Although seriously questionable legally, if the authors guild was able to prove that Text-To-Speech of copyright books was copyright infringement then that would be absolutely huge.
Other people have addressed the disability angle. I'll add one thing: I bet you that this dispute isn't about the law on text-to-speech, but rather, about the nature of Amazon's agreements with authors. I.e., the issue most likely isn't that copyright law forbids TTS being used with copyrighted works, but rather, that Amazone signed some piece of paper that implicitly restricts them from competing with audiobook editions of many of their Kindle books.
You are wrong about the details, but let me focus on this part of your comment, where you are wrong about the generals.
Ones and zeros are ones and zeros. The problem is that sometimes your sensor says "one" when it should say zero, and vice-versa. This is why the data in the discs is stored using error correcting codes, in order to keep the those errors at a low rate, and to be able to detect them when they do happen.
But read errors will still happen at some rate, and there are two general strategies for dealing with them:
Retry erroneous reads until the data is read successfully;
Resign yourself to lossy reads, and degrade gracefully when they happen.
General-purpose computer storage devices must take the first strategy, because computers can't work at all without bit-perfect reads. CD Audio, on the other hand, was designed to allow lossy readers that degrade gracefully. A CD player that hits a small enough read error can guess a sample value from the neighboring ones and keep going.
I don't know about you, but neither my Windows CD nor my linux CD have any problems reading with lossless precision.
At the most primitive level, any attempt to read a bit from a disc is subject to random errors at some rate. Just the bare fact that data is digitally encoded doesn't automatically make that go away. What digital encoding allows you to do, however, is to store checksums along with the data. The checksums, in turn, allow you to demonstrate, to a very high degree of confidence, whether you've read the data accurately. This means that the way you get a bit-perfect read from digital media is by verifying the checksums of all your raw reads, and redoing the ones that fail until you get them right.
CD-ROM devices are required to make bit-perfect reads of the computer data stored on the disc. Audio CD players, on the other hand, are not required to do so. Playing audio successfully consists of producing an analog output signal, and in real time. This means that read errors can be much more easily tolerated, especially if correcting them would delay the music reproduction.
Or, in other terms: audio CD players are allowed to perform lossy reads of audio CDs, because it makes them cheaper to build, and they're going to output analog anyway.
Audio CD players do not usually perform bit-perfect reads of the discs. They have plenty of latitude to ignore small read errors, since the listeners are very unlikely to notice small errors. What this new technology allegedly does is press CDs that are easier for the players to read; i.e., reduce the rate of read errors with improved discs.
This is a way to make sure your data writes work better.
Um, no, this doesn't seem to have to do anything with consumer CD writers.
Here's my guess at the real story: Sony figured out a way to use Blu-Ray pressing infrastructure to press CDs too, making it unnecessary to have CD-specific machinery at their plants. This lowers their manufacturing costs for CDs, and as a mostly unimportant side benefit, the resulting CDs are easier to read accurately by existing CD players. Some PR hack at Sony got hold of this info, and decided to turn it into a silly marketing campaign.
Actually, this sort of thing would most likely make CDs and Blu-Ray disks cheaper to press. If the same machines can press either of them, that allows for economies of scale that aren't possible with different machines for each.
Not that I expect CDs to actually become cheaper because of this, of course.
Fun fact: not all countries do so. Several countries in Asia have responded to the problem of parents aborting female fetuses by limiting doctors' ability to inform their patients of the embryo's sex, and prohibiting people from aborting on the basis of the embryo's sex.
This is where you go astray. If Amazon doesn't like the higher rates, they should stop purchasing the rights. If they stop purchasing, the publishers will have to reduce rates to the point where they start purchasing again. This way, the rights are priced according to their actual value.
Or e-books are not made at all, because nobody is willing to compensate the authors at the rate that the authors would minimally like to be compensated for e-books.
In no case should Amazon be forced to stop offering text-to-speech capability.
I'm sympathetic to TTS features in general, but I think your statement here is too strong. I can think of one exception that is perfectly fair: Amazon should not offer TTS in Kindle if its agreements with the content providers forbid it from doing so. If the content providers can successfully argue that the terms of their agreements with Amazon forbid the latter from offering TTS, then Amazon is in the wrong, period.
Part of the problem here for Amazon is that they are both the licensee and the TTS provider. This wouldn't happen if the licensee and the TTS provider were separate parties, assuming then that the TTS provider wasn't be bound by the licensee's terms. So, for example, if there was a competitive market in e-book readers that could all use Amazon's Kindle format, then those reader manufacturers would be absolutely free to provide such a feature.
But since audo books and e-books have different roylaty rates, if you change their ratio then you chance the total earnings to the authors. thus you need to re-adjust the roylaty rates so that the authors get the same total earnings.
I was just making a similar point in another comment, so I have to agree here. It really seems to me that there is no real argument about whether it is permissible to use text-to-speech (TTS) technology to have an e-book read out loud to the user; even TFA, confused as it is, seems to acknowledge this:
"True, you can already get software that will read aloud whatever is on your computer. But Kindle 2 is being sold specifically as a new, improved, multimedia version of books -- every title is an e-book and an audio book rolled into one. And whereas e-books have yet to win mainstream enthusiasm, audio books are a billion-dollar market, and growing. Audio rights are not generally packaged with e-book rights. They are more valuable than e-book rights. Income from audio books helps not inconsiderably to keep authors, and publishers, afloat."
So it really looks to me like they might have some legitimate complaint about whether Amazon has violated the terms of its agreements with the publishers by (a) making their proprietary e-book reader capable of reading their proprietary format e-books out loud, and (b) marketing it as an alternative to audio books. But this seems like a short-to-medium term dispute between Amazon and the copyright holders, not something that can stop TTS-capable e-book readers from flooding the market in the long run. And as you point out, well, the authors are going to have to figure out how to do new kinds of licensing deals for e-book editions in this new world.
Why? Copyright holders receive royalties on audiobooks because audiobooks are a derivative work. That makes sense. TTS technology is not a derivative work. It allows you to create a derivative work, but so do a pair of scissors.
But in the case of Kindle, Amazon is combining a proprietary e-book format with a proprietary device that both displays the books as text, and reads them aloud. They cannot invoke your argument as easily as a third-party TTS provider would be able to.
I think the real issue here isn't whether it's unlawful to use TTS technology to read e-books out loud; I've read TFA and I understand that the author actually agrees it's not. The real issue seems to be a relatively minor one about whether Amazon violated some agreement with its publishers by combining e-books and TTS.
Basically, it looks like the right to make print editions of books is priced under the assumption of additional revenue from audio editions, and that the publishers licensed their work to Amazon under terms and prices similar to those they do for print editions, with the assumption that the Kindle editions would no more cut into their audiobook revenues than print editions could. Clearly this wasn't their brightest moment. If TTS on e-book readers becomes ubiquitous (as I expect it will), well, then they'll have to learn license and price e-book rights better.
There are all sorts of stupid "danger" indicators in all sorts of damn devices, and interaction designers have been telling us for years how worthless such indicators normally are. Airplane accidents is a field rife with fatal examples. Basically, a user that's concentrated in achieving a broader task that involves making a phone call will tend to have their locus of attention focused on things other than the roaming indicator. (And don't get me started on receiving important calls with a hands-free device while driving near an international border.)
Basically, $28,000 roaming bills are potential big inadvertent consequence of using the phone in the normal, habitual manner. A real solution to the problem has to make it impossible for such use to result in such bills, period. Such an extraordinarily large bill should only be possible through performing a sequence of proportionately extraordinary actions, in a context that makes the billing consequences clear.
If the company spends a fortune then tries to charge you for costs and a percentage profit, then the argument works.
Only if they can actually collect. Maybe you will agree with my point if I modify it to say the following: a company should not incur the costs of providing a customer a service unless they are reasonably sure that they will actually collect on the bill.
I will say the following thing, though: a company that opaquely encourages its customers to inadvertently rack up really big bills that it reasonably expects a large number of them not to be able to pay, and prices the services to make a profit after the expected number of no-pays, is a company that is really in the business of providing involuntary predatory loans. Consumers should be protected from practices like that.
In the case in this article, basically, the consumer needs to be protected from companies like the operator of the in-ship phone service, who could have taken some very simple steps to prevent this from happening. Simplest, probably most effective example: not route calls or data for anybody's phone unless the phone's user has requested access to the in-ship phone service.
Why is it the responsibility of the provider to monitor how much you use your cell phone?
This is not about responsibility. It's about the phone company's self-interest. It is in the self-interest of the phone company to make it impossible for people to inadvertently rack up big bills that they cannot possibly pay.
If today I incur the expenses to provide you a service, and then bill you for it at the end of the month, what I'm doing is in an important regard equivalent to extending you short-term credit. It is absolutely in my interest to only extend such implicit credit when I can be reasonably sure I will be paid. Therefore, I should not allow people to rack up $28,000 bills unless I am reasonably sure that they will indeed pay them up.
There are several problems in this one specific case:
Third-party billing: the phone company is required to bill the customer for charges assessed by a third party, over whom they have hardly any control. This is broken at many levels, not least of which is the legal level. (Not just when it comes to roaming, but also cramming.)
The devices make it way too easy for the customer to inadvertently rack up enormous, unintended charges. The UI cue in nearly all phones is the network name display, which is just too subtle for most people.
A superior system would require the phone to query the current network for its billing rates, to be aware of the user's usual rates, and intelligently notify the user before they attempt to place calls that would incur in higher than normal charges. The notification should spell out as clearly and concretely as possible how large the charges could be--e.g., "A 10 minute phone call at this rate would cost $1,234.56," or "displaying a typical newspaper web page would cost $5 at this rate."
Please allow me to stress the following point again: this is all in the interest of the provider.
Funny you would glorify one of the most retarded things about java. The sane thing to do is to have the assignment operator do the same thing whatever the type. Either always by value, or always by reference. DO NOT make me have to find out if the types of a and b are primitive or not each time I see a = b.
Java's primitive-vs-object distinction is indeed annoying, but it really doesn't matter in Java. It behaves exactly as it would behave if each primitive variable was a reference to an interned, immutable object. In fact, a better implementation of Java that eliminated the primitive/object distinction from the language semantics would be best served by stipulating that the objects of the primitive type are always immutable, and that any two objects of primitive type that are equal by content are always also equal by reference. Then the compiler could transparently remove the reference indirection.
Because major multinational Joe-Blow, Inc., is not comfortable buying joe-blow.tld1 and letting whoever wants to buy joe-blow.tld2, which they could use in a way that damages Joe-Blow's reputation. There is some legal recourse to that, depending on the specifics, but that takes time and is anything but guaranteed. Why risk it? So most large companies (and indeed, many smaller companies) just buy a bunch of permutations of their domain name, including different TLDs and common misspellings.
But I'd argue that this approach is only feasible given a limited number of TLDs in the first place. If gTLDs mean they can't corner the market on registrations of names related to their company name, why would they try?
As you point out yourself, they already can't do it, given all the ccTLDs in existence. Doesn't that just go to show that how many domains Joe-Blow Inc. registers isn't determined by how many TLDs there are?
Right, so then why not just issue a separate serial number that identifies the host and link based on that? Why bother allocating names in a hierarchical system if not in order to organize?
Because the serial numbers scheme would require a centralized authority to issue the numbers. If you wanted a new hostname, you'd have to submit a request to this centralized authority, and wait until it was processed. Under DNS, if you are the holder of the example.com domain, you have the authority to assign names within example.com, and to delegate the authority to assign names within subdomains thereof. When you want to add a new hostname within example.com, you don't need to submit a request to some overworked office that takes a week to process it; you do it yourself at your DNS servers. When the folks you've given a subdomain to need to add a hostname within that subdomain, they also do it at their DNS servers, they don't need to bug you.
And then there's the issue of who answers queries for the mapping of the serial numbers to IP addresses. If we wanted to have a distributed system for answering queries about which IP each serial number points to, we'd need a system largely equivalent to DNS anyway, except more complicated, because you wouldn't be able to tell anything about a serial number's assignee without examining a database that contains all of the serial numbers. Compare this to DNS, where the hierarchical nature of FQDNs means that each domain server only needs to store a subset of the assignments. Each prefix of an FQDN determines set of DNS servers that can answer questions about the rest of the name. This is what allows the system to get away with no server in the system needing to store anything like the full table of assignments.
This just goes to demonstrate the point I started this thread with: the point of the hierarchy of domain names is to delegate the authority to assign stable hostnames and respond to queries about assignments. A stable host identifiers system that used purely numeric, meaningless identifiers would have a exact same technical reasons for giving the same sort of hierarchical structure to its numeric host identifiers, so that a stable host ID would be a sequence of numbers, each associated with an authority in the context of the preceding ones.
If you're not going to use DNS to find and identify anything and instead are only going to use Google, then why bother at all? Why not just stick to IPs? Or if you need a dynamic way of identifying things independently of IPs, why not just issue some other serial number?
I'm not questioning the value of DNS as a layer of abstraction between high-level protocols like HTTP from low-level protocols like IP and TCP. The ability to change what IP address a request to a symbolic hostname gets sent to is of certainly valuable, because a host's IP address is determined by details about routing that are not relevant to a protocol like HTTP. Using symbolic hostnames allows you to make hosting or routing changes that invalidate a host's IP address without breaking existing references to that host.
I still don't think DNS is a good content addressing scheme for non-technical end users.
What? This is adding more structure, if done right. Which imparts more structure, to group all animals in a flat bucket called 'animals' (.com) or to pull them out into 'sponges', 'worms', 'molluscs', 'insects', 'chordates', etc (new TLD's)?
But what if I wanted to classify animals by color, instead of taxon?
The big problem with centralized hierarchical organizational schemas is that the same domain of things can be classified orthogonally according to different criteria, and different classifications are appropriate for different tasks. So you will never be able to choose a "correct" one true classification.
Granted, it's not a multi-level hierarchy, but then again, when was the last time you used the old Yahoo! directory? Ontological organization has been voted off the Internet, for better or worse.
I think you're misunderstanding what I'm getting at. The reason we have different TLDs certainly is about organization as well as delegation of authority. If it was only about authority, why bother having.net,.org,.com,.info, and.biz addresses?
But my larger point is that that use of TLDs is wrong. You will not solve the organization problems with DNS, simply because there is no one universal scheme of classification that you can apply to all sites in al contexts. Different users of the net will need different classification schemes, and there is no technical reason why DNS should impose one.
The reason we have those TLDs is because of an attempt to make DNS serve the function I'm arguing it should not serve.
Schools are supposed to have EDU addresses, government organizations are supposed to have GOV. I believe that ORG addresses were supposed to be non-profits, COM were supposed to be for commercial entities, and NET addresses were supposed to be mainly for ISPs and companies providing network services. And then countries were supposed to have their own addresses for sites specific to their country. The idea was partially so that you'd know the type of organization you were connecting to by its name.
True, but again: why should DNS encode that as part of the domain name? If you want to know what kind of organization you're connecting to, why not just search the web for information about them? Or, why not consult some sort of trusted, non-DNS database that catalogues organizations of that kind?
Examples of tools we already have that do this sort of job better than DNS: (a) Google, (b) the Better Business Bureau online database, (c) SSL certificate authorities.
So I'm just thinking they should go back to the drawing board and come up with a TLD structure that makes sense. Or maybe even reevaluate the whole DNS idea, because the TLD issue isn't the only, or even the biggest, problem with DNS.
I don't think the TLD structure needs to make any particular amount of sense. Organizing the net with DNS is just using the wrong tool to solve the problems.
I'm not sure that's a sign that we have too many, but perhaps rather that the attempt at organizing these things has been a failure so far. Personally, I'd like to see the whole thing reevaluated, figuring out what the goals of the organization are, and then figuring out what organization is likely to achieve those goals based on the reality of the Internet today (i.e. spammers, phishing, fraud, domain squatting).
But the point of DNS isn't to "organize" the content of the net. The point of DNS is to delegate and distribute authority for maintaining the records of a database of domain names to IP addresses, and answering queries about the records. If I have example.com, that means I can assign mappings within example.com, or delegate authority to assign mappings within a subdomain to some other party. That's all it is.
If we want to organize the content of the internet, we have better tools for doing so, like search engines or directories.
You're failing to ask a crucial question here: why do we insist in continuing to use domain names as an end-user content addressing mechanism? Why not use web directories, search engines, bookmarks and portals instead?
There are technical reasons for DNS as part of the low-level plumbing of the internet, certainly, but are there really any good reasons for a non-technical user to be aware of it?
It is a disgusting thing to do to the Internet, to remove the last semblance of hierarchy and structure in the naming system.
This claim makes no sense. DNS is just as hierarchical under the traditional and the gTLD model.
The hierarchy of domain names is really about delegation of authority for assigning IP addresses for symbolic names. The folks who manage the.edu domain have the authority to assign symbolic hostnames that end in.edu, and additionally, to delegate assignments in subdomains thereof to other parties, like the folks who manage the mit.edu and stanford.edu domains. And these latter folks have the same authority within those subdomains, and so on.
So, the TLDs are a relic of the early days of the Internet, where it was a network connecting computers at US military installations and universities, so very few top-level authorities were needed. Today, when the Internet is so large and pervasive, there is really no good argument for limiting the number of TLDs.
It will be a great stream of new revenue for registrars though.
Will it, really? I mean, if any organization can register a TLD and sell subdomains within it, that would drastically increase the supply of domain names. The prices of domains should go down, in that case, since if you can't get joe-blow.tld1 you could get joe-blow.tld2.
Of course, this is assuming that the domain registrars don't form a cartel. But the point is that generic TLDs aren't a big deal; a cartel of domain registrars is.
I think your argument relies on fixing the sense of the term "insurance" in a specific way that is convenient to your argument; in that regard, I think it begs the question. If you define "insurance" more abstractly, as something like "a scheme where most participants overpay for what they get, so that a minority of them underpay," then health expenditures based on taxes are a form of insurance.
I think the actual substance of the point, that compulsory, equal-pay insurance schemes are best implemented as government programs paid from taxes, is correct, though.
(Of course, when I say 'insurance companies should be able to set costs based whatever they want', I don't mean they should be exempt from anti-discrimination laws. Those are something entirely different.)
I don't think the two statements can be easily reconciled. If we can forbid the insurance companies from using race to set costs, why can't we forbid them from using other factors? (For example, prohibiting health insurance companies, in their current form in the USA, from using DNA test results as a cost factor.)
Just ask some neighbor kid to install your software for you, one that's too young to enter a legal agreement. Seems much more simple, and unlike this cat device, gives you plausible deniability to claim "I didn't even realize there was a EULA, let alone agree to it."
Let's grant that this works, for the sake of argument. The company is taking action against you in a court, you argue that you did not agree to the EULA on these grounds, and your argument is accepted. What comes next?
Well, this case is analogous to you taking some goods or service from somebody on the mistaken belief that they were giving it away for free. If all parties agree that you mistakenly and reasonably believed the goods or servie was being given away, all that shows is that you didn't steal (since you had no mens rea). You still did take something from them without their permission, and may still be liable for that. Probable outcome: you return the goods, or pay them for the service, and the matter is dropped.
So in the case where your EULA argument somehow worked, the best you could get away with is the vendor agreeing to drop the matter if you surrender all of your copies of their software.
Slashdot Mirror
Other people have addressed the disability angle. I'll add one thing: I bet you that this dispute isn't about the law on text-to-speech, but rather, about the nature of Amazon's agreements with authors. I.e., the issue most likely isn't that copyright law forbids TTS being used with copyrighted works, but rather, that Amazone signed some piece of paper that implicitly restricts them from competing with audiobook editions of many of their Kindle books.
You are wrong about the details, but let me focus on this part of your comment, where you are wrong about the generals.
Ones and zeros are ones and zeros. The problem is that sometimes your sensor says "one" when it should say zero, and vice-versa. This is why the data in the discs is stored using error correcting codes, in order to keep the those errors at a low rate, and to be able to detect them when they do happen.
But read errors will still happen at some rate, and there are two general strategies for dealing with them:
General-purpose computer storage devices must take the first strategy, because computers can't work at all without bit-perfect reads. CD Audio, on the other hand, was designed to allow lossy readers that degrade gracefully. A CD player that hits a small enough read error can guess a sample value from the neighboring ones and keep going.
At the most primitive level, any attempt to read a bit from a disc is subject to random errors at some rate. Just the bare fact that data is digitally encoded doesn't automatically make that go away. What digital encoding allows you to do, however, is to store checksums along with the data. The checksums, in turn, allow you to demonstrate, to a very high degree of confidence, whether you've read the data accurately. This means that the way you get a bit-perfect read from digital media is by verifying the checksums of all your raw reads, and redoing the ones that fail until you get them right.
CD-ROM devices are required to make bit-perfect reads of the computer data stored on the disc. Audio CD players, on the other hand, are not required to do so. Playing audio successfully consists of producing an analog output signal, and in real time. This means that read errors can be much more easily tolerated, especially if correcting them would delay the music reproduction.
Or, in other terms: audio CD players are allowed to perform lossy reads of audio CDs, because it makes them cheaper to build, and they're going to output analog anyway.
Audio CD players do not usually perform bit-perfect reads of the discs. They have plenty of latitude to ignore small read errors, since the listeners are very unlikely to notice small errors. What this new technology allegedly does is press CDs that are easier for the players to read; i.e., reduce the rate of read errors with improved discs.
Um, no, this doesn't seem to have to do anything with consumer CD writers.
Here's my guess at the real story: Sony figured out a way to use Blu-Ray pressing infrastructure to press CDs too, making it unnecessary to have CD-specific machinery at their plants. This lowers their manufacturing costs for CDs, and as a mostly unimportant side benefit, the resulting CDs are easier to read accurately by existing CD players. Some PR hack at Sony got hold of this info, and decided to turn it into a silly marketing campaign.
Actually, this sort of thing would most likely make CDs and Blu-Ray disks cheaper to press. If the same machines can press either of them, that allows for economies of scale that aren't possible with different machines for each.
Not that I expect CDs to actually become cheaper because of this, of course.
Fun fact: not all countries do so. Several countries in Asia have responded to the problem of parents aborting female fetuses by limiting doctors' ability to inform their patients of the embryo's sex, and prohibiting people from aborting on the basis of the embryo's sex.
Or e-books are not made at all, because nobody is willing to compensate the authors at the rate that the authors would minimally like to be compensated for e-books.
I'm sympathetic to TTS features in general, but I think your statement here is too strong. I can think of one exception that is perfectly fair: Amazon should not offer TTS in Kindle if its agreements with the content providers forbid it from doing so. If the content providers can successfully argue that the terms of their agreements with Amazon forbid the latter from offering TTS, then Amazon is in the wrong, period.
Part of the problem here for Amazon is that they are both the licensee and the TTS provider. This wouldn't happen if the licensee and the TTS provider were separate parties, assuming then that the TTS provider wasn't be bound by the licensee's terms. So, for example, if there was a competitive market in e-book readers that could all use Amazon's Kindle format, then those reader manufacturers would be absolutely free to provide such a feature.
I was just making a similar point in another comment, so I have to agree here. It really seems to me that there is no real argument about whether it is permissible to use text-to-speech (TTS) technology to have an e-book read out loud to the user; even TFA, confused as it is, seems to acknowledge this:
So it really looks to me like they might have some legitimate complaint about whether Amazon has violated the terms of its agreements with the publishers by (a) making their proprietary e-book reader capable of reading their proprietary format e-books out loud, and (b) marketing it as an alternative to audio books. But this seems like a short-to-medium term dispute between Amazon and the copyright holders, not something that can stop TTS-capable e-book readers from flooding the market in the long run. And as you point out, well, the authors are going to have to figure out how to do new kinds of licensing deals for e-book editions in this new world.
But in the case of Kindle, Amazon is combining a proprietary e-book format with a proprietary device that both displays the books as text, and reads them aloud. They cannot invoke your argument as easily as a third-party TTS provider would be able to.
I think the real issue here isn't whether it's unlawful to use TTS technology to read e-books out loud; I've read TFA and I understand that the author actually agrees it's not. The real issue seems to be a relatively minor one about whether Amazon violated some agreement with its publishers by combining e-books and TTS.
Basically, it looks like the right to make print editions of books is priced under the assumption of additional revenue from audio editions, and that the publishers licensed their work to Amazon under terms and prices similar to those they do for print editions, with the assumption that the Kindle editions would no more cut into their audiobook revenues than print editions could. Clearly this wasn't their brightest moment. If TTS on e-book readers becomes ubiquitous (as I expect it will), well, then they'll have to learn license and price e-book rights better.
There are all sorts of stupid "danger" indicators in all sorts of damn devices, and interaction designers have been telling us for years how worthless such indicators normally are. Airplane accidents is a field rife with fatal examples. Basically, a user that's concentrated in achieving a broader task that involves making a phone call will tend to have their locus of attention focused on things other than the roaming indicator. (And don't get me started on receiving important calls with a hands-free device while driving near an international border.)
Basically, $28,000 roaming bills are potential big inadvertent consequence of using the phone in the normal, habitual manner. A real solution to the problem has to make it impossible for such use to result in such bills, period. Such an extraordinarily large bill should only be possible through performing a sequence of proportionately extraordinary actions, in a context that makes the billing consequences clear.
Only if they can actually collect. Maybe you will agree with my point if I modify it to say the following: a company should not incur the costs of providing a customer a service unless they are reasonably sure that they will actually collect on the bill.
I will say the following thing, though: a company that opaquely encourages its customers to inadvertently rack up really big bills that it reasonably expects a large number of them not to be able to pay, and prices the services to make a profit after the expected number of no-pays, is a company that is really in the business of providing involuntary predatory loans. Consumers should be protected from practices like that.
In the case in this article, basically, the consumer needs to be protected from companies like the operator of the in-ship phone service, who could have taken some very simple steps to prevent this from happening. Simplest, probably most effective example: not route calls or data for anybody's phone unless the phone's user has requested access to the in-ship phone service.
This is not about responsibility. It's about the phone company's self-interest. It is in the self-interest of the phone company to make it impossible for people to inadvertently rack up big bills that they cannot possibly pay.
If today I incur the expenses to provide you a service, and then bill you for it at the end of the month, what I'm doing is in an important regard equivalent to extending you short-term credit. It is absolutely in my interest to only extend such implicit credit when I can be reasonably sure I will be paid. Therefore, I should not allow people to rack up $28,000 bills unless I am reasonably sure that they will indeed pay them up.
There are several problems in this one specific case:
A superior system would require the phone to query the current network for its billing rates, to be aware of the user's usual rates, and intelligently notify the user before they attempt to place calls that would incur in higher than normal charges. The notification should spell out as clearly and concretely as possible how large the charges could be--e.g., "A 10 minute phone call at this rate would cost $1,234.56," or "displaying a typical newspaper web page would cost $5 at this rate."
Please allow me to stress the following point again: this is all in the interest of the provider.
Java's primitive-vs-object distinction is indeed annoying, but it really doesn't matter in Java. It behaves exactly as it would behave if each primitive variable was a reference to an interned, immutable object. In fact, a better implementation of Java that eliminated the primitive/object distinction from the language semantics would be best served by stipulating that the objects of the primitive type are always immutable, and that any two objects of primitive type that are equal by content are always also equal by reference. Then the compiler could transparently remove the reference indirection.
But I'd argue that this approach is only feasible given a limited number of TLDs in the first place. If gTLDs mean they can't corner the market on registrations of names related to their company name, why would they try?
As you point out yourself, they already can't do it, given all the ccTLDs in existence. Doesn't that just go to show that how many domains Joe-Blow Inc. registers isn't determined by how many TLDs there are?
Because the serial numbers scheme would require a centralized authority to issue the numbers. If you wanted a new hostname, you'd have to submit a request to this centralized authority, and wait until it was processed. Under DNS, if you are the holder of the example.com domain, you have the authority to assign names within example.com, and to delegate the authority to assign names within subdomains thereof. When you want to add a new hostname within example.com, you don't need to submit a request to some overworked office that takes a week to process it; you do it yourself at your DNS servers. When the folks you've given a subdomain to need to add a hostname within that subdomain, they also do it at their DNS servers, they don't need to bug you.
And then there's the issue of who answers queries for the mapping of the serial numbers to IP addresses. If we wanted to have a distributed system for answering queries about which IP each serial number points to, we'd need a system largely equivalent to DNS anyway, except more complicated, because you wouldn't be able to tell anything about a serial number's assignee without examining a database that contains all of the serial numbers. Compare this to DNS, where the hierarchical nature of FQDNs means that each domain server only needs to store a subset of the assignments. Each prefix of an FQDN determines set of DNS servers that can answer questions about the rest of the name. This is what allows the system to get away with no server in the system needing to store anything like the full table of assignments.
This just goes to demonstrate the point I started this thread with: the point of the hierarchy of domain names is to delegate the authority to assign stable hostnames and respond to queries about assignments. A stable host identifiers system that used purely numeric, meaningless identifiers would have a exact same technical reasons for giving the same sort of hierarchical structure to its numeric host identifiers, so that a stable host ID would be a sequence of numbers, each associated with an authority in the context of the preceding ones.
I'm not questioning the value of DNS as a layer of abstraction between high-level protocols like HTTP from low-level protocols like IP and TCP. The ability to change what IP address a request to a symbolic hostname gets sent to is of certainly valuable, because a host's IP address is determined by details about routing that are not relevant to a protocol like HTTP. Using symbolic hostnames allows you to make hosting or routing changes that invalidate a host's IP address without breaking existing references to that host.
I still don't think DNS is a good content addressing scheme for non-technical end users.
But what if I wanted to classify animals by color, instead of taxon?
The big problem with centralized hierarchical organizational schemas is that the same domain of things can be classified orthogonally according to different criteria, and different classifications are appropriate for different tasks. So you will never be able to choose a "correct" one true classification.
Yup. For better, I'd insist.
But my larger point is that that use of TLDs is wrong. You will not solve the organization problems with DNS, simply because there is no one universal scheme of classification that you can apply to all sites in al contexts. Different users of the net will need different classification schemes, and there is no technical reason why DNS should impose one.
The reason we have those TLDs is because of an attempt to make DNS serve the function I'm arguing it should not serve.
True, but again: why should DNS encode that as part of the domain name? If you want to know what kind of organization you're connecting to, why not just search the web for information about them? Or, why not consult some sort of trusted, non-DNS database that catalogues organizations of that kind?
Examples of tools we already have that do this sort of job better than DNS: (a) Google, (b) the Better Business Bureau online database, (c) SSL certificate authorities.
I don't think the TLD structure needs to make any particular amount of sense. Organizing the net with DNS is just using the wrong tool to solve the problems.
But the point of DNS isn't to "organize" the content of the net. The point of DNS is to delegate and distribute authority for maintaining the records of a database of domain names to IP addresses, and answering queries about the records. If I have example.com, that means I can assign mappings within example.com, or delegate authority to assign mappings within a subdomain to some other party. That's all it is.
If we want to organize the content of the internet, we have better tools for doing so, like search engines or directories.
You're failing to ask a crucial question here: why do we insist in continuing to use domain names as an end-user content addressing mechanism? Why not use web directories, search engines, bookmarks and portals instead?
There are technical reasons for DNS as part of the low-level plumbing of the internet, certainly, but are there really any good reasons for a non-technical user to be aware of it?
This claim makes no sense. DNS is just as hierarchical under the traditional and the gTLD model.
The hierarchy of domain names is really about delegation of authority for assigning IP addresses for symbolic names. The folks who manage the .edu domain have the authority to assign symbolic hostnames that end in .edu, and additionally, to delegate assignments in subdomains thereof to other parties, like the folks who manage the mit.edu and stanford.edu domains. And these latter folks have the same authority within those subdomains, and so on.
So, the TLDs are a relic of the early days of the Internet, where it was a network connecting computers at US military installations and universities, so very few top-level authorities were needed. Today, when the Internet is so large and pervasive, there is really no good argument for limiting the number of TLDs.
Will it, really? I mean, if any organization can register a TLD and sell subdomains within it, that would drastically increase the supply of domain names. The prices of domains should go down, in that case, since if you can't get joe-blow.tld1 you could get joe-blow.tld2.
Of course, this is assuming that the domain registrars don't form a cartel. But the point is that generic TLDs aren't a big deal; a cartel of domain registrars is.
I think your argument relies on fixing the sense of the term "insurance" in a specific way that is convenient to your argument; in that regard, I think it begs the question. If you define "insurance" more abstractly, as something like "a scheme where most participants overpay for what they get, so that a minority of them underpay," then health expenditures based on taxes are a form of insurance.
I think the actual substance of the point, that compulsory, equal-pay insurance schemes are best implemented as government programs paid from taxes, is correct, though.
I don't think the two statements can be easily reconciled. If we can forbid the insurance companies from using race to set costs, why can't we forbid them from using other factors? (For example, prohibiting health insurance companies, in their current form in the USA, from using DNA test results as a cost factor.)
Let's grant that this works, for the sake of argument. The company is taking action against you in a court, you argue that you did not agree to the EULA on these grounds, and your argument is accepted. What comes next?
Well, this case is analogous to you taking some goods or service from somebody on the mistaken belief that they were giving it away for free. If all parties agree that you mistakenly and reasonably believed the goods or servie was being given away, all that shows is that you didn't steal (since you had no mens rea). You still did take something from them without their permission, and may still be liable for that. Probable outcome: you return the goods, or pay them for the service, and the matter is dropped.
So in the case where your EULA argument somehow worked, the best you could get away with is the vendor agreeing to drop the matter if you surrender all of your copies of their software.