable beyond loading the desktop. Modern operating systems have a lot of stuff to load into memory so really that number should be closer to 256MB. The 300Mhz part also seems odd to me — it’s too high. I have trouble believing that any modern OS actually requires a 300MHz processor. Maybe to do stuff like play games or decode DVDs, but certainly not to simply load and manage a GUI. My old jukebox computer was a 166 Pentium and I was able to run Red Hat 8 with a fancy Gnome desktop just fine (it had 192MB of RAM).

So just for chuckles, I put together a Pentium 200MMX and installed Windows XP Pro on it. To be fair this was a fairly beefy P200: 512MB of SDRAM on a VIA MVP3Pro mainboard, an MSI GeForce4MX AGP graphics card, and a 7200rpm hard drive. Threw in an SB live for sound too. All of this in an old school AT case. I would have killed for a system like this back in 1997.

And XP worked and worked well. Boot time averaged around a minute. It was maybe a bit slow pulling up certain features — for some reason the Start Menu took a few seconds to load — but it was certainly usable. It’s difficult to gauge the actual load speed of Internet Explorer as much of it is loaded into memory at startup but the initial launch of the web browser took about 15 seconds. Slim software like Abiword loaded up in under 10 seconds. But the system certainly took it’s time loading up bloatware like Media Player. And it while it was able to play an MP3 (no visualizations!) just couldn’t be expected to play a DVD in real-time, no matter how good a DVDROM drive I put in the unit.

So what’s the moral of the blog entry: with a modern operating system it’s all about the RAM, not the processor. An “underpowered” Pentium 200 can capably run the functional aspects of XP by making sure it’s real needs were meet: lots or RAM and a fast hard drive.