A touch screen keyboard is not better than a hardware keyboard for a "creation" device.
A touch screen input system may be better for creation tasks if it has appropriate customizations for each task. A hardware keyboard is ergonomically better and provides better feel and feedback, so a general "touch screen keyboard" that copies the limitation of a hardware keyboard (same layout for all tasks) won't be very good. But something that adapts to tasks may be very good for "creation" tasks (particularly those that aren't primarily about creation of text, for which typical hardware keyboards are fairly well optimized.)
Re:Theoretical performance vs real-world performan
on
Knuth Got It Wrong
·
· Score: 1
But everyone already knows that (or should). Knuth knows that as well, caches and paging systems are not unknown things to him.
I agree. In GP I was clarifying what the issue was, not defending the idea in the TFS headline that "Knuth was wrong". A more accurate (though less headline-worthy) title would be "A naive reading of Knuth may be more misleading than most people are aware."
Except it's a bounded increase. That's my whole point. There is nothing slower than the hard drive to cache from.
That still doesn't make the assumption of constant memory cost valid, even in the asymptotic case, unless all cache misses are created equal such that cache misses act like direct random access, and the algorithm asymptotically approaches a fixed cache miss rate.
Since hard drives aren't random access devices, anyway, even ignoring cacheing, this is unlikely.
I agree with all of your post except this part. Algorithmic complexity theory is about orders of magnitude, not about precise numbers. That's why we have O(n) as a complexity class but not O(2n), O(3n) as separate classes; saying that an algorithm has O(n) worst-case time performance is saying that the time it takes to run it is approximated by some linear function of the problem size.
Right. And if that is true based on the assumption that all memory accessed by the system has an equal and constant access costs, but in the real world (because of cache) memory does not have equal and constant access cost, an algorithm which has O(n) theoretical performance under the assumption of equal and constant memory access cost may not scale even approximately linearly on a real world system -- it may scale quadratically or even worse, depending on it smemory access pattern and the structure of the underlying cache layers.
(Assuming that consistent cache misses result in constant time [albeit slow] access, the big-O of the algorithm would still be a valid asymptotic limit, though misleading in terms of real-world performance, but I'm not sure that assumption is valid of real world caches. The big point is that any a priori performance analysis rests on some set of assumptions about the characteristics of a system, and says nothing about the performance of a system that does not share those characteristics.)
What we really care about when we call it O(n) is that that's better than something like O(n^2) (a.k.a. polynomial time [wikipedia.org] or just "P") or O(log n), no matter what computational model you assume.
That's nonsense unless you mean "computational model" in a sense that is irrelevant to the immediate issue. If all you know is that an algorithm that has an asymptotic running time that is O(n) under the assumption of constant-cost memory access, you don't know the order of the asymptotic limit of its running time on a system that doesn't have constant-cost memory access. Under a particular scheme of non-constant-cost memory access, it might be O(n), it might be O(n log n), it might be O(n^n), it might be something else.
If the assumptions under which the performance scaling were calculated don't apply to your situation, neither does the calculation.
Sure, but not enough to really make a big enough impact to warrant an entirely new system. There's always going to be a bottleneck on performance somewhere, so even if they build a new system with the latest and greatest gfx technology, the visual experience is still limited by something as simple as the read speed of the disc drive. Enjoy those shiny graphics as the cutscenes sputter along...
Read speed of disc drives has also improved in the past five years. Also, its quite possible a new generation of console could use a primary game media other than discs.
Perhaps not, but it's the only way they'd advance that translates into sales. When you see marketing materials, do you see extreme closeup screenshots of the processor? No... you see screenshots of the games in action... the graphical detail.
Sure, if you are looking in magazines. If you are looking at in-store demos (which are also marketing material) you see the graphics, sure, but also the actual gameplay. Other improvements besides in the output resolution can impact the quality of the gameplay.
Re:Cache behavior is hard and not portable
on
Knuth Got It Wrong
·
· Score: 1
I am sympathetic with his argument, but trying to figure out the best algorithm for a given OS and hardware combo is very difficult and usually not portable.
Assuming that the number of actual caching strategies is smaller than the number of hardware and OS combos (which it would seem likely to be, since at least some of those should share caching strategies), what is really needed is analysis of the determinants of algorithm performance that take into account different caching strategies. That's still difficult compared to simple ideal assumptions like constant memory access cost, but simple metrics that don't capture real world behavior well are of limited utility.
Re:Theoretical performance vs real-world performan
on
Knuth Got It Wrong
·
· Score: 5, Informative
Yes it's true. In some real-world applications an algorithm encounters it's worst case running time more than the predicted theoretical average case running time.
That's actually not the issue.
The issue is that in the real world, the assumptions underlying the calculation of algorithmic complexity may not hold. For instance, Knuth's analysis that the author of the article here holds to be misleading (not, as the Slashdot title suggests, "wrong") calculates the complexity based on the assumption of ideal random access memory, that is, memory for which all accesses are equal cost.
In real-world computers using caches (as pretty much all real-world computers do, often at many different levels) that assumption does not hold -- access to some parts of the memory address space are much more expensive than accesses to other parts of the address space, and which parts of the address space are expensive changes over time (and how it changes over time potentially depends on the caching strategy used at every level lower than the level at which the algorithm is implemented.)
This means that algorithms in the real world can scale worse than their theoretical "worst-case", if that theoretical worst-case scaling is based on the assumption of constant memory access cost, since that assumption does not hold in the real world.
Well yeah, but if it's illegal and you do it, you can be sued.
Wrong.
If it is illegal, you can be sued for doing it, whether or not you actually have done it.
A major purpose of the trial -- which comes after you have been sued -- is to determine whether or not you did what you were sued for doing. If there was away to assure you were guilty before you were sued, we wouldn't need trials.
I personally find it funny that while the Kinect hardware involves not touching anything, it's name is, by dictionary defintion, about two things touching each other.
You have a dictionary that defines "kinect"?
I mean, sure, you can view it as a radical misspelling of "connect".
Or it could be shorthand for being separated from a conjoined twin (a "kin-ectomy").
Or, as is actually the case, it could be a non-word designed to serve not as a description, but as a trademark (non-words having no definition, and making, in some respects, "better" trademarks) that suggests "connect" and "kinetic", without either being part of any definition.
To the best of my knowledge, there has not been any new breakthroughs in television technology, so a new console wouldn't provide anything in the way of graphical advancement
1. "Graphical advancement" doesn't need changes in TV technology; better graphics hardware on the same TV (so same resolution and refresh) will produce better visual experience. GPUs have advanced since the current console generation was introduced.
2. 1080p120 and 1080p240 TVs (and 3D TV technologies) are new, and could be supported by a new console generation. No doubt, many people would point out that those have limited market penetration, but people pointed that out about HDTV when the current generation of consoles were introduced.
3. Graphical advancement isn't the only way consoles could advance. Better onboard networking, better CPU, etc.
In most other industries this would be considered illegal as a clear conflict of interest.
Outside of some areas of government work and a handful of tightly-regulated industries, "clear conflicts of interest" aren't illegal, and are, in fact, fairly common. Certain conflicts of interest may, while not illegal in and of themselves, be prohibted by particular contracts (particularly employment contracts), but most aren't even there (for instance, its a pretty clear conflict of interest to work for one company and to own stock in a competitor -- if its voting stock, there is a double conflict of interest -- but except in the case of an executive-level employee, this would rarely be prohibited.)
Because if H-1B's didn't lower labor costs by undercutting the cost of hiring locally, there'd be no economic incentive to use them, and therefore no rational profit-maximizing corporation would ever hire anyone on an H-1B visa.
H1-Bs have the same cost of living you do - you're competing on an even playing field.
Presuming, of course, that you are willing to accept the same lifestyle: H-1B's from with a lower standard of living than the US (i.e., almost every other country on Earth) are naturally willing to accept, as an improvement, a standard of living lower than an equivalently educated American would tend to expect, and, therefore, a lower salary.
The alternative is the same person working from India.
No, its not. Because if that would be as cheap or cheaper, the profit-maximizing firm that hired the H-1B worker would instead outsource to India. If the H-1B option was gone, a some of the work that goes to H-1Bs would go to overseas outsourcing, some of it would go to domestic workers, and some of it would go to different firms than are doing it now (some of them in the US with the same choices, some of them not.)
Every time a company brings a worker from India at those wages to here at these wages on an H1-B, I win.
Since, in either case, you aren't doing the work, I don't see how you "win" even if your premises were correct: work that you don't get because it is outsourced overseas and work that you don't get because it goes to someone on an H-1B (and, for completeness, work that you don't get because it goes to a different local worker) are all equivalent: they all represent work that you don't get.
Since when did USA invaded Bosnia? It was a NATO attack, there was no invasion and they just bombed the hell out of Serbia until they deposed Milosevic by themselves which effectively led to the independence of Bosnia (as well as other parts of ex-Yugoslavia).
You are confusing the 1999 NATO-Yugoslavia war over Kosovo with the 1995 execution of Operation Deliberate Force against Bosnian Serb forces in Bosnia.
The independence of Bosnia was established before the NATO-Yugoslavia war began.
Saudi Arabia, Iraq, the UAE, Kuwait, etc. would disagree. Seems they made a good amount of money off their natural mineral resources.
Yes, because they held a particular resource which has both very valuable, found in quantities sufficient to meet domestic needs and provide a surplus for export in a small number of countries, and because that resource is one which most of the exporting countries were effectively able to form a cartel around.
If you look at any other resource extraction scenario, it usually is pretty dismal for the producer.
In each of your examples - Brazil, Nigeria, Angola - the wealthy ruling class is wealthy (fabulously so) because they have deemed that the benefits from the sales of natural resources should flow not to the country in general, but to the rulers in particular.
Actually, that's even more true in the your examples (except Kuwait and the UAE, and perhaps even there, or at least Kuwait, if you consider the large number of non-citizen foreign workers.)
But immense wealth was made for each of these countries (and is still being made today) from their mining and mineral resources.
Not as immense as you might think, which is why the average wealth (to which the distribution between the elites and the lowly is irrelevant) in the nation's in GPs list is fairly low.
A representative democracy is one where individual decisions are made by representatives of the people, as opposed to a direct democracy (see some ancient Greek city states and modern Switzerland) where decisions are made directly by the people.
IIRC, modern Switzerland is (like many US states) a mostly-representative democracy where certain decisions are made through direct democracy, not a direct democracy on, say, the Athenian model.
The so-called "gross debt" does, but the gross debt includes intragovernmental debt (money that the federal government owes itself), and so isn't the real number of interest, which is the public debt, which is somewhere in the neighborhood of 60-some percent of GDP.
The US government will *never* be able to retire this debt, not in the lifetime of anyone alive today.
Not being able to retire the debt is pretty much irrelevant. There's no reason to put much emphasis on doing that. Ideally, in the long-term (and in the short-term during expansions) you want the debt, or perhaps more specifically the debt service expense, to decline in proportion to GDP (in the short-term, during recessions, it makes sense for it to rise in proportion to GDP).
There may be real concerns from the debt:GDP ratio, but not being able to retire the debt is pretty much irrelevant. Retiring the debt -- or even paying it down in absolute terms -- isn't a particularly high priority.
Apple IS trying to control the flow of their customers personal data. They are preventing analytics...
No, they aren't preventing analytics. They are restricting analytics based on whether the analytics provider competes with Apple outside of the advertising market. So, yes, they are trying to control the flow of their customers personal data, but not to protect the customers privacy, only to assure that it goes to a place that doesn't give money to people competing with Apple in the mobile hardware, mobile OS, or development environment markets.
Trying to paint this is as good for the end-user or some kind of privacy benefit is ridiculously naive.
This type of trading provides liquidity and makes a market.
Sure, it "makes a market" -- in that it increases the probability that someone will be available willing to buy at some price when someone is willing to sell, and vice versa -- but it does so by divorcing that market from the thing that make markets efficient at realizing utility, to wit, the proximity between the market actors and the costs of production and/or benefits of direct use of the things being bought and sold.
Natural liquidity in a market may have particular benefits, but mindlessly chasing liquidity as an unqualified good is idiotic.
According to the Constitution there are rights we cannot be forced to give up because they were not given to us by men.
Wrong. The Constitution doesn't make claims about the origin of rights, nor does prevent any of them from being taken away through amendment (except the "right" to participate in the slave trade, and that only for a set period of years after the Constitution was adopted.)
You are probably confusing the Constitution with the Declaration of Independence.
Which is why I a quite frustrated when I first saw iGoogle and thought it was their new look. They should have tested this on iGoogle;
iGoogle (and Google Calendar and some other Google properties) have had themes including (but not limited to) background images for quite some time, so insofar as it makes sense to "test" a non-iGoogle background-image function on iGoogle, they already did.
Here's a chrome plugin that does kills the fade-in in a narrowly targeted way:
Of course, in Chrome the address bar by default works almost exactly like the search box on the Google homepage, so there's no real need to use the Google homepage on Chrome.
And what role is Google serving if they spend two years hosting the patents in their database while the USPTO spends 24 months looking for a contractor who will presumably charge money to do the same thing?
Well, I dunno, but if I was going to guess who was going to be in a position to offer the lowest bid to USPTO, I'd think it would be the one that had already sunk the costs for much of the work, and who thus would most likely have the least additional cost if they got the contract.
Looking at the "striking examples" I have to ask, does anyone other than 14-year-old gamers build computers anymore? Whatever happened to six flat sides?
If you check any place that sells cases (newegg, etc.) you'll find plenty, for just about any motherboard size.
But that's not the point. This article was pretty clearly talking about a situation where a 2 car family is trying to figure out which of their 2 cars is going to be the most beneficial to upgrade. For that to happen, you are going to need to know the figures for your existing vehicle.
Yes, and the figures you are going to need to know are the actual fuel costs of your current vehicles and the expected fuel costs of the new vehicles. Neither gallons per mile nor the existing label gets you there without a bunch of other calculations, though if you have good records of actual costs and mileage on the existing cars, it will be a step or two shorter if you start with gallons/mi on the label, true.
OTOH, for lots of other common comparison use case, especially for comparing new cars against each other, the est. $/year is more useful. (And, for people who don't keep extensive records of their own, the est. $/year would continue to be more useful if the EPA also kept -- say, on the web -- the ratings for older vehicles updated to the assumptions used for newer vehicles.)
Gallons/mile is only useful superior to the est. $/year -- or even mpg -- for meaningful comparisons when you use other information and do calculations based on it, not as a direct comparator. And if you are willing to do that kind of calculation, its a simple calculation to derive gallons per mile from mpg.
LOL...yeah, right. You are talking about the same people who can NOT figure out whether 10->20 will save you more than 33->50, and now you expect that they are going to be able to make this calculation on their own?
If they can't do the calculation on their own, they can't make a meaningful comparison in the precise use case you refer to unless you further assume that the usage patterns for the two cars are identical, which is typically not the case in two-car families. After all, fuel economy doesn't tell you which saves more gas without considering the usage pattern.
Slashdot Mirror
A touch screen input system may be better for creation tasks if it has appropriate customizations for each task. A hardware keyboard is ergonomically better and provides better feel and feedback, so a general "touch screen keyboard" that copies the limitation of a hardware keyboard (same layout for all tasks) won't be very good. But something that adapts to tasks may be very good for "creation" tasks (particularly those that aren't primarily about creation of text, for which typical hardware keyboards are fairly well optimized.)
I agree. In GP I was clarifying what the issue was, not defending the idea in the TFS headline that "Knuth was wrong". A more accurate (though less headline-worthy) title would be "A naive reading of Knuth may be more misleading than most people are aware."
That still doesn't make the assumption of constant memory cost valid, even in the asymptotic case, unless all cache misses are created equal such that cache misses act like direct random access, and the algorithm asymptotically approaches a fixed cache miss rate.
Since hard drives aren't random access devices, anyway, even ignoring cacheing, this is unlikely.
Right. And if that is true based on the assumption that all memory accessed by the system has an equal and constant access costs, but in the real world (because of cache) memory does not have equal and constant access cost, an algorithm which has O(n) theoretical performance under the assumption of equal and constant memory access cost may not scale even approximately linearly on a real world system -- it may scale quadratically or even worse, depending on it smemory access pattern and the structure of the underlying cache layers.
(Assuming that consistent cache misses result in constant time [albeit slow] access, the big-O of the algorithm would still be a valid asymptotic limit, though misleading in terms of real-world performance, but I'm not sure that assumption is valid of real world caches. The big point is that any a priori performance analysis rests on some set of assumptions about the characteristics of a system, and says nothing about the performance of a system that does not share those characteristics.)
That's nonsense unless you mean "computational model" in a sense that is irrelevant to the immediate issue. If all you know is that an algorithm that has an asymptotic running time that is O(n) under the assumption of constant-cost memory access, you don't know the order of the asymptotic limit of its running time on a system that doesn't have constant-cost memory access. Under a particular scheme of non-constant-cost memory access, it might be O(n), it might be O(n log n), it might be O(n^n), it might be something else.
If the assumptions under which the performance scaling were calculated don't apply to your situation, neither does the calculation.
Read speed of disc drives has also improved in the past five years. Also, its quite possible a new generation of console could use a primary game media other than discs.
Sure, if you are looking in magazines. If you are looking at in-store demos (which are also marketing material) you see the graphics, sure, but also the actual gameplay. Other improvements besides in the output resolution can impact the quality of the gameplay.
Assuming that the number of actual caching strategies is smaller than the number of hardware and OS combos (which it would seem likely to be, since at least some of those should share caching strategies), what is really needed is analysis of the determinants of algorithm performance that take into account different caching strategies. That's still difficult compared to simple ideal assumptions like constant memory access cost, but simple metrics that don't capture real world behavior well are of limited utility.
That's actually not the issue.
The issue is that in the real world, the assumptions underlying the calculation of algorithmic complexity may not hold. For instance, Knuth's analysis that the author of the article here holds to be misleading (not, as the Slashdot title suggests, "wrong") calculates the complexity based on the assumption of ideal random access memory, that is, memory for which all accesses are equal cost.
In real-world computers using caches (as pretty much all real-world computers do, often at many different levels) that assumption does not hold -- access to some parts of the memory address space are much more expensive than accesses to other parts of the address space, and which parts of the address space are expensive changes over time (and how it changes over time potentially depends on the caching strategy used at every level lower than the level at which the algorithm is implemented.)
This means that algorithms in the real world can scale worse than their theoretical "worst-case", if that theoretical worst-case scaling is based on the assumption of constant memory access cost, since that assumption does not hold in the real world.
Wrong.
If it is illegal, you can be sued for doing it, whether or not you actually have done it.
A major purpose of the trial -- which comes after you have been sued -- is to determine whether or not you did what you were sued for doing. If there was away to assure you were guilty before you were sued, we wouldn't need trials.
Wrong.
Not doing something doesn't stop you from getting accused of doing it, and a lawsuit is, after all, just a very formal accusation.
Specifically relevant to copyright and P2P, I suggest you familiarize yourself with the history of Media Sentry and Media Defender.
You have a dictionary that defines "kinect"?
I mean, sure, you can view it as a radical misspelling of "connect".
Or it could be shorthand for being separated from a conjoined twin (a "kin-ectomy").
Or, as is actually the case, it could be a non-word designed to serve not as a description, but as a trademark (non-words having no definition, and making, in some respects, "better" trademarks) that suggests "connect" and "kinetic", without either being part of any definition.
1. "Graphical advancement" doesn't need changes in TV technology; better graphics hardware on the same TV (so same resolution and refresh) will produce better visual experience. GPUs have advanced since the current console generation was introduced.
2. 1080p120 and 1080p240 TVs (and 3D TV technologies) are new, and could be supported by a new console generation. No doubt, many people would point out that those have limited market penetration, but people pointed that out about HDTV when the current generation of consoles were introduced.
3. Graphical advancement isn't the only way consoles could advance. Better onboard networking, better CPU, etc.
Outside of some areas of government work and a handful of tightly-regulated industries, "clear conflicts of interest" aren't illegal, and are, in fact, fairly common. Certain conflicts of interest may, while not illegal in and of themselves, be prohibted by particular contracts (particularly employment contracts), but most aren't even there (for instance, its a pretty clear conflict of interest to work for one company and to own stock in a competitor -- if its voting stock, there is a double conflict of interest -- but except in the case of an executive-level employee, this would rarely be prohibited.)
Because if H-1B's didn't lower labor costs by undercutting the cost of hiring locally, there'd be no economic incentive to use them, and therefore no rational profit-maximizing corporation would ever hire anyone on an H-1B visa.
Presuming, of course, that you are willing to accept the same lifestyle: H-1B's from with a lower standard of living than the US (i.e., almost every other country on Earth) are naturally willing to accept, as an improvement, a standard of living lower than an equivalently educated American would tend to expect, and, therefore, a lower salary.
No, its not. Because if that would be as cheap or cheaper, the profit-maximizing firm that hired the H-1B worker would instead outsource to India. If the H-1B option was gone, a some of the work that goes to H-1Bs would go to overseas outsourcing, some of it would go to domestic workers, and some of it would go to different firms than are doing it now (some of them in the US with the same choices, some of them not.)
Since, in either case, you aren't doing the work, I don't see how you "win" even if your premises were correct: work that you don't get because it is outsourced overseas and work that you don't get because it goes to someone on an H-1B (and, for completeness, work that you don't get because it goes to a different local worker) are all equivalent: they all represent work that you don't get.
You are confusing the 1999 NATO-Yugoslavia war over Kosovo with the 1995 execution of Operation Deliberate Force against Bosnian Serb forces in Bosnia.
The independence of Bosnia was established before the NATO-Yugoslavia war began.
Yes, because they held a particular resource which has both very valuable, found in quantities sufficient to meet domestic needs and provide a surplus for export in a small number of countries, and because that resource is one which most of the exporting countries were effectively able to form a cartel around.
If you look at any other resource extraction scenario, it usually is pretty dismal for the producer.
Actually, that's even more true in the your examples (except Kuwait and the UAE, and perhaps even there, or at least Kuwait, if you consider the large number of non-citizen foreign workers.)
Not as immense as you might think, which is why the average wealth (to which the distribution between the elites and the lowly is irrelevant) in the nation's in GPs list is fairly low.
IIRC, modern Switzerland is (like many US states) a mostly-representative democracy where certain decisions are made through direct democracy, not a direct democracy on, say, the Athenian model.
The so-called "gross debt" does, but the gross debt includes intragovernmental debt (money that the federal government owes itself), and so isn't the real number of interest, which is the public debt, which is somewhere in the neighborhood of 60-some percent of GDP.
Not being able to retire the debt is pretty much irrelevant. There's no reason to put much emphasis on doing that. Ideally, in the long-term (and in the short-term during expansions) you want the debt, or perhaps more specifically the debt service expense, to decline in proportion to GDP (in the short-term, during recessions, it makes sense for it to rise in proportion to GDP).
There may be real concerns from the debt:GDP ratio, but not being able to retire the debt is pretty much irrelevant. Retiring the debt -- or even paying it down in absolute terms -- isn't a particularly high priority.
No, they aren't preventing analytics. They are restricting analytics based on whether the analytics provider competes with Apple outside of the advertising market. So, yes, they are trying to control the flow of their customers personal data, but not to protect the customers privacy, only to assure that it goes to a place that doesn't give money to people competing with Apple in the mobile hardware, mobile OS, or development environment markets.
Trying to paint this is as good for the end-user or some kind of privacy benefit is ridiculously naive.
Sure, it "makes a market" -- in that it increases the probability that someone will be available willing to buy at some price when someone is willing to sell, and vice versa -- but it does so by divorcing that market from the thing that make markets efficient at realizing utility, to wit, the proximity between the market actors and the costs of production and/or benefits of direct use of the things being bought and sold.
Natural liquidity in a market may have particular benefits, but mindlessly chasing liquidity as an unqualified good is idiotic.
Wrong. The Constitution doesn't make claims about the origin of rights, nor does prevent any of them from being taken away through amendment (except the "right" to participate in the slave trade, and that only for a set period of years after the Constitution was adopted.)
You are probably confusing the Constitution with the Declaration of Independence.
iGoogle (and Google Calendar and some other Google properties) have had themes including (but not limited to) background images for quite some time, so insofar as it makes sense to "test" a non-iGoogle background-image function on iGoogle, they already did.
Of course, in Chrome the address bar by default works almost exactly like the search box on the Google homepage, so there's no real need to use the Google homepage on Chrome.
Well, I dunno, but if I was going to guess who was going to be in a position to offer the lowest bid to USPTO, I'd think it would be the one that had already sunk the costs for much of the work, and who thus would most likely have the least additional cost if they got the contract.
If you check any place that sells cases (newegg, etc.) you'll find plenty, for just about any motherboard size.
Yes, and the figures you are going to need to know are the actual fuel costs of your current vehicles and the expected fuel costs of the new vehicles. Neither gallons per mile nor the existing label gets you there without a bunch of other calculations, though if you have good records of actual costs and mileage on the existing cars, it will be a step or two shorter if you start with gallons/mi on the label, true.
OTOH, for lots of other common comparison use case, especially for comparing new cars against each other, the est. $/year is more useful. (And, for people who don't keep extensive records of their own, the est. $/year would continue to be more useful if the EPA also kept -- say, on the web -- the ratings for older vehicles updated to the assumptions used for newer vehicles.)
Gallons/mile is only useful superior to the est. $/year -- or even mpg -- for meaningful comparisons when you use other information and do calculations based on it, not as a direct comparator. And if you are willing to do that kind of calculation, its a simple calculation to derive gallons per mile from mpg.
If they can't do the calculation on their own, they can't make a meaningful comparison in the precise use case you refer to unless you further assume that the usage patterns for the two cars are identical, which is typically not the case in two-car families. After all, fuel economy doesn't tell you which saves more gas without considering the usage pattern.