(Gee, thanks for the civil reply. You gain a lot of karma points that way, I'm sure.)
In any case, the contention was not (if you read a bit more carefully) that a single-user metric should count the same user twice. Obviously. Rather, the argument is that at a single-user metric is not really a good one to measure 'market browser share' at all, because it overstates the usage by low-use, occasional users, and understates usage by high-use, constant users. As a website developer, sure, I'm interested in both per-visitor and per-page metrics - but the latter is much more important to me because it more accurately tallies who's using my site the most.
But since your position in Microsoft management is pretty secure, that distinction's probably not important to you.
Even if you think of this (as another commenter has it) as a "unique workflow", I think it misses an even greater source of error in NetApplication's approach. Consider:
- My grandma uses IE as preloaded on her Windows PC and goes to, say, Gmail (yes, at least I got her off Hotmail:) once a day to get her cat pictures. She's counted as a single unique visitor by NetApp. - I go to Gmail with Chrome in the morning and live on it all day, loading hundreds if not thousands of pages during that time. Despite that, I'm _still_ counted as a single unique visitor by NetApp. Even though the "eyeball time" (the real "browser usage") between me and grandma is vastly different.
I think the "single user" metric has an inherent biased towards low-usage, unsophisticated users - the ones most likely not to have replaced IE as loaded on their systems. So it makes sense (not even counting the geo-weighting issues) that they'd have IE's share much higher than anyone else's. Though no single approach is perfect, that's why I think of the two StatCounter's is netter. (And frankly it's always been more in line with other metrics - like the Wikimedia stats - that seem unbiased and cut a wide swath of the Net. NetApplications has always been the outlier.)
That $70 is for "Basic Phones" and so doesn't apply to anything Verizon deems a "smartphone". Which is pretty much everything that can deal with mobile data in a non-trivial way. Not a valid comparison unless you're my grandma.
It's pretty obvious the new pricing is a "screw the single user who wasn't using a bazillion bytes of data or talk time" plan. Maybe it's Verizon's way of saying I should get hitched? My mom would like that, for sure.
Or maybe they're abandoning the singles market in dense urban centers to Sprint (about the only place Sprint works well, so it plays to their strengths).
Honestly, I was thinking about switching to Verizon when my AT&T contract ended later this month or when the next iPhone came out. It would have only been about $5 more. Now it will be $25 or more greater. Non-starter for me, and I suspect many others.
For certain definitions of the the term, sure, it's both; for others, not.
"Mistake" was put in quotes in what I wrote because of the original poster's implication was that all bugs were things completely avoidable, serious failures that were something the programmer should rectify for free and that not doing for a customer so was immoral/improper - like mis-assembled car that should have its manufacturing defect(s) covered under warranty. Custom-written software isn't in the same category, IMHO. If you want perfection, you have to pay for testing to perfection - not something most clients are willing (understandably) to do.
I sure would not want to program for you. In 25 years of independent development, I only saw the bizarre belief you express from a single one client. I gave them two alternatives:
(a) "After the initial acceptance period, we'll fix all bugs for free...but of course you need to pay my team for 5-man-weeks of testing and QA time so that we can both be assured it's perfect first", OR (b) "You'll pay us time and materials to fix any significant bugs you find, but we'll only charge you for 5-man-days of testing and QA time beforehand and you'll work with us to discover any others we missed as you use the software."
Needless to say, not being stupid, they took option (b) and we probably only ended up charging them for a few minor fixes.
A software bug is not "a mistake". It's an inevitable part of the process, one that can be mitigated by good design, good coding, good management, and good testing. But all of those things take time and money. There's no magic zero-cost shortcut to perfection in any non-trivial project.
Replying to my own posting. Reading later comments above, the phosphor-pumping step (necessary to produce pleasant-looking "white" light rather than monochromatic base-LED light) cuts into the efficiency greatly. As in, "white" LEDs are a mix of monochromatic LEDs and phosphors that re-emit their light in more broad-spectrum and more complex ways. At an efficiency cost, of course. Hadn't considered that - makes sense.
I do have a few LEDs in my house and like them a lot. But then other than not being dimmable, I didn't mind CFLs much either - the "not fully bright immediately" thing was to me an advantage, not a disadvantage, especially before morning coffee.
Curious. A straight technical question I'm not quite getting:
We've long been told incandescent bulbs are barely 2% efficient - that 98% of their output is "heat" (meaning any electromagnetic energy not in the visible spectrum) and only 2% is "light" (meaning EM in the visible range). Kewl, fine.
So if a 100W-equivalent LED-based bulb consumes (for the sake of round numbers) 25 watts, it's four times as efficient as the equivalent incandescent, meaning 8% overall efficiency (4x2% = 8%), right? But I thought LEDs (the diodes themselves) were supposed to be 20-50% efficient, not 8%? I know there are supporting electronics and all, but why the huge discrepancy between the LEDs themselves and their instantiation in a consumer fixture?
Perhaps someone with actual experience in the field can enlighten (ahem) me on this.
Those following Groklaw closely through the long months of pre-trial and actual trial understand this partial verdict in a much fuller context. (Unfortunately, most of the posters here seem not to fall into that category.)
Overall, it was a great day for Google, and Wall Street got that immediately: GOOG is up about 1.75% on the day, and ORCL down by a similar amount. So clearly the people with money on the line wouldn't agree with TFA's headline here. (Which, to be fair, is how most of the uninformed news media coverage is spinning it, so Slashdot is hardly alone in getting it wrong.
Nerd fail. Was actually pinging from the wrong shell, which includes the latency of my laptop's wireless link. So the actual CenturyLink DSL latencies are 5ms or so less for me: 15-20ms to the local Qwest.net DNS server, 70-75ms to Google DNS.
We now return you to more competent geek programming.
+1 on this. Without more information from the OP about the test nor its destination, it's nearly impossible to draw any conclusions.
FWIW, I have CenturyLink 12.0mb/s DSL in the Seattle metro area. I just pinged their (CenturyLink's) local DNS server (205.171.3.x) from my router, and have latency consistently in the 20-25ms range - which I consider perfectly fine. (traceroute shows five hops total, bwt, all in the qwest.net network).
Pinging Google's public DNS server, outside of qwest.net at 8.8.8.8, gives 75-80ms latency. Again, just fine in my book. ./tsg/
I think GoDaddy sucks large avian eggs on a variety of levels (morally and technically); I try hard to convince any client who comes to me with domains or hosting there to switch to something else - anything else. (I usually recommend Pairnic.com as a registrar and Pair.com for hosting, but other great options exist.) </disclosure>
That said, I transferred a client's domain from GoDaddy just last night (to Pairnic) and the process was exceedingly quick and smooth. Since GoDaddy allows you to log in an "approve" a transfer in process, the whole thing was done in 20 minutes from start to finish. Never had one go that fast, so I have to give credit to GoDaddy where it's due here. At least in this case, they were not putting up any roadblocks.
But yes, I'm glad to be almost completely rid of them. Even if Danica is smokin' hot.
Just 'cuz I was curious, and it has some peripheral bearing on the question - assuming 19.816 gm/cm^3 for the density of Pu (more than lead) and also assuming (since it's the UK) we're talking "tons" = metric tonnes = 1000kg = 10^6 gm -
87 x 10^6 gm / 19.816 gm/cm^3 = 4.39 x 10^6 cm^3 = 4.39 m^3.
4.39 cubic meters is a single cube 1.637 meters on a side (or a little more than 5 feet/side, for us backward Yanks). More or less the size of a smallish SUV, yes?
Of course their Pu isn't, one hopes, stored all in one solid cube, which would probably exceed critical mass by some large factor. But still, it's not a massive physical quantity of material you're talking about here./TSG/
The article is completely wrong in relation to CentOS (which aims for 100% binary compatibility to RH). AFAIK, they don't care which patches were applied by RH because they're duplicating the kernel down to the last byte. As you say, they'll just compile the tarball and off she goes.
The article is correct as far as an entity like Oracle is concerned, which aims to put in its own additions and "improvements".
I'm of two minds about whether RH is evil or prudent to do this, but on balance I've got no lost love for Larry Ellison, so I give RH the benefit of the doubt on this one.
Just an FYI: reading through the transcript I kept seeing things like "NOUN 37" and "VERB 12" - I thought these might be redactions for national security or censorship of Very Bad Words (ala the Nixon White House tapes and "expletive deleted" - but I'm dating myself to know about that). But they actually seem to be the way the internal shipboard guidance computer was controlled, with two part commands, one being an action (not surprisingly, "VERB yy") and one being an object to be acted upon ("NOUN xx"). Details here:
Interestingly, this is not at all unlike how the original Fortran code for ADVENT (the seminal "Collossal Cave Adventure") was architected, even down to the terminology used.
As fellow author, Will Belegon, noted, if Amazon is going to start pulling books with incest in them: "I just re-read Genesis 19: 30-38 and realized that Lot's daughters got him drunk, had sex with him and bore sons. I demand you follow your clear precedent and remove The Bible from Kindle."
Based on preliminary information I heard from the Facebook launch announcement today [...] users will not have the ability to declare chats or related conversations to be "off the record" -- everything will apparently be recorded. Individual users will have the ability to archive or delete their own copies of transcripts, but it appears that there is explicitly not a functionality similar to Google's "off the record" chat feature, which permits users to declare that their conversations with given individuals should not be routinely preserved. "It just didn't make sense for us," were pretty much the words that Zuckerberg used in
response to a question on this topic."
True...but just to be clear (and I speak as someone who owns a business in Washington), the specific tax in question is the state "Business and Occupation" tax, which (for manufacturing activities, at least) is a tax levied at.484% of the gross revenue of the business - not the net income, not the net profit, but the gross total of checks that came in the door. Yes, it's pretty bizarro, but then without a state personal or corporate income tax, they do what they can to keep the lights on in the Capital Building.
All by way of saying that.484% adds up to a pretty tidy sum when levied on Microsoft's gross licensing revenue, worldwide.
At the time of my post, there were only ~10 comments, and many (including the article summary itself) were not positive. Glad to know that others share a good opinion of it.
...who thought this ad was pretty damn brilliant? Low-key, sure, but also sweet, memorable, and focused on the product/service itself rather than hype and glitz? I thought it promoted the both the company's values and the value of what they provide to their customers extraordinarily well.
Maybe I'm just not cynical enough, but it sure gets my vote./tsg/
Slashdot Mirror
(Gee, thanks for the civil reply. You gain a lot of karma points that way, I'm sure.)
In any case, the contention was not (if you read a bit more carefully) that a single-user metric should count the same user twice. Obviously. Rather, the argument is that at a single-user metric is not really a good one to measure 'market browser share' at all, because it overstates the usage by low-use, occasional users, and understates usage by high-use, constant users. As a website developer, sure, I'm interested in both per-visitor and per-page metrics - but the latter is much more important to me because it more accurately tallies who's using my site the most.
But since your position in Microsoft management is pretty secure, that distinction's probably not important to you.
lol; "better" not "netter". Though that's that's probably true, too. :)
Even if you think of this (as another commenter has it) as a "unique workflow", I think it misses an even greater source of error in NetApplication's approach. Consider:
- My grandma uses IE as preloaded on her Windows PC and goes to, say, Gmail (yes, at least I got her off Hotmail :) once a day to get her cat pictures. She's counted as a single unique visitor by NetApp.
- I go to Gmail with Chrome in the morning and live on it all day, loading hundreds if not thousands of pages during that time. Despite that, I'm _still_ counted as a single unique visitor by NetApp. Even though the "eyeball time" (the real "browser usage") between me and grandma is vastly different.
I think the "single user" metric has an inherent biased towards low-usage, unsophisticated users - the ones most likely not to have replaced IE as loaded on their systems. So it makes sense (not even counting the geo-weighting issues) that they'd have IE's share much higher than anyone else's. Though no single approach is perfect, that's why I think of the two StatCounter's is netter. (And frankly it's always been more in line with other metrics - like the Wikimedia stats - that seem unbiased and cut a wide swath of the Net. NetApplications has always been the outlier.)
Original "has": http://icanhascheezburger.com/2007/01/11/i-can-has-cheezburger-3/
"Haz" would have been way funnier here, though.
"I CAN BE HAZ MAT?!"
I can has worldwide pandemic?
That $70 is for "Basic Phones" and so doesn't apply to anything Verizon deems a "smartphone". Which is pretty much everything that can deal with mobile data in a non-trivial way. Not a valid comparison unless you're my grandma.
It's pretty obvious the new pricing is a "screw the single user who wasn't using a bazillion bytes of data or talk time" plan. Maybe it's Verizon's way of saying I should get hitched? My mom would like that, for sure.
Or maybe they're abandoning the singles market in dense urban centers to Sprint (about the only place Sprint works well, so it plays to their strengths).
Honestly, I was thinking about switching to Verizon when my AT&T contract ended later this month or when the next iPhone came out. It would have only been about $5 more. Now it will be $25 or more greater. Non-starter for me, and I suspect many others.
For certain definitions of the the term, sure, it's both; for others, not.
"Mistake" was put in quotes in what I wrote because of the original poster's implication was that all bugs were things completely avoidable, serious failures that were something the programmer should rectify for free and that not doing for a customer so was immoral/improper - like mis-assembled car that should have its manufacturing defect(s) covered under warranty. Custom-written software isn't in the same category, IMHO. If you want perfection, you have to pay for testing to perfection - not something most clients are willing (understandably) to do.
I sure would not want to program for you. In 25 years of independent development, I only saw the bizarre belief you express from a single one client. I gave them two alternatives:
(a) "After the initial acceptance period, we'll fix all bugs for free...but of course you need to pay my team for 5-man-weeks of testing and QA time so that we can both be assured it's perfect first", OR
(b) "You'll pay us time and materials to fix any significant bugs you find, but we'll only charge you for 5-man-days of testing and QA time beforehand and you'll work with us to discover any others we missed as you use the software."
Needless to say, not being stupid, they took option (b) and we probably only ended up charging them for a few minor fixes.
A software bug is not "a mistake". It's an inevitable part of the process, one that can be mitigated by good design, good coding, good management, and good testing. But all of those things take time and money. There's no magic zero-cost shortcut to perfection in any non-trivial project.
Replying to my own posting. Reading later comments above, the phosphor-pumping step (necessary to produce pleasant-looking "white" light rather than monochromatic base-LED light) cuts into the efficiency greatly. As in, "white" LEDs are a mix of monochromatic LEDs and phosphors that re-emit their light in more broad-spectrum and more complex ways. At an efficiency cost, of course. Hadn't considered that - makes sense.
I do have a few LEDs in my house and like them a lot. But then other than not being dimmable, I didn't mind CFLs much either - the "not fully bright immediately" thing was to me an advantage, not a disadvantage, especially before morning coffee.
Curious. A straight technical question I'm not quite getting:
We've long been told incandescent bulbs are barely 2% efficient - that 98% of their output is "heat" (meaning any electromagnetic energy not in the visible spectrum) and only 2% is "light" (meaning EM in the visible range). Kewl, fine.
So if a 100W-equivalent LED-based bulb consumes (for the sake of round numbers) 25 watts, it's four times as efficient as the equivalent incandescent, meaning 8% overall efficiency (4x2% = 8%), right? But I thought LEDs (the diodes themselves) were supposed to be 20-50% efficient, not 8%? I know there are supporting electronics and all, but why the huge discrepancy between the LEDs themselves and their instantiation in a consumer fixture?
Perhaps someone with actual experience in the field can enlighten (ahem) me on this.
Overall, it was a great day for Google, and Wall Street got that immediately: GOOG is up about 1.75% on the day, and ORCL down by a similar amount. So clearly the people with money on the line wouldn't agree with TFA's headline here. (Which, to be fair, is how most of the uninformed news media coverage is spinning it, so Slashdot is hardly alone in getting it wrong.
Nerd fail. Was actually pinging from the wrong shell, which includes the latency of my laptop's wireless link. So the actual CenturyLink DSL latencies are 5ms or so less for me: 15-20ms to the local Qwest.net DNS server, 70-75ms to Google DNS.
We now return you to more competent geek programming.
+1 on this. Without more information from the OP about the test nor its destination, it's nearly impossible to draw any conclusions.
/tsg/
FWIW, I have CenturyLink 12.0mb/s DSL in the Seattle metro area. I just pinged their (CenturyLink's) local DNS server (205.171.3.x) from my router, and have latency consistently in the 20-25ms range - which I consider perfectly fine. (traceroute shows five hops total, bwt, all in the qwest.net network).
Pinging Google's public DNS server, outside of qwest.net at 8.8.8.8, gives 75-80ms latency. Again, just fine in my book.
.
I think GoDaddy sucks large avian eggs on a variety of levels (morally and technically); I try hard to convince any client who comes to me with domains or hosting there to switch to something else - anything else. (I usually recommend Pairnic.com as a registrar and Pair.com for hosting, but other great options exist.)
</disclosure>
That said, I transferred a client's domain from GoDaddy just last night (to Pairnic) and the process was exceedingly quick and smooth. Since GoDaddy allows you to log in an "approve" a transfer in process, the whole thing was done in 20 minutes from start to finish. Never had one go that fast, so I have to give credit to GoDaddy where it's due here. At least in this case, they were not putting up any roadblocks.
But yes, I'm glad to be almost completely rid of them. Even if Danica is smokin' hot.
Just 'cuz I was curious, and it has some peripheral bearing on the question - assuming 19.816 gm/cm^3 for the density of Pu (more than lead) and also assuming (since it's the UK) we're talking "tons" = metric tonnes = 1000kg = 10^6 gm -
/TSG/
87 x 10^6 gm / 19.816 gm/cm^3 = 4.39 x 10^6 cm^3 = 4.39 m^3.
4.39 cubic meters is a single cube 1.637 meters on a side (or a little more than 5 feet/side, for us backward Yanks). More or less the size of a smallish SUV, yes?
Of course their Pu isn't, one hopes, stored all in one solid cube, which would probably exceed critical mass by some large factor. But still, it's not a massive physical quantity of material you're talking about here.
Mod parent up - it's correct.
The article is completely wrong in relation to CentOS (which aims for 100% binary compatibility to RH). AFAIK, they don't care which patches were applied by RH because they're duplicating the kernel down to the last byte. As you say, they'll just compile the tarball and off she goes.
The article is correct as far as an entity like Oracle is concerned, which aims to put in its own additions and "improvements".
I'm of two minds about whether RH is evil or prudent to do this, but on balance I've got no lost love for Larry Ellison, so I give RH the benefit of the doubt on this one.
Just an FYI: reading through the transcript I kept seeing things like "NOUN 37" and "VERB 12" - I thought these might be redactions for national security or censorship of Very Bad Words (ala the Nixon White House tapes and "expletive deleted" - but I'm dating myself to know about that). But they actually seem to be the way the internal shipboard guidance computer was controlled, with two part commands, one being an action (not surprisingly, "VERB yy") and one being an object to be acted upon ("NOUN xx"). Details here:
http://history.nasa.gov/afj/compessay.htm
Interestingly, this is not at all unlike how the original Fortran code for ADVENT (the seminal "Collossal Cave Adventure") was architected, even down to the terminology used.
As fellow author, Will Belegon, noted, if Amazon is going to start pulling books with incest in them: "I just re-read Genesis 19: 30-38 and realized that Lot's daughters got him drunk, had sex with him and bore sons. I demand you follow your clear precedent and remove The Bible from Kindle."
Based on preliminary information I heard from the Facebook launch announcement today [...] users will not have the ability to declare chats or related conversations to be "off the record" -- everything will apparently be recorded. Individual users will have the ability to archive or delete their own copies of transcripts, but it appears that there is explicitly not a functionality similar to Google's "off the record" chat feature, which permits users to declare that their conversations with given individuals should not be routinely preserved. "It just didn't make sense for us," were pretty much the words that Zuckerberg used in response to a question on this topic."
http://www.google.com/buzz/lauren4321/Am7dw5mhpRi/Facebooks-new-chat-email-feature-apparently
(said items probably a hell of a lot more useful than the actual Windows 1.0 software ever was...)
All by way of saying that .484% adds up to a pretty tidy sum when levied on Microsoft's gross licensing revenue, worldwide.
At the time of my post, there were only ~10 comments, and many (including the article summary itself) were not positive. Glad to know that others share a good opinion of it.
...who thought this ad was pretty damn brilliant? Low-key, sure, but also sweet, memorable, and focused on the product/service itself rather than hype and glitz? I thought it promoted the both the company's values and the value of what they provide to their customers extraordinarily well.
Maybe I'm just not cynical enough, but it sure gets my vote. /tsg/
Worth reading Lauren Weinstein's blog post take on this - trenchantly dead-on, as usual:
Murdoch's Folly: Block Google & Kill Fair Use -- Plus a Nasty Truth
http://lauren.vortex.com/archive/000633.html