Bogonomicon

Dean's performance in Iowa has caused a great deal of soul searching out in the Blogsphere- if Dean doesn't get the nomination, will you support who does? My answer is clear- NO.

Here's the problem I have with being ABB. My entire life has been one long game of one step forward, two steps back. And, post-Clinton, the Democratic strategy has lead to two steps backwards, one step sideways. All pretension of making forward progress has been dropped- Democratic language today is one of maybe hoperfully keeping some of what we have now. Maybe they won't overturn Roe v. Wade. Maybe they won't entirely repeal the first amendment. Maybe they won't bankrupt social security. And I'm supposed to be thankfull because in 2004 we might take a step sideways, instead of another two steps back? If we're lucky?

Bush is bad. That I agree whole heartedly with. But if we don't change something fundamental in the Democratic party- if we don't heed the wakeup calls of 2000 and 2002, we'll be right back here, with a new Bush, come 2008, or 2012. Nixon should have destroyed the Republican party for a decade at least- but 6 years later Reagan and Bush Sr. came right back into power. The ABB crowd isn't looking at the long term. The worst defeat possible for Bush wouldn't stop the Republicans from shedding their image like a snake shedding it's skin, and come right back into power in 2008. We need to start taking 2-3 steps forward, and that means radical change. We need to aim for more than status quo.

Unlike the majority of Americans, I was paying attention to this race 4-6 months ago. When the conventional wisdom was that the only way to win was to run as Republican-lite, which, when you inspected it, turned out to be exactly like being a Republican, except they felt guilty about it. Edwards, Kerry, Lieberman- they all voted for the war resolution and the PATRIOT act. Now they adopt liberal/progressive attitudes, like sheep donning wolf's clothing. My prediction: should one of these people actually win the nomination, they will immediately tack hard to the starboard, talking about wooing the soccer mom and NASCAR dad demographics, and ditch their liberal disguises. Clark will likely do the same- his advisors are almost universally DLCers, who will strongly advise a hard shift to the right. They all promise only one step sideways- not the radical leap forward we need to stave off long-term doom.

The New Deal coalition- and the Democrats- was a supermajority and thus dominated politics between 1932 and 1968, a run of 36 years. Take a long hard look at the Republicans of the day, especially Eisenhower- they'd be too liberal for the Democrats these days (Eisenhower tried to cut defense spending! Gasp!) Johnson, signing the Civil Rights act into law, killed it by causing the white southern protestant southern demographic (and a non-trivial amount of the immigrant catholic northern demographic) to pull out. Identity politics started a 36-year slide that culminated (to date) in the 2002 debacle.

But I think the fundamental political bent of this country is still New Deal/Great Society progressivism. Notice how even Bush has to cloak his proposals in Orwellian doublespeak titles to make them sound liberal/progressive- the Clean Skies act, the No Child Left Behind act, etc. If you wandered around and asked people if they beleived all children should be educated, that jobs should be safe and pay real wages, that everyone should have health care, that the environment should be protected, etc- people will say yes. And if we could just get people in the south to vote their economic interests and not their prejudices, we could put the New Deal supermajority back together. And go back to two steps forward, one step sideways.

But maybe I'm wrong. Maybe the country really is becoming deeply conservative. In which case, the Democrats are doomed anyways. The question then becomes not if we can defeat Bush this time, it becomes when will Bush (or his ideological brother) will be back. Wether we make a doomed stand now, or surrender peicemeal, won't change the inevitable outcome- goverment by the cronyists, of the cronyists, and for the cronyists. And most probably the effective death of democracy and freedom in the US. It's only when we can stave off the impending doom that making a stand now becomes important. If not now, when do we make a stand? In not here, where? We've been retreating for thirty years- when do we stand and fight?

I choose here. I choose now. Death before dishonor, damn the torpedoes, I'm for Dean.

Any article with that title gets my attention. Of course, I think the x86 architecture should have gotten a bullet in the brain no later than 1982, but that's me.

The problem is that circa 1982 was about the last time it was really feasible to kill the x86. The one lesson learned from the Itanic debacle is that people do not want to upgrade hardware and software simultaneously. The only times I know of that people have done that is when their old hardware/software architecture has died out from under them, and there is no choice but to do a simultaneous upgrade and vendor switch. The exception to this rule- Apple- proves the rule. When Apple switched to PowerPC from 68K, the new PowerPCs were fast enough to emulate the old 68K faster than a previous-generation 68K ran native. The first generations of Apple Power Macs mainly ran 68K code. The next generation of hardware has to run the previous generation's software at least as fast as the previous generation's hardware, with a promise of more performance in the future- that is the ironclad rule. And it's been true for at least 20 years.

The early-mid 1990's was an interesting time for Intel- and, I think, a missed opportunity to kill the x86. AMD was at least a generation behind Intel- the K6, Athlon, and Opteron were still in the future. Apple was proving that virtual machine emulation had arrived, and at least Sun/Java was listening. Right around the time Intel was going for the Pentium, or maybe Pentium Pro, they could have instead gone for a serious RISC chip, gotten an easy 50%+ speed increase, and spent most of it on emulation/JIT compilation of the x86 code into something sane. Unlike Apple, Intel would have had to help port 2-3 OSs- Windows, OS/2, and maybe NeXTStep, depending upon the exact time and wether Intel cared. In retrospect, only Windows really mattered. This is what Intel has actually final decided to do with the Itanic- unfortunately, I think it's too little, too late.

Unfortunately, the window closed, sometime back in the nineties. AMD released the K6, then the Athlon, and now the Opteron. Intel no longer has the generation or so of lead on it's competition- one could argue (and I will, elsewhere) that AMD has a generational advantage over Intel. Nor is there the easy big wins anymore. The advantage RISC has is it does the same work in fewer transistors- except that in modern CPUs the actual CPU tends to be just shoved in a corner, the vast majority of the chip is cache. And cache is the same between x86 and RISC.

Intel's hope was that VLIW would provide enough of an advantage over RISC that they would still be able to pull and Apple- get enough of a speed increase out of a new architecture to pay for the emulation of the old one. Unfortunately for Intel it didn't work. Wether it would have worked if Itanium had been released in 1998 like originally planed I'm not sure. But the performance advantage of the Itanium over the x86 and RISC competitors today is small enough that emulation isn't really a feasible possibility- a 50% speed advantage is minimum to pull that trick. And the Itanium's speed advantage disappears, and becomes a huge speed disadvantage, when using anything except the most advanced compilers. While the x86s and RISCs have a lot of ability to make up for suboptimal codes, the Itanium has none. Watch that first step, it's a doozy. Unfortunately, most code is compiled with not very advanced compilers.

And even code that is compiled with an advanced compiler gets run on the wrong CPU. The Itanium requires the compiler to "bundle" instructions that can be executed in parallel. Imagine a set of 12 instructions which can be broken into either 2 groups of six instructions, or 3 groups of 4 instructions. And that there are two different Itanium CPUs, one with four instruction pipelines and one with six. If you compile the 12 isntructions as 2 groups of six instructions, it executes in 2 clocks on the 6-pipeline CPU, but it takes 4 clock cycles on the 4 pipeline CPU (the CPU can execute 4 instructions from the first bundle in the first clock, then the remaining 2 instructions from that bundle in the second clock- remember it can't mix isntructions from different bundles!- and the same break up happens with the second bundle), a 33.3% speed reduction from breaking the instructions into 3 bundles of 4 instructions each. But breaking the instructions into three bundles takes 3 instructions, even on the 6-pipeline machine (the extra two pipelines are never used), a 50% speed reduction from it's optimal. Optimizing for one machine deoptimized for the other, often signifigantly. So we will always be dealing with suboptimal code.

Notice one thing- AMD's Opteron CPUs run the "old" 32-bit code at full speed, signifigantly faster than the previous generation of Athlons. And AMD made it easy for 64-bit OSs to run 32-applications (in a manner very similiar to how the 386 and successors could run 16-bit apps inside a 32-bit OS). No emulators need be included in the OS. This means that you can justify buying Opteron computers right now and using them to run your old 32-bit apps on your old 32-bit OSs. Then, in 6-18 months when you're upgrading your software, you'll be able to upgrade to a 64-bit OS, and maybe those apps which would really benefit from 64-bits. Other apps can happily stay at 32-bits. Or not- the Opteron is perfectly justifiable as a 32-bit only CPU. This makes users much more willing to upgrade- they don't have to do everything right now, and they aren't locked into any future plans.

Unfortunately, this means that we aren't going to be able to kill the x86 anytime soon. 64-bits will be enough room that we won't need to go to 128-bit machines for 15-25 years, assuming Moore's Law continues for that time frame. Unfortunately, that's the point where Moore's Law as we currently understand it starts hitting real hard physics limitations- another 30 years of doubling clock rate every 18 months means computers will be running at ~10¹⁵ cycles/second. At that frequency, the quantum energy of a photon is enough to break up atoms and change their atomic structures. The next shot to kill the x86 is going to be right around the time Moore's Law runs out.

I think hitting the end of Moore's Law will bring RISC back into play- as we start being able to super-optimize and tune our chips without constantly have the parameters of how many transistors we have and what process we're designing for changing out from under us all the time. But the end of Moore's Law will have profound and far-reaching consequences that I will dig into predicting some other time.

The latest missives from SCO has that Linux is evil because it could be used by the Axis of Evil.

I'd add a sarcastic comment here, but none seems to be sufficiently saracastic for the situation. What do you do when legitimate news is less believable than parody?

Today's required reading is this post over at Orcinus, about the differences between theories about real conspiracies (which do exist- as Nixon, COINTELPRO, and Iran-Contra proved) and conspiracy theories. Especially this part:

Real conspiracies, by their very nature, have the following characteristics:

-- They are limited in scope, their purpose being usually to achieve only a singular, often narrow, purpose.

-- They are limited in duration in time.

-- They include only a limited number of participants.

-- As the boundaries of these limits increase, the likelihood of the conspiracy failing or being exposed rises exponentially.


Conspiracy theories, in direct contrast, almost universally feature the following qualities:

-- They are broad-ranging in nature, and usually boil down to a massive plot to enslave, murder or politically oppress all of mankind or at least large numbers of people.

-- They are believed to have existed for long periods of time, in some cases for hundreds of years.

-- They involve large numbers of people, notably significant numbers of participants in high positions in government or the bureaucracy.

-- The long-term success of these conspiracies is always credited to willing dupes in the media and elsewhere.

There is a natural human attraction to conspiracy theories, I think- they're the political equivelent of ghost stories told around camp fires. In that sense, I enjoy them as next as the next guy, and endulge in them on occassion. The problem comes in with the fact that most people have never been taught logical thinking or the concepts of a proof, and don't want to deal with shades of uncertainity between TRUE and FALSE. And that the truth of a proposition should be readily decidable, if not pre-decided. Which is a problem for me, as I like playing in the maybe-world of what-if. What is interesting is what is not yet known. Yesterday's weather doesn't interest anyone (well, except maybe climatologists), it's tomorrow's weather which is interesting.

Gwynne Dyer, who did the seminal (and largely forgotten) series War thinks that Al Qeida may help get Bush re-elected. It makes sense- Bush is the best recruiter for Al Qeida ever.

This opens up the whole discussion of wether Bush will attempt his own October Surprise. Another terrorist attack on US soil is October Surprise theory #1 (with a bullet). Actually, I wouldn't be surprised that the black-ops side of the CIA didn't have a few cells of muslem terrorists kicking around who think they're part of Al Qeida but really report to Langley. After all, the CIA trained Osama in how to be a terrorist- it's not like we don't have contacts. The only problem now is that the CIA hates the Bush administration for a) blaming them for the bogus intelligence on Iraq (unfairly- I'm not a fan of the CIA, but in this case they are innocent), and b) outing an agent (Plame). Which doesn't make it impossible (you're telling me there isn't a single christian conservative working out of the black ops side of the CIA that doesn't think Bush is answer to their prayers?), but does make it less likely. I have to agree with Mr. Dyer- if it happens it's most likely to be true-blue Al Qeida (unless it's a home-grown militia).

David Kay, who resigned from being the lead WMD hunter in Iraq, has stated on the record that there were no WMDs in Iraq. No nukes, no chem, no bio.

The Bush Whitehouse, meanwhile, is proving that denial ain't just a river in Egypt.

I have to admit, I like this author's explanation of what Dean's real problem is:

Sam Smith, editor and publisher of The Progressive Review and a long-time observer of Washington politics, nailed it when he wrote a few days ago that Dean's biggest problem is his failure to properly bow down to the permanent power culture.

"Dean failed to accept the fact that before you can get elected by the people you have to be selected by the crowd in charge," wrote Smith. "You don't just run for president in the Democratic Party (unless you're a Sharpton or Kucinich doomed from the start); you ask permission nicely just like Clinton did. Show the elite that you want to come to Washington to serve them, not lead others. . . . It's bad enough when a Georgia peanut farmer like Carter tries it, but Dean came out of the establishment himself so his crime was worse: betrayal rather than naivete. And he paid the price.

"It's not political. Washington is a place where more things are done illegally or under the table than just about anywhere in the world. Where your laws are made - and broken - as Mark Russell used to say. And it's the world's most powerful private club. If you want to get ahead here the first thing you've got to do is shut your mouth. And show you respect the people who really run the place. Dean didn't do that."

Actually, I'd argue that the whole point of the Dean campaign is to break the back of the Democratic power elite. The whole idea of having a Democratic power elite is kind of wrong- an oxymoron, like jumbo shrimp, compassionate conservative, or Microsoft Works. My opinion is that even ignoring the philosophical problems, the leadership of the Democratic campaign should get fired for incompetence and disloyalty.

But the problem is that the Democratic power elite is nigh onto indistinguishable in many respects from the Republican power elite. Both of them are selected by a media that is not so much liberal or conservative, as pro-big-business, aka corporatist. The big corporations (and the billionaires who control them) are their customers- and the first rule of any business is never torque off your customers. It just helps that the mainstream media is a big business these days as well- that just makes sliding into the corporatist frame that much easier.

Which is why I don't have a lot of hope that the media will let up on Dean, despite the fact that he is no longer technically the front runner. Time will tell, however.

Krugman joins me in calling electronic voting machines a threat to our very Republic.

We all make fun of countries with sham elections- you know the ones, where everyone turns out to vote El Presidente back into office. But that is exactly what we ourselves are becoming. The real problem here is we don't need some grand Illuminati conspiracy- just one, maybe three, people in the right places to totally subvert democracy in the US.

What is deeply annoying about this is that computers could make voting easier, and be every bit as secure. What is needed, at an absolute minimum, is that the machine print out a ballot (not difficult nor expensive with how cheap bubblejets/laserjets have gotten recently) which the voter holds in their hands. The computer just helps put the ballot together. Inserting the printed ballot in the accept slot casts the vote. Then you just recount some statistically signfigant random subset of ballot machines (the computers should know exactly how many votes are in each ballot machine). Any difference between what the computer thinks and the actual slips of paper in the locked box triggers a total system recount, with the paper ballots being the real votes. And since which ballot boxes get recounted get picked by random after the election (but before the results are certified), any attempt at rigging the election (or any software bug that effects the recount) would almost certainly be detected. We could do so much better than any electronic voting machine I've seen. It'd almost be hard not to.

From the department of some good news for a change, we have this press release that the MPAA has finally admitted that CSS- the Content Scrambling System, the encryption scheme for DVDs, is no longer a trade secret.

It's the story with everything- Republicans breaking the law, Democrats being dumb and victimized, Robert Novak releasing private documents, and- the cherry on top- a lesson in computer security. It's the news article that showed up on Bug Traq (a computer security mailing list), Slashdot, and the Minnesota for Dean mailing list- and was on topic in all three. It seems that for over a year now, Republicans have been exploiting a weakness int the Congressional servers to read, and release to the press, confidential Democratic memos.

Now, as amusing as it might be to see several Republican congressional staffers, and maybe the odd Republican congressman, frog marched out of the Capitol in handcuffs for computer crimes (I don't think this will happen- that would require the Democrats have guts), what is most disturbing about this story is the utter and complete lack of technological savvy of the Democrats. They don't even know enough to know they need to hire someone.

It's the whole wise prince issue again. Machiavelli (required reading, IMHO) said that a wise prince will pick wise advisors- and listen to them. A foolish prince may, by chance or inheritance, end up with wise advisors, but he will not heed them. Howard Dean has demostrated that even if he himself isn't capable of setting up and locking down a Linux/Samba server, he knows enough to hire people who do- and listens to them. Not only about server issues- I comment that Dean was a guest blogger on the Lawrence Lessig blog (that was the point where he won my heart), is releasing his community building software under GPL, and now has had his position paper on the internet reviled by the Cato Institue.

The Republicans have also demostrated some technological awareness- at least at the script kiddy level. Too bad they're blackhats. Which makes them the social '31337 (I'm from the "never apologize, never explain" school of joke telling- if you don't get it, pass on). The Democrats, meanwhile, have demostrated less technological awareness that the Pointy Haired Boss. Which should be expected of the party which gave us two Microsoft "settlements" neither of which is worth the paper they're written on, the DMCA, the Sony Bono copyright extension, and three different incarnations of the CPA (internet censorship).

Which is part of the reason why I'm so disgusted with the Democratic party leadership- and so enamored of Dean.

Time for some unapologetic hardware geeking. The news is that the new P4 core will have a longer pipeline. I find this interesting news. Especially with the news that the double-pumped adder was going away.

When the original P4 came out, I formed the suspicion that something had gone wrong with the P4 design. My theory, in short (and I have to emphasize based upon almost no evidence- this is just me going off) was that the surprise assault of the 1GHz athlon had scared Intel. Up until that point, they had most of their engineers working on what would become the Itanium (and was at the time called Merced), with a much smaller crew working on the follow up to the P3 and last of the x86 line at Intel. But AMD stole a march on them, and in it's fear it pulled engineers off the Itanium and threw them at the P4 with marching orders to make the P4 go fast- and by that, management meant up the clock speed as much as possible. Intel had for years been touting the concept that clock speed == performance, something which is flatly untrue but easily for people not familiar with hardware to fall for.

Intel's sudden shift of hardware engineers had two effects. The first was that it delayed the Itanium and the Itanium 2, by probably something like 12-24 months. Enough so that instead of having incredible performance, they only had good performance- and that plus the cost to switch over killed those CPUs.

The second effect it had was on the P4. The easiest way to raise the clock rate on a CPU is increase the number of stages. Modern CPUs act a lot like assembly lines these days. The "work" of executing instructions is broken down into multiple stages, and different instructions can be executing in different stages. Sort of like building a car- you have one station putting the engine in, and the next stage attaching the doors, both stations working in parallel on different cars.

How many stages the CPU will have is general decided very early in the design process. You need to balance the work done at each stage to maximize the overall speed of the processor. It doesn't help if putting the engine in takes 10 minutes, and attaching the doors only 2- the people working on the doors will spend most of their day standing around waiting on the people putting the engine in. Same thing with CPUs. But when the orders came down to up the P4's clock rate, it was decided to simply double the number of pipeline stages. The theory here is that since each stage is now doing only half the work, it only needs half the time- allowing Intel to double the clock rate. Note, the actual amount of work being done doesn't increase (actually, it goes down some), but that wasn't important. What was important was having a really high clock rate.

The P4 succeeded, at least enough to force AMD to introduce it's marketinghertz, aka "performance ratings". But I still think something deep inside the P4 was wrong. Since the original work balancing between the stages wasn't completely redone, some stages were unbalanced. They took longer to complete than other stages. One of the stages that was too light- that completed too early, was the addition stage. So Intel decided to make lemonaid from their lemons and simply brag about their "double-pumped adder". At least until people figured out how difficult it was to take advantage of this "feature".

Fast forward to today. Now we're hearing that a) the double pumped adder is going away, and b) the new tweaked P4 core is going to have more stages. This sounds to me an awful lot like Intel went back through and split the heavy/slow stages into two, helping balance the pipeline again. This would lay the ground work for a serious speed bump if P4 processors, and just in time. As I've mentioned earlier, the Athlon-64/Opteron line is comming on strong. And while I don't see it reclaiming the clock rate crown from Intel, they could get close. Close enough to scare a jumpy Intel, and take a clear lead in real performance.

But, as usual, the situation isn't that simple. AMD has made the switch to Silicon On Insulator- SOI. The problem is heat. You can make a chip run faster by pumping more power into it. Vague analogy here- this is kind of like putting a bigger engine into a car- it makes the car go faster, at least until the frame bends and the powertrain breaks. With CPUs, increasing the power increases the speed, but also increases the heat. Increase the power too much and the chip melts. Literally. SOI allows you to get the same performance from the same process with less power- or, equivelently, more performance at the same process with the same power.

Unfortunately, the smaller the process the more power it takes. Which is why CPUs have gotten ever more elaborate heat sinks as time went on.
This spells trouble for Intel, as we shall see.

AMD switched to SOI with their 130nm process for the Athlon-64s and Opterons. This was an aggressive decision from AMD- a process shrink, a major switch in process technology (going to SOI is about as difficult as a process shrink, it appears to me), a 64-bit extension, a new processor core, an integrated memory controller, and a new bus standard, all in one heroic leap. This was aggressive because a delay in any of these would cause a delay in all of them. The SOI + 130nm conjoined allowed Intel to get half a step or so ahead of AMD, because Intel didn't do SOI when it went to 130nm. Nor much of anything else, 130nm was more of a standard process shrink.

AMD has cleared the hurdle. It looks like it's getting the last fab problems out of the way and is ramping up production. And, thanks to some combination of good design and SOI, AMD's 130nm processors are quite handily competing with the best Intel has to offer. Intel, meanwhile, is facing severe heat problems. See this article for some more info (AMD is using the same process as IBM's 970 fabs- IBM and AMD are swapping process technology). Especial as Intel tries to move to 65nm, and then 45nm, technologies- see this article.

The heat problem, however, is mainly a question fo real performance- how fast the transistors can toggle back and forth. You can raise the clock speed, however, without raising performance- do half as much work per step in twice as many steps. This inclines me to beleive that what Intel is going after is phantom speed boosts- bump the speed of the clock but not the speed of the processor. Without moving to SOI- which almost certainly won't happen in '04- it will be hard for Intel to see real speed jumps. Sooner or later Intel will be forced to go SOI- Intel is just hoping later rather than sooner, because doing so would slow it down, and put it back on par with AMD's process.

Those of us geeky enough to read benchmarks will likely not be fooled by this phantom performance boost. But Intel may have to write off the geeky segment. Especially with the extra geek cool factor of 64 bits in AMD's favor. What Intel is targetting with this move is your average pointy haired boss, who still thinks clock speed is an accurate measure of performance. The idea here is to use this perception of a performance increase to help stave off massive defections.

It's a Hail Mary pass from Intel. But, as this week has shown me, sometimes Hail Mary passes connect. We shall see.

Interesting results from Iowa. Dean has come in third- a rather distant third, in fact. Which was a rude enough shock in and of itself. What was the real surprise was numbers one and two- Kerry and Edwards.

First off, this isn't doom for Dean. A disappointment, sure. A cause for a campaign rethink, I hope so. But not a doom. Remember the last non-incumbant who won both IA and NH and the WH was Carter in 76- almost 30 years ago. And there is a long list of candidates who won Iowa and yet still lost the nomination. Iowa simply isn't that important.

Second, Gephardt is out- if not officially yet, then soon. Which has several positive effects for Dean. First is that Gephardt was by far the worst of the negative campaigners- which is one very good explanation of what happened. Gephardt's negative campaigning hurt him, but it also hurt Dean, enough for Kerry and Edwards to pull out surprise wins. And second, Gephardt was a real threat come mini-tuesday- three of the seven states having primaries/caucuses that day are MO, OK, and ND. Gephardt would have simply owned MO (his home state), and would have been very strong in OK and ND (fellow bread basket/midwestern ag/union states). I could have seen Gep taking three of the seven states comming mini-tuesday- meaning that Dean would have had to sweep Clark in all four states to beat Gep's take. Now, with Gep out, those three states are in play again (with no homefield advantage for any candidate).

Neither the Kerry not the Edwards campaigns have a lot of money (I could argue that Kerry's campaign is $6 million in the hole). Nor do they have a lot of time to raise that money- NH is in a week, mini-tuesday in two. Nor will the money just automatically start rolling in- the big donors might decide Iowa as a fluke and require the campaigns to prove them wrong by winning again. And both only have large campaign presences in one other state (NH for Kerry, SC for Edwards). They won Iowa by focusing on it. This is the classic problem with Iowa winners, and why so many of them go on to lose the nomination.

The states they selected for the second victories are also interesting- as it puts them in competition with Clark as much as Dean. Kerry especially with the whole war-hero thing. It's just a matter of time before they light into each other, especially if Dean can retake the underdog position- a position he does better at.

There are several things I think Dean can, must, do at this point. First and foremost, stop being quite such a tightwad. Yes, not having money after you win the nomination is a problem- but only if you win the nomination first. Edward's result came from his willingness to spend money on airtime. Volunteers help, but not as much as a good advertising campaign. More paid campaign staffers would be nice too.

The image the campaign itself puts out needs to be changed. How the heck did Dean lose his position as the health care candidate? To Gephardt of all people? This doesn't mean change positions, it means change the subject, which positions you talk about. Actually, talking about the war hurts Dean. In many ways it's his weakest spot- and by harping on it, he helped remind people of his lack of foreign policy experience. And it reinforces the whole "Angry Liberal Dean" meme. Focusing on other issues allows Dean to highlight his positives, and shuck the whole angry liberal image.

Third, and most importantly, win New Hampshire, preferrably by large (>10 pts) margins. Because if Dean doesn't win NH, it will be to one of Clark, Kerry, or Edwards, and that would put Dean into Hail Mary pass territory.

In a little noticed ruling recently, Microsoft was given permission to punish OEMs who sell PCs without OSs. Or at least not reward them, which amounts to the same thing. All Microsoft has to do is price it's OSs out of reach of the OEMs (who are in a low margin business and don't have a lot of profit to give up), and then "reward" those OEMs who play ball with the ability to actually make a profit. Or at least not be driven into bankruptcy quite so fast.

When this agreement was reached, I argued that a) it was too weak, b) it wouldn't get enforced, c) Microsoft would feel it couldn't afford to abide by the terms no matter what, and d) we'd be back in court with them in a couple of years anyways. We've done consent agreements with Microsoft before- the '95 consent agreement, violations of which lead to the most recent round of litigation. We need to stop giving them community service they don't bother showing up for anyways.

Note that my prediction that this would be the year of Linux on the desktop stands. I took into account the current "consent agreement", assuming it wouldn't be enforced (indeed, I completely forgot about it). Remember that, according to Linux, all you are buying with a site license is the OS upgrades- you need to buy an original copy of the OS seperately (i.e. preinstalled on the machine) which you then throw away (you can never use it on another machine) before "upgrading" to whatever OS you bought the site license for. The fact that this looks suspiciously like needing to buy 2 OSs for the machine is begining to dawn on people, which is why I predict Linux on the Desktop this year.

The graph in this Salon article alone is worth the price of admission. It's a graph, 1993-2003, of the month by month measure of job creation. What's striking is that except for the very tail end of 2000, the best Bush has ever managed to do did not quite get up to the worst that Clinton ever managed to do.

But what really kinda surprises me about this is not that the Bush record is so bad- it's that even with the worst job record of any administration since Herbert Hoover and the Great Depression, Bush is widely considered not only to have a chance in the 2004 election, but actually to be unbeatable. All of those unemployeed people, people without heath insurance, people who didn't even get cost of living increases, people who have no job security, are going to run right out and vote for Bush because otherwise the boogy man might get them.

What we've got here is failure to communicate. The media, by playing up the threat of terrorism sensitises us to a irrelevent threat- terrorism does exist, but you are far more likely to die slipping in the bathtub tomorrow than to get killed by a terrorist. Heart disease- now heart disease is a real killer (although auto accidents snuff out the largest number of potiential years of life). The threat of getting laid off or downsized is, on the contrary, very real. But since it's not publicized, your job woes are therefor seen as a unique, exceptional circumstance. So the image, as perpetrated by the media, is that fighting terrorist is way more important than fighting unemployeement. And thus Bush's aura of Invicibility.

Feeling that maybe congress shouldn't have a say in who gets appointed to the bench, Bush made Pickering an appeals judge via recess appointment. Generally, midterm appointments are only done in the case of a surprise vacancy and with unobjectionable candidates, but I've always wondered how long that would last with this administration.

What this gives is another opportunity for the Democrats to gain points. It's blantantly obvious to anyone who pays attention care not one fig about democratic processes- witness the redistricting in Texas, the stunts DeLay has pulled, and now this. It's another opportunity to slam the Republicans, especially Bush.

We'll see if they take it.