Lets have a closer look at that 866 Passmark points.
Compared to a single AMD Athlon XP 2400+ with 431 points, it looks a bit suspect but possible (I'd expect less than perfect scaling when going from one to two processors).
The AMD Athlon X2 Dual Core BE-2300 (2 cores at 1.9 GHz) comes in at 1033 points. Seems plausible by comparison, considering the improvements in AMD's technology since the XP.
You are right about the Core, if you used the Core Duo (without the "2") as basis for comparison. That was the Yonah (http://en.wikipedia.org/wiki/Yonah_(microprocessor)) laptop processor by the way. But for a "fair" comparison, the Core 2 Duo, built for the desktop, is the better basis. If we pick the Core2Duo E4300, the very first (and weakest) Core2Duo that went to market, it has 1056 passmark points. Later models were better of course.
Getting back to Hairyfeet's idea of using a Pentium D, the fastest in the Passmark list (the Extreme Edition 965 @3.73 GHz) has 1318 points. The Celeron E3400 that I suggested as alternative gets 1711 points.
Considering tech dumpster diving, I like that sport:-) But even so, the minimum I'm looking for is AMD64 X2 and Core2Duo. Refurbishing old P4 machines is simply not worth my time anymore, unless I have a bunch of parts that are known to be OK. Stuff from dumpsters needs triage, because some of it got there for being defective;-) BTW, the machine I'm typing this post on has an Athlon 64 X2 Dual Core 4600+ (2 cores at 2.4 GHz), 2GB RAM, a NVidia 8600GT and 1266 Passmark points. Perfectly good for surfing, office stuff and older games. But it tends to have problems with new games, so an upgrade is on the way...
I mostly agree, but the idea of using a Pentium D makes me cringe. TDPs of 95W or more, and being a Netburst part, it is clearly inferior to the Core architecture. If you have a socket 775 board, I would check if it can run something like the Celeron E3400 (Wolfdale core). This one is also a low cost dual core, will likely run rings around the Pentium D and is cooler with only 65W TDP.
This said, for $175 I would look around if someone has a "moderately old" PC to sell. At that price, you may be able to get an early Core2Duo with a mainboard that already has a PCI Express graphics slot. Maybe it even comes with a halfway decent graphics card.
I think the non-Windows market is large enough these days that at least one or two motherboard manufacturers will make boards where the user has a way to disable "Secure Boot" or add his own keys.
These may be more expensive though... I could imagine that this last refuge is in the market for server hardware, and instead of a $100 ASUS, you'd have to buy a $250 Tyan or such:-(
Maybe, but they are not really consequent about it. If the goal was getting rid of x86 dependencies, Vista (and later OSes) should have been centered around.net, with a XP virtual machine for compatibility only and no more new features in XP mode.
With the above strategy, Microsoft might have succeeded in pushing most Windows development to.net (possibly at the expense of driving more customers to Linux).
As it is, the virtual machine came later (in Win7) and only for Professional/Enterprise/Ultimate. And the Win32 API was extended further so it remained an attractive development target. My guess is that there won't be too many.net-only applications that will migrate painlessly from x86 to ARM.
Modern CPUs want a lot of memory bandwidth. Modern GPUs want a staggering amount of memory bandwidth (the GTX580, for example, has around 200 gigabytes per second and even the 8600GTS has 32 gigabytes per second whereas a dual-channel DD3-1600 system only has 25 gigabytes per second). Stick both of those on one chip and you're going to be starving both of them unless you have at least four memory channels.
Good point. Actually, even a HD66xx-class discrete GPU will run significantly faster if it gets GDDR5 memory instead of DDR3. That is 64 gigabytes per second to make the GPU really happy. And with the next generation Llano, that problem will be more pronounced as the GPU gets faster and the performance will be closer to a HD67xx.
I wonder how much it would cost to put a 4-channel memory controller into the Llano (except the obvious need for yet another socket format). AMD certainly could do it, they already have 4-channel memory for the Opterons. But would it be economical?
The AMD Phenom (AM3 socket) series has one advantage over Intel for consumer cpus, and that is they all support ECC while Intel's consumer SandyBridge does not (caveat: you have to find an AMD mobo which supports ECC, not all of them do even though the socket format does). You have to move on to Intel's Xeon SandyBridge to get ECC, and there the pricing premium becomes significant. Very few people want the added cost of ECC (I seem to be the only one who really cares:-( ) for a consumer cpu. Intel clearly has pricing power here too.
Well, there's two of us who care. Once, I've seen a colleague at work despairing over a randomly crashing PC. Not fun, and it turned out to be bad RAM. I guess ECC would have prevented that or at least given a clear error message. As a consequence of Intel's policies, I just got myself an AMD Phenom II instead of an Intel Sandy Bridge for my current upgrade. The Xeons were just too expensive for my taste...
45 nm is correct for the AMD Phenom II series. But the "Llano" APUs for low-end desktops are already in 32 nm. So you could say Intel is half a step ahead right now. Overall, however, I agree that AMD is under pressure and cannot afford artificial delays in their products.
What they still have are some niches where Intel has slacked off or does not compete for other reasons. The most important one right now are the APUs. AMD's Brazos platform does well on netbooks, and IMHO the LLano is a good choice for cheap consumer PCs. Another one is (desktop) CPUs with support for ECC RAM: Intel does not support that feature in its current "Core iX" CPUs at all, Presumably because they want to extract extra money from customers who need it by making them buy the much more expensive Xeons. Well, I'm a bit paranoid about reliability myself and thus I just ended up ordering a Phenom II instead of a (otherwise superior) Sandy Bridge CPU.
If you choose your appliances for moderate energy consumption, then even less is possible. My electrical power consumption (not running my own washing machine, no electrical air conditioning or heating) is around 1000kWh/year or ~3kWh/day. Yet I don't think I'm missing out on much quality of life.
This is made possible by mostly using compact fluorescent lamps, a computer built from energy efficient parts and a well insulated apartment that does not heat up too much in summer, despite being under the roof.
The common method is using an ATM skimmer to copy the card, and a camera to record typing in of the code. No mugging necessary. Sometimes the keypad is faked too.
Even as a usually law-abiding citizen, I might be tempted to steal that camera thingy if i find it. The fact that it was put there by criminals would greatly reduce my pangs of conscience;-)
But... at the same time, you have dozens of competitors emulating WoW too. And all of them taken together will likely take away only a small part of WoW's market share. The little they take is then somehow divided up amongst these dozens of copies.
So each WoW copy will acquire much less customers than they think, and I guess it is actually a bad monetary decision to copy WoW;-)
Eve PvP is a lot about numbers. Half a dozen frigates can usually kill a cruiser, but the frigates are cheaper to buy. The background is that firepower (and room for equipment) in Eve goes up much slower than price and ship size. So if you have enough buddies, together you can go ahead and attack anything.
The flip side is that the end game really forces you into PvP alliances that can raise sizable fleets (PvE gets boring at some point). Those alliances tend to demand regular participation in patrols and fleet battles. Ultimately, this was one of my reasons to leave EvE, the time commitment was too much on top of a 40hr week at work. But I still think it is a very well designed game.
I have two computers with similar hardware and OS at work and at home. Windows XP SP3 plus security updates, IDE disk, dual core processors of not too different performance.
My private computer at home (no virus scanner, no login scripts) boots reasonably fast. Two minutes maximum until I have a responsive desktop.
The machine at work is loaded down with McAffee, logs into a Windows domain and has a few login scripts to process. A typical start of the work day is like this: -I boot and log in (that part is reasonably fast, it does not take much longer than on my private machine) -I go to the coffee maker and make my first coffee of the day. Usually I take the time for the deluxe version with frothed milk. When I come back, the desktop is marginally usable, but the hard disk is still busy. I suspect a preemptive scan by McAffee, but the Task Manager shows something like 98% idle, so whatever it is hides its activity. -About 10 minutes after booting, hard disk activity goes back to "normal" and I can actually get work done.
There is still a lot of desert we could place photovoltaics in. The main reason this is not done big time yet is that fossil energy is so cheap. Looking at the EEX (http://www.eex.com/de/) right now, the average price is around 5 Euro-cents per kWh.
Renewables have a long way to go before they can compete with that. But compared to the prices the end user pays (typically 20-25 cents/kWh in Germany), photovoltaics are getting close and wind power is already cheaper. So I think our civilisation could survive on that, if we develop less wasteful habits in using energy and raw materials.
Good example, but also one that will be irrelevant soon. Those patents appear to be from the early 1990s and will run out soon. Besides, lithium-ion batteries (http://en.wikipedia.org/wiki/Lithium-ion) are increasingly displacing NiMH for cars anyway, as the energy density is superior.
The problem with that approach is that the network (at least the "last mile" leading to customers' residences) is a natural monopoly. See http://en.wikipedia.org/wiki/Natural_monopoly for a definition. For a competitor, it is usually not worthwhile to build a parallel network if he can reach only a few customers. Result: The incumbent ISP can like a fuckhead and get away with it.
A way to solve that would be a public network where the customer can choose his provider and the provider can then rent the wire from the customer's house to the next telephone exchange. Germany got that one halfway right: When the telecommunications branch of the former Deutsche Post (public mail and telecom authority) was privatized, the new company "Deutsche Telekom" also got the network - under the condition that they rent out the "last mile" to competitors if the customer wants to go with one of those. A new regulation authority controls the price for that rent. As a result, Germany actually has DSL competition in most places. Of course, there is still a lot of bickering between Deutsche Telekom and the competition about how much rent is fair, and the regulation authority is needed to keep the Deutsche Telekom from charging excessive rates. But by and large it works.
I probably wouldn't care too much as an end user. And not being hurt directly, it might be difficult to make a lawsuit out of it even if I cared. What damages to claim? Amazon, however, might have reason to take this to court. They also have much more resources to fight out a lawsuit. Which does, unfortunately, make a difference. Maybe some state attorney who has a clue about the internet might also be interested, but don't hold your breath for that.
I predict that there will always be some interesting games without excessive DRM. The may be a small fraction of the market, but they are there. In some cases, indie projects. In some cases, older mainstream titles where the publisher does not insist on DRM anymore. Case in point, I recently bought the X3 Gold Edition. One of the games (X3 Reunion) is on the installation DVD without copy protection, for the other (X3 Terran Conflict) there is a patch from the vendor to remove copy protection.
Now I'm only waiting for my new joystick to be delivered...
Interesting that such extremely tiny microcontrollers are still being made. I'd have expected things to start at something like the ATtiny85 (just googled at http://www.atmel.com/dyn/products/product_card.asp?part_id=3612). Still small at 8 kByte Flash and 0.5 kByte RAM, but it gives you a bit more to work with. The price difference is size and money is small. 8 pins vs. 6, and the first online store where I found it charges $2.40.
Unless you are planning some mass produced, really cheap product that needs a microcontroller, I'd say optimizing your code to fit in the ATtiny4 comes under "premature optimization".
Which does not mean I disagree with pride in 'doing it right', but in most cases finishing the product might be more important than squeezing another dollar out of the production costs.
This is slightly offtopic and anecdotal, but there seems to be progress in fixing the import filters.
I have a particularly problematic Word document that both OpenOffice 3.3 and LibreOffice 3.3 could not load without problems. OO 3.3 would drop some text embedded in graphics, LO 3.3 messed up the table of contents. LO 3.4.2 ist the first version that imports the file without obvious problems:-)
If you choose strings with length count, the format of the length field should be fixed in the specification. If necessary, use multiple string types. As in "shortstring": 8bit unsigned length field, holds up to 255 characters "string": 16bit unsigned length field, holds up to 65535 characters "longstring": 32 bit unsigned length field, holds up to 2^32-1 characters
In general, I think making the length of "integer" depend on the architecture is a stupid idea, because of breakage when the size of the address space changes. Better to have a separate type "pointer" and add larger integer types as needed. That way, application programmers have a stable definition and systems programmers have a pointer type that adapts in size.
Firstly, it makes it seem like the address+length format is a no-brainer, but there are quite a lot of problems with that. It would have had the undesirable consequence of making a string larger than a pointer. Alternatively, it could be a pointer to a length+data block, but then it wouldn't be possible to take a suffix of a string by moving the pointer forward. Furthermore, if they chose a one-byte length, as the article so casually suggests as the correct solution (like Pascal), it would have had the insane limit of 255-byte strings, with no compatible way to have a string any longer. (Though a size_t length would make more sense.) Furthermore, it would be more complex for interoperating between languages -- right now, a char* is a char*. If we used a length field, how many bytes would it be? What endianness? Would the length be first or last? How many implementations would trip up on strings > 128 bytes (treating it as a signed quantity)? In some ways, it is nice that getaddrinfo takes a NUL-terminated char* and not a more complicated monster. I'm not saying this makes NUL-termination the right decision, but it certainly has a number of advantages over addr+length.
A lot of those questions can be handled by a clear specification of the string types. Things like "Would the length be first or last?" or "how long is the length field" obviously belong in the specification (and someone who does not think that far should not be allowed to design programming languages other people have to use).
Some more of these are a general problem if you mix languages and port to other systems. Sure, "a char* is a char*" works for strings. But how many programs handle only strings? Most likely, you have some integers as well, and then the endianness of the length field (which is probably some sort of integer) comes back as a more general question about the endianness of integers. So you have to deal with that problem anyway.
So I think the downsides of Pascal style strings are overrated. My two cents on how a useful specification for the 16bit systems of the 1970s could look: 1) A "string" consists of a length field at the beginning and ASCII characters after the length field. 2) The length field is an unsigned 16bit integer. Endianness depends on the architecture (whatever the CPU normally uses). For exchange between computers, the application programmer is responsible for creating a unified format.
This string format would be only one byte longer for the same number of characters (you don't need the NUL byte at the end) and allow pretty long strings of 64kByte. If that extra byte really hurts, add a second "shortstring" data type with 8bit unsigned integer as length field.
In some countries (for instance Germany) it is actually legally impossible to sign away all of the creator's rights. You can transfer the rights for commercial reproduction, but not, for instance, the right to prevent defacing of the work.
There are some cases where architects sued building owners over changes to a building - and those lawsuits tend to be successful.
I actually know a project manager who worked like this. The guy was laid off about two years ago, and while I don't know management's actual reasons for firing him, it was the best decision they made in a loooong time:-)
Slashdot Mirror
Lets have a closer look at that 866 Passmark points.
Compared to a single AMD Athlon XP 2400+ with 431 points, it looks a bit suspect but possible (I'd expect less than perfect scaling when going from one to two processors).
The AMD Athlon X2 Dual Core BE-2300 (2 cores at 1.9 GHz) comes in at 1033 points. Seems plausible by comparison, considering the improvements in AMD's technology since the XP.
You are right about the Core, if you used the Core Duo (without the "2") as basis for comparison. That was the Yonah (http://en.wikipedia.org/wiki/Yonah_(microprocessor)) laptop processor by the way. But for a "fair" comparison, the Core 2 Duo, built for the desktop, is the better basis. If we pick the Core2Duo E4300, the very first (and weakest) Core2Duo that went to market, it has 1056 passmark points. Later models were better of course.
Getting back to Hairyfeet's idea of using a Pentium D, the fastest in the Passmark list (the Extreme Edition 965 @3.73 GHz) has 1318 points. The Celeron E3400 that I suggested as alternative gets 1711 points.
Considering tech dumpster diving, I like that sport :-) ;-)
But even so, the minimum I'm looking for is AMD64 X2 and Core2Duo. Refurbishing old P4 machines is simply not worth my time anymore, unless I have a bunch of parts that are known to be OK. Stuff from dumpsters needs triage, because some of it got there for being defective
BTW, the machine I'm typing this post on has an Athlon 64 X2 Dual Core 4600+ (2 cores at 2.4 GHz), 2GB RAM, a NVidia 8600GT and 1266 Passmark points. Perfectly good for surfing, office stuff and older games. But it tends to have problems with new games, so an upgrade is on the way...
I mostly agree, but the idea of using a Pentium D makes me cringe. TDPs of 95W or more, and being a Netburst part, it is clearly inferior to the Core architecture.
If you have a socket 775 board, I would check if it can run something like the Celeron E3400 (Wolfdale core). This one is also a low cost dual core, will likely run rings around the Pentium D and is cooler with only 65W TDP.
This said, for $175 I would look around if someone has a "moderately old" PC to sell. At that price, you may be able to get an early Core2Duo with a mainboard that already has a PCI Express graphics slot. Maybe it even comes with a halfway decent graphics card.
I think the non-Windows market is large enough these days that at least one or two motherboard manufacturers will make boards where the user has a way to disable "Secure Boot" or add his own keys.
These may be more expensive though... :-(
I could imagine that this last refuge is in the market for server hardware, and instead of a $100 ASUS, you'd have to buy a $250 Tyan or such
Maybe, but they are not really consequent about it. If the goal was getting rid of x86 dependencies, Vista (and later OSes) should have been centered around .net, with a XP virtual machine for compatibility only and no more new features in XP mode.
With the above strategy, Microsoft might have succeeded in pushing most Windows development to .net (possibly at the expense of driving more customers to Linux).
As it is, the virtual machine came later (in Win7) and only for Professional/Enterprise/Ultimate. And the Win32 API was extended further so it remained an attractive development target. .net-only applications that will migrate painlessly from x86 to ARM.
My guess is that there won't be too many
Modern CPUs want a lot of memory bandwidth. Modern GPUs want a staggering amount of memory bandwidth (the GTX580, for example, has around 200 gigabytes per second and even the 8600GTS has 32 gigabytes per second whereas a dual-channel DD3-1600 system only has 25 gigabytes per second). Stick both of those on one chip and you're going to be starving both of them unless you have at least four memory channels.
Good point. Actually, even a HD66xx-class discrete GPU will run significantly faster if it gets GDDR5 memory instead of DDR3. That is 64 gigabytes per second to make the GPU really happy. And with the next generation Llano, that problem will be more pronounced as the GPU gets faster and the performance will be closer to a HD67xx.
I wonder how much it would cost to put a 4-channel memory controller into the Llano (except the obvious need for yet another socket format). AMD certainly could do it, they already have 4-channel memory for the Opterons. But would it be economical?
The AMD Phenom (AM3 socket) series has one advantage over Intel for consumer cpus, and that is they all support ECC while Intel's consumer SandyBridge does not (caveat: you have to find an AMD mobo which supports ECC, not all of them do even though the socket format does). You have to move on to Intel's Xeon SandyBridge to get ECC, and there the pricing premium becomes significant. Very few people want the added cost of ECC (I seem to be the only one who really cares :-( ) for a consumer cpu. Intel clearly has pricing power here too.
Well, there's two of us who care. Once, I've seen a colleague at work despairing over a randomly crashing PC. Not fun, and it turned out to be bad RAM. I guess ECC would have prevented that or at least given a clear error message. As a consequence of Intel's policies, I just got myself an AMD Phenom II instead of an Intel Sandy Bridge for my current upgrade. The Xeons were just too expensive for my taste...
45 nm is correct for the AMD Phenom II series. But the "Llano" APUs for low-end desktops are already in 32 nm. So you could say Intel is half a step ahead right now. Overall, however, I agree that AMD is under pressure and cannot afford artificial delays in their products.
What they still have are some niches where Intel has slacked off or does not compete for other reasons. The most important one right now are the APUs. AMD's Brazos platform does well on netbooks, and IMHO the LLano is a good choice for cheap consumer PCs.
Another one is (desktop) CPUs with support for ECC RAM:
Intel does not support that feature in its current "Core iX" CPUs at all, Presumably because they want to extract extra money from customers who need it by making them buy the much more expensive Xeons. Well, I'm a bit paranoid about reliability myself and thus I just ended up ordering a Phenom II instead of a (otherwise superior) Sandy Bridge CPU.
If you choose your appliances for moderate energy consumption, then even less is possible.
My electrical power consumption (not running my own washing machine, no electrical air conditioning or heating) is around 1000kWh/year or ~3kWh/day. Yet I don't think I'm missing out on much quality of life.
This is made possible by mostly using compact fluorescent lamps, a computer built from energy efficient parts and a well insulated apartment that does not heat up too much in summer, despite being under the roof.
The common method is using an ATM skimmer to copy the card, and a camera to record typing in of the code. No mugging necessary. Sometimes the keypad is faked too.
Even as a usually law-abiding citizen, I might be tempted to steal that camera thingy if i find it. The fact that it was put there by criminals would greatly reduce my pangs of conscience ;-)
But... at the same time, you have dozens of competitors emulating WoW too. And all of them taken together will likely take away only a small part of WoW's market share. The little they take is then somehow divided up amongst these dozens of copies.
So each WoW copy will acquire much less customers than they think, and I guess it is actually a bad monetary decision to copy WoW ;-)
Eve PvP is a lot about numbers. Half a dozen frigates can usually kill a cruiser, but the frigates are cheaper to buy. The background is that firepower (and room for equipment) in Eve goes up much slower than price and ship size. So if you have enough buddies, together you can go ahead and attack anything.
The flip side is that the end game really forces you into PvP alliances that can raise sizable fleets (PvE gets boring at some point). Those alliances tend to demand regular participation in patrols and fleet battles.
Ultimately, this was one of my reasons to leave EvE, the time commitment was too much on top of a 40hr week at work. But I still think it is a very well designed game.
More like McAffee and login scripts.
I have two computers with similar hardware and OS at work and at home. Windows XP SP3 plus security updates, IDE disk, dual core processors of not too different performance.
My private computer at home (no virus scanner, no login scripts) boots reasonably fast. Two minutes maximum until I have a responsive desktop.
The machine at work is loaded down with McAffee, logs into a Windows domain and has a few login scripts to process. A typical start of the work day is like this:
-I boot and log in (that part is reasonably fast, it does not take much longer than on my private machine)
-I go to the coffee maker and make my first coffee of the day. Usually I take the time for the deluxe version with frothed milk. When I come back, the desktop is marginally usable, but the hard disk is still busy. I suspect a preemptive scan by McAffee, but the Task Manager shows something like 98% idle, so whatever it is hides its activity.
-About 10 minutes after booting, hard disk activity goes back to "normal" and I can actually get work done.
Were they to add some of the non-hybrid technology used in hybrids (regenerative braking, etc.) they could probably put that rating up pretty quickly
Already happening. See http://en.wikipedia.org/wiki/Start-stop_system or http://en.wikipedia.org/wiki/Mild_hybrid for various levels of this. Of course, those are not quite as efficient as a Prius with its more sophisticated systems.
There is still a lot of desert we could place photovoltaics in. The main reason this is not done big time yet is that fossil energy is so cheap. Looking at the EEX (http://www.eex.com/de/) right now, the average price is around 5 Euro-cents per kWh.
Renewables have a long way to go before they can compete with that. But compared to the prices the end user pays (typically 20-25 cents/kWh in Germany), photovoltaics are getting close and wind power is already cheaper. So I think our civilisation could survive on that, if we develop less wasteful habits in using energy and raw materials.
Good example, but also one that will be irrelevant soon.
Those patents appear to be from the early 1990s and will run out soon. Besides, lithium-ion batteries (http://en.wikipedia.org/wiki/Lithium-ion) are increasingly displacing NiMH for cars anyway, as the energy density is superior.
The problem with that approach is that the network (at least the "last mile" leading to customers' residences) is a natural monopoly. See http://en.wikipedia.org/wiki/Natural_monopoly for a definition. For a competitor, it is usually not worthwhile to build a parallel network if he can reach only a few customers. Result: The incumbent ISP can like a fuckhead and get away with it.
A way to solve that would be a public network where the customer can choose his provider and the provider can then rent the wire from the customer's house to the next telephone exchange. Germany got that one halfway right:
When the telecommunications branch of the former Deutsche Post (public mail and telecom authority) was privatized, the new company "Deutsche Telekom" also got the network - under the condition that they rent out the "last mile" to competitors if the customer wants to go with one of those. A new regulation authority controls the price for that rent.
As a result, Germany actually has DSL competition in most places. Of course, there is still a lot of bickering between Deutsche Telekom and the competition about how much rent is fair, and the regulation authority is needed to keep the Deutsche Telekom from charging excessive rates. But by and large it works.
I probably wouldn't care too much as an end user. And not being hurt directly, it might be difficult to make a lawsuit out of it even if I cared. What damages to claim?
Amazon, however, might have reason to take this to court. They also have much more resources to fight out a lawsuit. Which does, unfortunately, make a difference.
Maybe some state attorney who has a clue about the internet might also be interested, but don't hold your breath for that.
I predict that there will always be some interesting games without excessive DRM. The may be a small fraction of the market, but they are there.
In some cases, indie projects. In some cases, older mainstream titles where the publisher does not insist on DRM anymore. Case in point, I recently bought the X3 Gold Edition. One of the games (X3 Reunion) is on the installation DVD without copy protection, for the other (X3 Terran Conflict) there is a patch from the vendor to remove copy protection.
Now I'm only waiting for my new joystick to be delivered...
Interesting that such extremely tiny microcontrollers are still being made. I'd have expected things to start at something like the ATtiny85 (just googled at http://www.atmel.com/dyn/products/product_card.asp?part_id=3612). Still small at 8 kByte Flash and 0.5 kByte RAM, but it gives you a bit more to work with. The price difference is size and money is small. 8 pins vs. 6, and the first online store where I found it charges $2.40.
Unless you are planning some mass produced, really cheap product that needs a microcontroller, I'd say optimizing your code to fit in the ATtiny4 comes under "premature optimization".
Which does not mean I disagree with pride in 'doing it right', but in most cases finishing the product might be more important than squeezing another dollar out of the production costs.
This is slightly offtopic and anecdotal, but there seems to be progress in fixing the import filters.
I have a particularly problematic Word document that both OpenOffice 3.3 and LibreOffice 3.3 could not load without problems. OO 3.3 would drop some text embedded in graphics, LO 3.3 messed up the table of contents. LO 3.4.2 ist the first version that imports the file without obvious problems :-)
If you choose strings with length count, the format of the length field should be fixed in the specification. If necessary, use multiple string types. As in
"shortstring": 8bit unsigned length field, holds up to 255 characters
"string": 16bit unsigned length field, holds up to 65535 characters
"longstring": 32 bit unsigned length field, holds up to 2^32-1 characters
In general, I think making the length of "integer" depend on the architecture is a stupid idea, because of breakage when the size of the address space changes. Better to have a separate type "pointer" and add larger integer types as needed. That way, application programmers have a stable definition and systems programmers have a pointer type that adapts in size.
Firstly, it makes it seem like the address+length format is a no-brainer, but there are quite a lot of problems with that. It would have had the undesirable consequence of making a string larger than a pointer. Alternatively, it could be a pointer to a length+data block, but then it wouldn't be possible to take a suffix of a string by moving the pointer forward. Furthermore, if they chose a one-byte length, as the article so casually suggests as the correct solution (like Pascal), it would have had the insane limit of 255-byte strings, with no compatible way to have a string any longer. (Though a size_t length would make more sense.) Furthermore, it would be more complex for interoperating between languages -- right now, a char* is a char*. If we used a length field, how many bytes would it be? What endianness? Would the length be first or last? How many implementations would trip up on strings > 128 bytes (treating it as a signed quantity)? In some ways, it is nice that getaddrinfo takes a NUL-terminated char* and not a more complicated monster. I'm not saying this makes NUL-termination the right decision, but it certainly has a number of advantages over addr+length.
A lot of those questions can be handled by a clear specification of the string types. Things like "Would the length be first or last?" or "how long is the length field" obviously belong in the specification (and someone who does not think that far should not be allowed to design programming languages other people have to use).
Some more of these are a general problem if you mix languages and port to other systems. Sure, "a char* is a char*" works for strings. But how many programs handle only strings? Most likely, you have some integers as well, and then the endianness of the length field (which is probably some sort of integer) comes back as a more general question about the endianness of integers. So you have to deal with that problem anyway.
So I think the downsides of Pascal style strings are overrated. My two cents on how a useful specification for the 16bit systems of the 1970s could look:
1) A "string" consists of a length field at the beginning and ASCII characters after the length field.
2) The length field is an unsigned 16bit integer. Endianness depends on the architecture (whatever the CPU normally uses). For exchange between computers, the application programmer is responsible for creating a unified format.
This string format would be only one byte longer for the same number of characters (you don't need the NUL byte at the end) and allow pretty long strings of 64kByte.
If that extra byte really hurts, add a second "shortstring" data type with 8bit unsigned integer as length field.
In some countries (for instance Germany) it is actually legally impossible to sign away all of the creator's rights. You can transfer the rights for commercial reproduction, but not, for instance, the right to prevent defacing of the work.
There are some cases where architects sued building owners over changes to a building - and those lawsuits tend to be successful.
I actually know a project manager who worked like this. The guy was laid off about two years ago, and while I don't know management's actual reasons for firing him, it was the best decision they made in a loooong time :-)