WTF : Watched the Fucking

The fuck you say? I thought the people working on Battleship did it out of a simple love of the craft of film making.

When I was a kid we only had white. The black pixels had to sit at the back of the TV.

If a storage medium is dense but a bit slow, it just means it has a place lower in the storage hierarchy

Rather like if a slashdot poster is dense and a bit slow, he has a place lower in the slashdot hierarchy

I remember an exchange here that went like this (Google can't find it, because the Googlebot has better things to do than remember four year old slashdot posts. Unlike me)

Primus: .... The Macs UI us hard to use. You need to click on the very narrow window border to resize the window.

Secundus: Very narrow border? What are you, some kind of spastic?

Tertius: Mac fans show their people skills once again.

That's actually a very interesting idea. Qualcomm made a lot of money from a designing their own ARM core, the Snapdragon, from the micro-architecture upwards instead of buying a hard macro from ARM like all the other SOC vendors. The end result was that Scorpion had clock speed advantage over contemporary hard macros.

Look at the number of phones that used it here

http://en.wikipedia.org/wiki/Snapdragon_(system_on_chip)

Clearly there's market for tuned up ARM cores. Intel as you say has an ARM license and the people to do this sort of project. They've got the cash too - I've heard that Qualcomm spent $50 million on developing the custom microarchitecture for Snapdragon but if you look at the number of design wins they got it was worth it. If Intel cancelled Itanium and moved the design team to an a custom ARM architecture project it seems like it would be cheaper. Assuming they've got enough people working on x64 of course - that clearly must be their priority. Then again, they only need to spend more than AMD to stay in the game - maybe they could rest on their laurels for a bit on x64 and deploy resources on ARM.

I don't really see ARM as being a competitor to x64 on netbooks because of software compatibility. But on phones it is pretty much the opposite - people are going to pick the best ARM core rather than switch over to x64 because ARM is a much better fit. Also I can see ARM being used in datacenters. In fact a lot of the problem with ARM is that there aren't many ARM SOCs with a fast memory controller because they are all tuned for low power. So ARM scores poorly compared to even an Atom at memory intensive benchmarks. Someone is going to do a desktop or server class ARM in the end anyway, why shouldn't it be Intel?

My HP Officejet 6500 is like that You can get a huge "full featured" driver, or a simpler "basic scan and print" one.

I used to download the basic one on XP (the full featured one never seemed to be stable), but since Windows 7 if you add it as a network printer you get the built in driver which is fine. No scanning, but if you plug in a USB cable you get TWAIN drivers installed.

I actually tried the basic driver to get network scanning but it failed halfway through install. Doing a quick Google search it seems like that is quite common and yet there is no fix for it. In fact if you read through the 'fixes' people have suggested, most of them end up with "add the printer from the control panel", which really means "use the drivers built in to Windows".

So the built in drivers - which are perfectly stable, the only problem is I need to plug in a USB cable to scan - are the only ones worth using. And they're based on the venerable UNIDRV, a technology which has been around since Windows 3.1. Basically UNIDRV lets the you create a text file which defines the escape sequences for a printer. No wonder they don't support scanning! in fact one of the GPD files mentions

*CodePage: 1252 *% Windows 3.1 US (ANSI) code page.

So it seems like HP actually maintain a "Full" driver, a "Basic" driver and the config files for UNIDRV which end up baked into Windows. You have to wonder why they bother, particularly as the "Basic" and "Full" drivers seem to be such a nightmare to support and even get working.

I always used to do a wipe and reload. However with recent machines - mostly netbooks - I've found that just replacing Norton antivirus with Microsoft Security Essentials and stopping any irritating OEM stuff that runs on startup seems to be fine.

Then again if Microsoft gave me a copy of Windows I could do a clean install on like they used to, I'd have a choice. Of course the reason they don't do that is because the crapware vendors probably subsidize the price of the machine so heavily that Windows Starter is essentially free, so it's a bit of a sensitive issue for both Microsoft and their OEMs.

Microsoft are so cack handed these days that this initiative will probably irritate the crapware vendors enough for them to stop subsidizing Windows if it generates any money for them - after all there's a difference between the user doing this (or paying a third party to do it) and Microsoft doing it. Clearly at the very least it will irritate the OEMs but putting pressure on them to reduce the amount of crapware they install, but that is not necessarily in Microsoft's interests because it pushes down the market value OEMs will attach to Windows.

Very plausibly Dell paid $50 for Windows and got a significant percentage of that back in cash from the crapware vendors. If Dell is under pressure to install less crapware they will look to save that cash on the price of a Windows license.

VLIW seems like a dead end to me. Intel - who really can afford to do R&D - spent a lot of time and money on Itanium and it seems like it's not really competitive with x64 even if you're willing to recompile your code for each micro architecture iteration.

It seems like the patents should run out quite soon on most of the core x86 stuff. And AMD were willing to license x64 to Transmeta and Via (Via already have an x86 license from Intel). So maybe x86 will become an open architecture that anyone can implement if they pay up some fees, a bit like MIPS.

It's worth pointing out that China has an unlicensed MIPS clone called Longson - basically they skipped the patented instructions. Unlike Lexra, they made them NOPS - Lexra made them fault and MIPS sued them on the grounds that it was possible to emulate them in software.

Despite all that I've read that by partnering with MIPS licensees it would theoretically be possible to sell Longsons in the US. Of course the problem is that MIPS is a bit outdated and is not really competitive with x86 for performance heavy stuff or for ARM for low power. And obviously ARM is very open - it is even possible to get a license that lets you build ARMS with custom micro architecture like Qualcomm do.

So it does make you wonder if at some point x64 will become similarly open. x64 chips tend to lead performance per core so there's a certain argument for this.