Applications that only support.docx read capabilities include Apple's iWork, IBM's Lotus Notes, Oracle's OpenOffice.Org and Google Docs, amongst others.
I also just checked, and it's true--I can't save as.docx with OO.O Writer. Read-only capability isn't good enough, since "[software] must have the ability to read and write the endorsed file format".
Not everything is language-specific. The question is vague, saying only "the new grad knew a hot emerging technology that a client wanted." Perhaps that technology is an off-the-wall database technology like RDF, or the ins-n-outs of a huge framework like Boost. Even among languages, there's the imperative language family that's all pretty similar (Java, Basic and variants, C, C++, C#,...) and there's the functional family (Lisp, Scheme, Erlang, Haskell,...). Knowing imperative languages certainly makes for an easy time learning a new imperative language, but if your hot new technology is just a functional language and you've only done imperative, you'll probably have a long learning curve ahead of you.
I tend to agree. An omniscient creator would know whether or not a particular tuning would fit its needs, regardless of the probabilities humans might assign to constant ranges.
Although science can prove/disprove specific doctrines of religions, it can never disprove the the existence of God, nor the faith therein.
I disagree slightly, since "God" is not well defined. If "God" is a bowl of talking spaghetti sitting right in front of me at this moment, science could disprove God's existence. If God is a being in another universe with no direct or indirect interactions with this universe, science couldn't disprove God's existence. People don't agree on who/what God is.
Well, yes and no. The weak anthropic principle does just draw the conclusion "we have won the lottery". The strong anthropic principle concludes "the universe has properties which at some point allowed life to exist, and we won the lottery". These conclusions also depend on whose version of the principle(s) you're using, and the fact that most (all?) formulations are tautologies confuses the issues further. TBH it's really quite a philosophical mess.
My brief take on it:
1. If there is a single universe, the anthropic principle does not refute an intelligent designer having to "fine tune" the universe for life.
2. If there are multiple universes, the anthropic principle does refute an intelligent designer having to "fine tune" the universe for life.
Either of these cases is untestable, so only of philosophical interest. It is important to remember "we exist" is a valid assumption for scientific reasoning, which is the real content of the anthropic principle. The rest is mostly just tongue-wagging.
A mere 0.001% of Americans would have been a sample of 30,800 rather than 1019 - over 30x as accurate?
That's not how it works. The standard error of the mean (which may or may not be applicable depending on your poll, but at least is illustrative) scales as 1/Sqrt(n) where n is the sample size. Multiplying your sample size by 30 multiplies your error by 1/Sqrt(30) ~= 0.18, giving you around ~5x the accuracy (i.e. 1/5th the error margin), not 30x. Also, you'll notice that formula doesn't involve the size of the true population. Scaling the population of the US by 1000 doesn't affect the error margin of a sample, relying on the standard error of the mean. Suggesting we require a certain percentage of the population be used in polls is therefore a novice mistake and glosses over the more complex statistical issues involved.
The most recent poll on Slashdot, posted yesterday, already has over 7900 votes. The prior poll from five days ago has over 27600 and ten days ago is up to 41300.
Slashdot's polls are well and truly not representative of Americans. Off the top of my head, they'd have heavy liberal, pro-science, anti-religion biases compared to the full population. (This is not necessarily a good or bad thing, it's just a thing.) Slashdot polls also self-select--you only answer the question if it interests you. Getting a representative sample of Americans is non-trivial and many polls stop at about ~1000 since they're happy with the error margin at that point (around 3% for most polls I've seen; this varies with the polling methods and confidence level). These are often conducted by phone, and calling enough people to get 1000 responses is a big job.
That said, ordering TFA's list is probably terrible statistics. Subsample sizes that round to 2% are only 16-25 people. To be fair, TFA does list several "ties" but it's unclear if those are true ties (i.e. 20 people each picked Jimmy Carter and Glenn Beck) or if they're statistical ties (i.e. the error margins allow the order of the two to be unclear at a particular confidence level). The article goes on to speculate on causes of Gates' rise in the list, which is just silly. Perhaps the cause is that people who preferred several of the people who beat Gates last year now generally prefer Obama. In this case none of Gates' actions contribute to his ranking.
I hate it when journalists report statistics. It's so often just terrible. They very often ignore error margins entirely and draw strong conclusions from weak data, out of ignorance. Journalists aren't statisticians (neither am I) but they could at least stick to reporting the facts.
I watch Jeopardy(!) pretty regularly, and wordplay that's this involved is rare. You might get a single category like this per week, and that's generous. (Each day's game has 12 categories, 6 per round.) I'd imagine Watson would completely choke on this one, since it's a case the designers would probably be negligent to spend a lot of time on. Still, if it would handle it, I'd be really impressed:).
I can't quite tell if you understand the wordplay. "Novel movie" is referring to a fake ("novel") conflation of a real novel and a real movie, so I don't know what you mean by "search for [a] novel movie"--it's too specialized to expect Watson to be able to do. To be clear, the novel is The Grapes of Wrath and the movie is The Wrath of Khan. Combining the two titles gives The Grapes of Khan's Wrath (perhaps a few others would be acceptable). The question doesn't give Ricardo Montelban's character's name since it's Khan, part of the answer.
The first sentence of TFA contains a typo, "I can still remember when my when I first switched over to Google on the recommendation of my brother’s girlfriend." The author admits "quality" is subjective but calls the test "an objective small sample size evaluation", which is inconsistent. The points scale (5 for a "good" first result, 3 for second, 1 for third) seems arbitrary--it's not at all what I would use, at least, which would probably use total elapsed time. The final totals are very sensitive to error, because of small sample size, and no error-related issues are considered. Any reasonable error bounds should make it impossible to say whether Google beat Bing or vice versa, statistically.
The author did at least try to be objective, but the only real information TFA's test contains is "Google and Bing aren't completely incomparable in quality."
In today’s exhibition of about 15 questions, Watson tallied $4,400, compared to $3,400 for Jennings and $1,200 Rutter.
15 questions is not really enough to say anything. Maybe the humans aren't used to buzzing in this context (I assume Watson's buzzer skills are basically static), they weren't "warmed up", or the categories favored Watson's strengths and their weaknesses. (Or maybe not.) Interestingly, if these 15 questions corresponded to 3 first round categories, every question was answered, from the money totals. I also find this interesting:
[Watson] buzzes in about half the time, and answers 85 to 95 percent of those questions correctly.
Watson handling certain wordplay would really impress me. For instance take "What is the Grapes of Khan's Wrath" answering "The Jodes leave Oklahoma and meet up with villain Ricardo Montelban's character in this novel movie". That would just be awful to automate, I imagine.
Humans don't have to listen to Trebek; they can read ahead if they want. TBH he reads pretty slowly and if you can tune him out you might be at an advantage. A computer might be able to "read" faster than a human, but I don't see that as unfair.
I've never seen "sic" used for logical fallacies, just spelling and some grammar. It's just a flag for writers to say, "I can spell but the person I'm quoting can't. Don't blame me."
Harder than hard, sometimes it'd be impossible with a single picture. The user might cover up one piece (say, a pawn) with another (say, a king) from their bad angle choice. Depending on the other pieces you might not be able to infer the location of the covered pawn. Similarly a bishop could probably be blocked by a queen, which is even worse.
Sometimes *I* have trouble figuring out what piece is what on a chess set, like in this random crystal set. Because I know it's probably early-game I can figure out which piece is what after looking at it for a bit, but still, that one would be horrible to feed a chess board recognizer.
My aunt likes Sudokus but isn't terribly good at them. I could see her enjoying a hint every now and then without having to wait for the next day's paper or type it in to a computer. Then again, there's no way she'd use this setup without me or someone like me getting her into smart phones.
Hmm, interesting. I think the devil would be in the details. Over time if's (and the like) and functions are removed or added. Matching functions up would be difficult without a human to guide the process. A simpler problem to automate would probably be to compare structs (or the language-specific equivalent). You could compare names and data types with order considerations. After fields are typed, they usually don't switch orders or data types (perhaps excepting signed/unsigned), so it'd be decently robust. You could automatically compare two routine's local variables and perhaps arguments/return values similarly. Comparing two routine's decision trees would be difficult and probably uninformative if the routines were short.
A lot of code similarity is from things that the compiler "doesn't notice", like where whitespace is used, naming conventions, (partial) abbreviations, code (including function argument) ordering, and file divisions. In this case each of these provides evidence that the code was derived.
Alright. The source files for the code in my citation are WinMTRNet.cpp (see also WinMTRNet.h) and net.c in the mtr-0.80.tar.gz. Both include the ICMPHeader struct with identical field names (not very strong evidence), the struct sequence with identical field names, and the structs nethost and s_nethost which share half a dozen field names precisely--when they do they are in the same order. These three structs are in the same order in each file.
The functions after new_sequence in the MTR source are, in order, net_send_query, net_process_ping, and net_process_return; the functions after GetNewSequence are SendQuery, ProcessPing, and ProcessReturn.
These are things I noticed looking at those two source files briefly; I didn't have to hunt around at all. It's reasonable to expect a fair amount of difference since WinMTR was started ~8 years ago and both sources have diverged somewhat since then. Expecting large blocks of identical code seems unreasonable because of the time frame, though I wouldn't be at all surprised to find numerous snippets like the above in a detailed analysis.
There appear to be large blocks of very suggestively similar code. Ordering and name conventions alone are enough for me that I would swear in court that the two are derivatives.
I didn't meant to imply it was conclusive evidence that WinMTR is a derivative work, just that it was evidence. I do believe it's conclusive evidence that one routine was derived from the other, though as you say that could mean MTR copied the code from WinMTR instead of the other way around, or that it was copied with permission (though this is quite unlikely since the maintainer for over the last decade is quoted in TFA and says nothing of the kind).
Yes, but the variable and field names are entirely identical, down to the underscore and capitalization. The non-identical WinMTR code tends to prefer CamelCase: "WinMTRNet::GetNewSequence" and "SaveXmit(index);".
Moreover, the order of declaration of "next_sequence" and "seq" are identical in each version (as are their precise names) and the order of the three lines I quoted is identical. Either of these two groups could be permuted without changing the resulting code--there are 2*6 = 12 possibilities in those lines alone, considering only line ordering, and they happen to match. So, I'd say it really can be implemented many other ways and this is probabilistically very strong evidence.
There are numerous other suggestive details I could point out. True, it's not utterly conclusive, but neither is DNA testing (which has a small but non-zero error rate).
Provably so... in the context of a particular theory. The axioms of that theory cannot be proven, except in the context of yet another theory, and the rules of inference suffer similarly. A great example of the arbitrariness of axioms is the Axiom of Choice or the Generalized Continuum Hypothesis. Fundamentally, they're on same shaky foundation to me as, say, allowing proof by contradiction. Even defining the word "consistent" doesn't work: either it is inconsistent, or we must first check its consistency to use it in truly rigorous deductive reasoning, which is circular since then we need it to define itself. What's more, we really need a notion of reasoning before we can completely rigorously discuss consistency, while reasoning is without foundation without a concept of consistency--they're circularly dependent.
It is at this point I thank God for emotion, which causes me to mostly ignore these almost certainly unhelpful questions. Because of it I'm really pretty happy accepting first order predicate logic and Peano arithmetic, as well as the ability for some humans to apply them to reality (whatever that is). Belief is really quite amazing. It's strongly abused, but utterly necessary.
I'm claiming my quote is evidence that WinMTR is a derivative work of MTR, so that the GPL applies to force WinMTR to be open source. It doesn't just "look like"; those lines are verbatim in both sources, and it's clear that one was derived from the other from the rest of the source quoted in my citation.
If you don't accept that as evidence, what would you accept?
Mostly I was hoping to either get a lawyer's opinion (there are a few on/.) or hear from someone who's dealt with a similar situation personally.
I can understand your reasoning. Indirectly, though, *my* rights would be violated as a consequence of the copyright holder's being violated. That he doesn't care and I do doesn't seem like it matters. But IANAL.
If I downloaded WinMTR and asked them for source code pursuant to the GPL (supposing Appnor did wrongfully change the license), would I have any legal recourse if they refuse, as someone who has never contributed to MTR? If the software were more important this would actually be an important situation to deal with, so I'm curious.
From the WinMTR site, "Your Cost: FREE! Provided you subscribe to our newsletter. You can opt out immediately if you don’t like it." More relevant quotes: "License: Commercial. We changed it from GPL since in the last 10 years there was no external development. Still, we plan to offer it for free", and the title of every last page, "WinMTR – Appnor's Free Network Diagnostic Tool" (emphasis mine).
Slashdot Mirror
Applications that only support .docx read capabilities include Apple's iWork, IBM's Lotus Notes, Oracle's OpenOffice.Org and Google Docs, amongst others.
I also just checked, and it's true--I can't save as .docx with OO.O Writer. Read-only capability isn't good enough, since "[software] must have the ability to read and write the endorsed file format".
Not everything is language-specific. The question is vague, saying only "the new grad knew a hot emerging technology that a client wanted." Perhaps that technology is an off-the-wall database technology like RDF, or the ins-n-outs of a huge framework like Boost. Even among languages, there's the imperative language family that's all pretty similar (Java, Basic and variants, C, C++, C#, ...) and there's the functional family (Lisp, Scheme, Erlang, Haskell, ...). Knowing imperative languages certainly makes for an easy time learning a new imperative language, but if your hot new technology is just a functional language and you've only done imperative, you'll probably have a long learning curve ahead of you.
I tend to agree. An omniscient creator would know whether or not a particular tuning would fit its needs, regardless of the probabilities humans might assign to constant ranges.
Although science can prove/disprove specific doctrines of religions, it can never disprove the the existence of God, nor the faith therein.
I disagree slightly, since "God" is not well defined. If "God" is a bowl of talking spaghetti sitting right in front of me at this moment, science could disprove God's existence. If God is a being in another universe with no direct or indirect interactions with this universe, science couldn't disprove God's existence. People don't agree on who/what God is.
Well, yes and no. The weak anthropic principle does just draw the conclusion "we have won the lottery". The strong anthropic principle concludes "the universe has properties which at some point allowed life to exist, and we won the lottery". These conclusions also depend on whose version of the principle(s) you're using, and the fact that most (all?) formulations are tautologies confuses the issues further. TBH it's really quite a philosophical mess.
My brief take on it:
1. If there is a single universe, the anthropic principle does not refute an intelligent designer having to "fine tune" the universe for life.
2. If there are multiple universes, the anthropic principle does refute an intelligent designer having to "fine tune" the universe for life.
Either of these cases is untestable, so only of philosophical interest. It is important to remember "we exist" is a valid assumption for scientific reasoning, which is the real content of the anthropic principle. The rest is mostly just tongue-wagging.
A mere 0.001% of Americans would have been a sample of 30,800 rather than 1019 - over 30x as accurate?
That's not how it works. The standard error of the mean (which may or may not be applicable depending on your poll, but at least is illustrative) scales as 1/Sqrt(n) where n is the sample size. Multiplying your sample size by 30 multiplies your error by 1/Sqrt(30) ~= 0.18, giving you around ~5x the accuracy (i.e. 1/5th the error margin), not 30x. Also, you'll notice that formula doesn't involve the size of the true population. Scaling the population of the US by 1000 doesn't affect the error margin of a sample, relying on the standard error of the mean. Suggesting we require a certain percentage of the population be used in polls is therefore a novice mistake and glosses over the more complex statistical issues involved.
The most recent poll on Slashdot, posted yesterday, already has over 7900 votes. The prior poll from five days ago has over 27600 and ten days ago is up to 41300.
Slashdot's polls are well and truly not representative of Americans. Off the top of my head, they'd have heavy liberal, pro-science, anti-religion biases compared to the full population. (This is not necessarily a good or bad thing, it's just a thing.) Slashdot polls also self-select--you only answer the question if it interests you. Getting a representative sample of Americans is non-trivial and many polls stop at about ~1000 since they're happy with the error margin at that point (around 3% for most polls I've seen; this varies with the polling methods and confidence level). These are often conducted by phone, and calling enough people to get 1000 responses is a big job.
That said, ordering TFA's list is probably terrible statistics. Subsample sizes that round to 2% are only 16-25 people. To be fair, TFA does list several "ties" but it's unclear if those are true ties (i.e. 20 people each picked Jimmy Carter and Glenn Beck) or if they're statistical ties (i.e. the error margins allow the order of the two to be unclear at a particular confidence level). The article goes on to speculate on causes of Gates' rise in the list, which is just silly. Perhaps the cause is that people who preferred several of the people who beat Gates last year now generally prefer Obama. In this case none of Gates' actions contribute to his ranking.
I hate it when journalists report statistics. It's so often just terrible. They very often ignore error margins entirely and draw strong conclusions from weak data, out of ignorance. Journalists aren't statisticians (neither am I) but they could at least stick to reporting the facts.
I watch Jeopardy(!) pretty regularly, and wordplay that's this involved is rare. You might get a single category like this per week, and that's generous. (Each day's game has 12 categories, 6 per round.) I'd imagine Watson would completely choke on this one, since it's a case the designers would probably be negligent to spend a lot of time on. Still, if it would handle it, I'd be really impressed :).
I can't quite tell if you understand the wordplay. "Novel movie" is referring to a fake ("novel") conflation of a real novel and a real movie, so I don't know what you mean by "search for [a] novel movie"--it's too specialized to expect Watson to be able to do. To be clear, the novel is The Grapes of Wrath and the movie is The Wrath of Khan. Combining the two titles gives The Grapes of Khan's Wrath (perhaps a few others would be acceptable). The question doesn't give Ricardo Montelban's character's name since it's Khan, part of the answer.
The first sentence of TFA contains a typo, "I can still remember when my when I first switched over to Google on the recommendation of my brother’s girlfriend." The author admits "quality" is subjective but calls the test "an objective small sample size evaluation", which is inconsistent. The points scale (5 for a "good" first result, 3 for second, 1 for third) seems arbitrary--it's not at all what I would use, at least, which would probably use total elapsed time. The final totals are very sensitive to error, because of small sample size, and no error-related issues are considered. Any reasonable error bounds should make it impossible to say whether Google beat Bing or vice versa, statistically.
The author did at least try to be objective, but the only real information TFA's test contains is "Google and Bing aren't completely incomparable in quality."
In today’s exhibition of about 15 questions, Watson tallied $4,400, compared to $3,400 for Jennings and $1,200 Rutter.
15 questions is not really enough to say anything. Maybe the humans aren't used to buzzing in this context (I assume Watson's buzzer skills are basically static), they weren't "warmed up", or the categories favored Watson's strengths and their weaknesses. (Or maybe not.) Interestingly, if these 15 questions corresponded to 3 first round categories, every question was answered, from the money totals. I also find this interesting:
[Watson] buzzes in about half the time, and answers 85 to 95 percent of those questions correctly.
Watson handling certain wordplay would really impress me. For instance take "What is the Grapes of Khan's Wrath" answering "The Jodes leave Oklahoma and meet up with villain Ricardo Montelban's character in this novel movie". That would just be awful to automate, I imagine.
Humans don't have to listen to Trebek; they can read ahead if they want. TBH he reads pretty slowly and if you can tune him out you might be at an advantage. A computer might be able to "read" faster than a human, but I don't see that as unfair.
I've never seen "sic" used for logical fallacies, just spelling and some grammar. It's just a flag for writers to say, "I can spell but the person I'm quoting can't. Don't blame me."
Harder than hard, sometimes it'd be impossible with a single picture. The user might cover up one piece (say, a pawn) with another (say, a king) from their bad angle choice. Depending on the other pieces you might not be able to infer the location of the covered pawn. Similarly a bishop could probably be blocked by a queen, which is even worse.
Sometimes *I* have trouble figuring out what piece is what on a chess set, like in this random crystal set. Because I know it's probably early-game I can figure out which piece is what after looking at it for a bit, but still, that one would be horrible to feed a chess board recognizer.
I'd like to think the GP meant "the part of your brain that does things with logic".
My aunt likes Sudokus but isn't terribly good at them. I could see her enjoying a hint every now and then without having to wait for the next day's paper or type it in to a computer. Then again, there's no way she'd use this setup without me or someone like me getting her into smart phones.
Hmm, interesting. I think the devil would be in the details. Over time if's (and the like) and functions are removed or added. Matching functions up would be difficult without a human to guide the process. A simpler problem to automate would probably be to compare structs (or the language-specific equivalent). You could compare names and data types with order considerations. After fields are typed, they usually don't switch orders or data types (perhaps excepting signed/unsigned), so it'd be decently robust. You could automatically compare two routine's local variables and perhaps arguments/return values similarly. Comparing two routine's decision trees would be difficult and probably uninformative if the routines were short.
A lot of code similarity is from things that the compiler "doesn't notice", like where whitespace is used, naming conventions, (partial) abbreviations, code (including function argument) ordering, and file divisions. In this case each of these provides evidence that the code was derived.
Alright. The source files for the code in my citation are WinMTRNet.cpp (see also WinMTRNet.h) and net.c in the mtr-0.80.tar.gz. Both include the ICMPHeader struct with identical field names (not very strong evidence), the struct sequence with identical field names, and the structs nethost and s_nethost which share half a dozen field names precisely--when they do they are in the same order. These three structs are in the same order in each file.
The functions after new_sequence in the MTR source are, in order, net_send_query, net_process_ping, and net_process_return; the functions after GetNewSequence are SendQuery, ProcessPing, and ProcessReturn.
These are things I noticed looking at those two source files briefly; I didn't have to hunt around at all. It's reasonable to expect a fair amount of difference since WinMTR was started ~8 years ago and both sources have diverged somewhat since then. Expecting large blocks of identical code seems unreasonable because of the time frame, though I wouldn't be at all surprised to find numerous snippets like the above in a detailed analysis.
There appear to be large blocks of very suggestively similar code. Ordering and name conventions alone are enough for me that I would swear in court that the two are derivatives.
I didn't meant to imply it was conclusive evidence that WinMTR is a derivative work, just that it was evidence. I do believe it's conclusive evidence that one routine was derived from the other, though as you say that could mean MTR copied the code from WinMTR instead of the other way around, or that it was copied with permission (though this is quite unlikely since the maintainer for over the last decade is quoted in TFA and says nothing of the kind).
Yes, but the variable and field names are entirely identical, down to the underscore and capitalization. The non-identical WinMTR code tends to prefer CamelCase: "WinMTRNet::GetNewSequence" and "SaveXmit(index);".
Moreover, the order of declaration of "next_sequence" and "seq" are identical in each version (as are their precise names) and the order of the three lines I quoted is identical. Either of these two groups could be permuted without changing the resulting code--there are 2*6 = 12 possibilities in those lines alone, considering only line ordering, and they happen to match. So, I'd say it really can be implemented many other ways and this is probabilistically very strong evidence.
There are numerous other suggestive details I could point out. True, it's not utterly conclusive, but neither is DNA testing (which has a small but non-zero error rate).
Provably so... in the context of a particular theory. The axioms of that theory cannot be proven, except in the context of yet another theory, and the rules of inference suffer similarly. A great example of the arbitrariness of axioms is the Axiom of Choice or the Generalized Continuum Hypothesis. Fundamentally, they're on same shaky foundation to me as, say, allowing proof by contradiction. Even defining the word "consistent" doesn't work: either it is inconsistent, or we must first check its consistency to use it in truly rigorous deductive reasoning, which is circular since then we need it to define itself. What's more, we really need a notion of reasoning before we can completely rigorously discuss consistency, while reasoning is without foundation without a concept of consistency--they're circularly dependent.
It is at this point I thank God for emotion, which causes me to mostly ignore these almost certainly unhelpful questions. Because of it I'm really pretty happy accepting first order predicate logic and Peano arithmetic, as well as the ability for some humans to apply them to reality (whatever that is). Belief is really quite amazing. It's strongly abused, but utterly necessary.
I'm claiming my quote is evidence that WinMTR is a derivative work of MTR, so that the GPL applies to force WinMTR to be open source. It doesn't just "look like"; those lines are verbatim in both sources, and it's clear that one was derived from the other from the rest of the source quoted in my citation.
If you don't accept that as evidence, what would you accept?
Mostly I was hoping to either get a lawyer's opinion (there are a few on /.) or hear from someone who's dealt with a similar situation personally.
I can understand your reasoning. Indirectly, though, *my* rights would be violated as a consequence of the copyright holder's being violated. That he doesn't care and I do doesn't seem like it matters. But IANAL.
If I downloaded WinMTR and asked them for source code pursuant to the GPL (supposing Appnor did wrongfully change the license), would I have any legal recourse if they refuse, as someone who has never contributed to MTR? If the software were more important this would actually be an important situation to deal with, so I'm curious.
From the WinMTR site, "Your Cost: FREE! Provided you subscribe to our newsletter. You can opt out immediately if you don’t like it." More relevant quotes: "License: Commercial. We changed it from GPL since in the last 10 years there was no external development. Still, we plan to offer it for free", and the title of every last page, "WinMTR – Appnor's Free Network Diagnostic Tool" (emphasis mine).