For starters, let me just say that I don't expect a very unbiased view from someone whose username is gillbates.
Now that that's out of the way...
Saying that "source code gives *you* options" is a nice aphorism, but in real life scenarios I don't think it means that much. Sure, we have the option of hiring any number of Linux/wireless engineers to fix Vendor X's (VX) bugs. Why would we want to do that when we're pushing VX* to fix their own bugs?!
VX has had a team of half a dozen engineers or so working on the driver for the last 4 or 5 months. In a different country. How much time is it going to take for new software guys to come up to speed on their work? Way too long.
The bottom line is that no man is an island, and neither is any company. Having source is nice, but I doubt even the biggest Linux company (Red Hat? Novell?) owns their entire toolchain in source. Everyone relies on their vendors, and the relationships we establish with our vendors represent a significant portion of our success. That being said, we implicitly trust our relationship with MS and WinCE right now, and while our relationship with VX is still very good, this latest problem has put a black mark on their record.
The funny thing is, I really do agree with your statement in principle. I'm just pointing out that in practice some times having the source doesn't really mean anything.
What you say about hiring is really true, at least in the Midwest. CE developers are h-a-r-d to find. Maybe they're all on the left and right coasts.
While I don't really disagree with you about Linux/CE, I'll present the same story but from a different standpoint. We use 3 main OSes for development: Linux, VxWorks and WinCE. The other day we received a code drop for a Linux wireless client from a major vendor everyone has heard of. It was a piece of crap. That's great that it was open source, but it was a piece of crap bit of open source. And when you're knee-deep in a project that's already late you don't have time to learn 30K lines of new source and figure out what they screwed up.
Meanwhile (at the same time), we uncovered three problems with CE:
- the DHCP client wasn't waiting long enough for a server response in a particular configuration
- the USB8023 driver wasn't getting enough data from a site survey on an RNDIS device due to an improperly sized buffer
- adding a second IP address to an adapter was producing some wacky behavior (that's the technical term:-)
Within two weeks, MS engineers had diagnosed the 2nd and 3rd problems (we figured out the first) and sent out hotfixes that were later addressed by a QFE (which, for CE, is like a mini service pack). If we had had full source we would've needed an engineer that was intimately familiar with the USB8023 driver and RNDIS, an engineer that was intimately familiar with the way adapters bind to a protcol in CE, and someone with a rudimentary knowledge of DHCP. Your company may be lucky enough to have all on hand, but most companies probably do not.
Three weeks later, Vendor X is still working on those Linux bugs. It doesn't really help that the damn thing is open source.
I realize that these are both anedcotal examples, but maybe it will be refreshing for some to see a viewpoint on/. other than "Linux is good, MS is bad." It's all much more complicated than that.
Two endnotes: - The CE source for the DHCP client is already distributed. It helped to have it on hand. - The kernel is a very small part of CE. Having the kernel source won't help most people's problems, because most people aren't making kernel changes. Most CE work is done at the driver level, the OAL (the OEM adaptation layer that sits between the hardware and the OS), and the application level.
No one is going to use 32K processes. Most people are just interested in process #'s 33 - 40 that everyone was having to write services for instead. (In CE a service doesn't take up a process slot.)
I see the moderators are giving points to anyone with a keyboard these days...
A few minor corrections: - Platform Builder doesn't cost $3,000. It costs $995. And that's after a 120-day free eval. - PB used to be its own toolchain and IDE. Now it's been integrated into VS2005. (I'm not sure how MS plans on doing the eval, though my beta was good for 180 days.) - WinCE core licenses (no Pocket IE, no Media Player) are about $3. Upgrading to Professional (including both IE and WMP) bump it up to about $13. This is actually *less* than an MS-DOS license. - "enough code to flash CE 6 onto a CE 5 device" Uh, what are you talking about? - Windows Mobile, which is a customer of the Windows CE group, is what your mobile handheld device runs. It's its own OS. So you're not just going to "drop the new kernel" in there. Most PDA OEM's, from what I've heard, aren't going to support a CE5 -> CE6 upgrade.
From the article: The replica goes on general display at Bletchley Park on September 23.
Hopefully they'll do more than just display it. I would love to hear the ticking sound of one running. (Incidentally, that's where the name "bombe" comes from.)
The most vocal group of people demanding a specific product and promising to buy it will usually not actually buy what they say they want. They are just looking to get something they can't have, and when they can have it, they don't want it anymore.
One of the best Slashdot quotes I've ever seen. Thanks.
Perhaps by leaving a few typos on the site, I am making their day a little easier!
As seen in a Park District publication I got in the mail the other day:
"If you find a mistake in this publication, please consider that they are there for a purpose. We publish something for everyone and some people are always looking for mistakes."
Despite the fact that I see this as proof that the scientific method works (they've rooted out phony research)
This scientific method works, but not because of the reason you gave. The scientific method works because one scientist's experimental results should be reproducible by other scientists.
In this case Dr. Hwang's research wasn't debunked by experimentation, but by his own co-authors who basically said that he just lied and made up stuff.
Professor Jochem Hauser, one of the scientists who put forward the idea, told The Scotsman that if everything went well a working engine could be tested in about five years.
This is the kind of stuff I used to get excited about reading PopSci when I was 12. By now I've seen a little more about how the world works.
Since when does everything go well on an experimental scientific project? That's just something the backers say to get funding. If this works at all, maybe it'll work in 20 years or so.
Wasn't $3 billion...and wasn't a typo, either
on
The 3 Billion Dollar Typo
·
· Score: 3, Informative
Apparently not even the submitters are reading the articles these days...oh well.
For anyont who RTFA'd, 610,000 shares at 1Y were offered, not bought. The error so far has cost about $224 million, and may eventually cost $250 million. That's a huge cost for a trader error, but it's not $3 billion.
And I don't think this qualifies as a typo. How about "data entry error"? Or how about software bug, since the number of shares sold was more than the number of available shares.
It's just ludicrous that some people still don't (or won't) believe that global warming is real. Things like this are direct proof. As long as big companies like ExxonMobil, PetroChina and those perfume companies keep spraying their products into the atmosphere this will only get worse.
If the US had signed the Koyto treaty this wouldn't have happened!
Quoting from an original spec I saw: - 500 MHz CPU - 1 GB Memory - WiFi - Megapixel LCD - 4 USB Ports - Battery and Cool Hand Crank - Linux
I really doubt that they can produce them for $100. Let's take a look at the BOM:
1. 500 MHz CPU w/ mainboard (even some generic Via or clone): $35 (and that sounds low to me) 2. 1 GB Memory (not sure what kind of memory we're talking about here, but 1GB flash costs about $50. In '06 that will drop to $25, but with flash storage doubling or quadrupling every year, who will want just 1GB in '06?): $50 (wear-leveling like Intel's Strataflash) 3. Wifi barebones chipset: $10 4. 1 Megapixel LCD: $35 (for whatever Mr. Negroponte has in mind) 5. 4 USB Ports (attached to motherboard): $0 6. Battery and cool hand crank: $20 7. Linux w/ special development: $0 (assuming their development costs are donated)
Things they didn't mention:
8. Fancy silvery yet rubber enclosure with custom hinge: $15 9. Modular keyboard (to support int'l character sets): $5 10. Touchpad-like device: $5 11. Modem? GSM Modem? Maybe it's integrated with the motherboard?: $0 13. Misc. (power-cord/strap, stylus, manual?, buttons by screen, etc.): $10 14. Assembly: $10 (if they can find a sweatshop to do it somewhere cheap like say Cambodia)
These are pretty low component costs. Maybe they're looking for manufacturers to sell these components at no profit. But who is going to handle the operations, QA, warranties (?), support, software updates/patches rollouts? These things all cost money!
My subtotal costs about $195, and that still seems pretty cheap.
I have a feeling Mr. Negroponte is "throwing it out there" to see what the response will be. Maybe he'll get some donations of materials/time/R&D/etc. Maybe he'll have to pare down the features. In the end the computer that he can build for $100 will be much less than what he's showing us right now.
Facial problems are the most humiliating of all
on
First Face Transplant
·
· Score: 1
I watched an amazing show the other day about a boy from Sumatra (?) whose face was being destroyed by a face-eating tumor. I believe this is the correct link:
Description:
Witness groundbreaking surgery to remove the largest facial tumors ever recorded, giving a five-year-old boy a chance for a normal life. The tumors, caused by a rare congenital disorder called fibrous dysplasia, stem from abnormal bone growth.
Who came up with the "2 terabytes uploaded to Usenet every day" figure that the article throws around? Is that one of those astroturf'd numbers? Whose servers are storing 730TB/year for free?!
Hemos wrote:
"we wrote at least one new entry for 190 straight days, including ones when one or both of us was tired, on vacation or not particularly inspired."
Heh. Boy, does this refrain ever sound familiar.
Oh yeah, sure. With the exception of Taco's diatribe against Blizzard last week, how much content do Slashdot editors* write in a week? Maybe 50 words?
* for all values of editors = submission moderators
Ars uses some free-lance writers. You probably won't get the same quality as a Caesar or Hannibal article, but most of them are quite good. And they still have to abide by the Ars writing guidelines. The editing staff at Ars is pretty good at make sure things stay up to par, although I'll admit that the example you listed is needlessly wordy.
FWIW, this article was discussed for a week on the Ars writer's forum before it was posted, so some pretty good eyeballs got a chance to check it over before it was posted on/.
Slashdot Mirror
For starters, let me just say that I don't expect a very unbiased view from someone whose username is gillbates.
Now that that's out of the way...
Saying that "source code gives *you* options" is a nice aphorism, but in real life scenarios I don't think it means that much. Sure, we have the option of hiring any number of Linux/wireless engineers to fix Vendor X's (VX) bugs. Why would we want to do that when we're pushing VX* to fix their own bugs?!
VX has had a team of half a dozen engineers or so working on the driver for the last 4 or 5 months. In a different country. How much time is it going to take for new software guys to come up to speed on their work? Way too long.
The bottom line is that no man is an island, and neither is any company. Having source is nice, but I doubt even the biggest Linux company (Red Hat? Novell?) owns their entire toolchain in source. Everyone relies on their vendors, and the relationships we establish with our vendors represent a significant portion of our success. That being said, we implicitly trust our relationship with MS and WinCE right now, and while our relationship with VX is still very good, this latest problem has put a black mark on their record.
The funny thing is, I really do agree with your statement in principle. I'm just pointing out that in practice some times having the source doesn't really mean anything.
* VX != VxWorks, for the record
What you say about hiring is really true, at least in the Midwest. CE developers are h-a-r-d to find. Maybe they're all on the left and right coasts.
:-)
/. other than "Linux is good, MS is bad." It's all much more complicated than that.
While I don't really disagree with you about Linux/CE, I'll present the same story but from a different standpoint. We use 3 main OSes for development: Linux, VxWorks and WinCE. The other day we received a code drop for a Linux wireless client from a major vendor everyone has heard of. It was a piece of crap. That's great that it was open source, but it was a piece of crap bit of open source. And when you're knee-deep in a project that's already late you don't have time to learn 30K lines of new source and figure out what they screwed up.
Meanwhile (at the same time), we uncovered three problems with CE:
- the DHCP client wasn't waiting long enough for a server response in a particular configuration
- the USB8023 driver wasn't getting enough data from a site survey on an RNDIS device due to an improperly sized buffer
- adding a second IP address to an adapter was producing some wacky behavior (that's the technical term
Within two weeks, MS engineers had diagnosed the 2nd and 3rd problems (we figured out the first) and sent out hotfixes that were later addressed by a QFE (which, for CE, is like a mini service pack). If we had had full source we would've needed an engineer that was intimately familiar with the USB8023 driver and RNDIS, an engineer that was intimately familiar with the way adapters bind to a protcol in CE, and someone with a rudimentary knowledge of DHCP. Your company may be lucky enough to have all on hand, but most companies probably do not.
Three weeks later, Vendor X is still working on those Linux bugs. It doesn't really help that the damn thing is open source.
I realize that these are both anedcotal examples, but maybe it will be refreshing for some to see a viewpoint on
Two endnotes:
- The CE source for the DHCP client is already distributed. It helped to have it on hand.
- The kernel is a very small part of CE. Having the kernel source won't help most people's problems, because most people aren't making kernel changes. Most CE work is done at the driver level, the OAL (the OEM adaptation layer that sits between the hardware and the OS), and the application level.
No one is going to use 32K processes. Most people are just interested in process #'s 33 - 40 that everyone was having to write services for instead. (In CE a service doesn't take up a process slot.)
I see the moderators are giving points to anyone with a keyboard these days...
A few minor corrections:
- Platform Builder doesn't cost $3,000. It costs $995. And that's after a 120-day free eval.
- PB used to be its own toolchain and IDE. Now it's been integrated into VS2005. (I'm not sure how MS plans on doing the eval, though my beta was good for 180 days.)
- WinCE core licenses (no Pocket IE, no Media Player) are about $3. Upgrading to Professional (including both IE and WMP) bump it up to about $13. This is actually *less* than an MS-DOS license.
- "enough code to flash CE 6 onto a CE 5 device" Uh, what are you talking about?
- Windows Mobile, which is a customer of the Windows CE group, is what your mobile handheld device runs. It's its own OS. So you're not just going to "drop the new kernel" in there. Most PDA OEM's, from what I've heard, aren't going to support a CE5 -> CE6 upgrade.
Hopefully they'll do more than just display it. I would love to hear the ticking sound of one running. (Incidentally, that's where the name "bombe" comes from.)
One of the best Slashdot quotes I've ever seen. Thanks.
It's thanks to those "love letters to God" you disdain that the palimpset survived at all.
Ah yes, you've presented both sides of the fence so well.
It would seem to me that what put Oracle's customers at risk was the security flaw itself, not someone's disclosure of it.
Perhaps more importantly, I don't think that anything that physics does breaks the laws of God.
(Go ahead, mod me to hell. It's not like the moderators are going to give this topic a fair hearing on /.)
As seen in a Park District publication I got in the mail the other day:
"If you find a mistake in this publication, please consider that they are there for a purpose. We publish something for everyone and some people are always looking for mistakes."
So you're saying I'm screwed in the 14th year?
This scientific method works, but not because of the reason you gave. The scientific method works because one scientist's experimental results should be reproducible by other scientists.
In this case Dr. Hwang's research wasn't debunked by experimentation, but by his own co-authors who basically said that he just lied and made up stuff.
This is the kind of stuff I used to get excited about reading PopSci when I was 12. By now I've seen a little more about how the world works.
Since when does everything go well on an experimental scientific project? That's just something the backers say to get funding. If this works at all, maybe it'll work in 20 years or so.
(It's posts like these that almost make me like the place again. ;-)
(BTW, you might consider adding your info to this Wikipedia article: http://en.wikipedia.org/wiki/Communication_during_ the_September_11%2C_2001_attacks Or, if you're interested in me piecing the relevant parts into the article, please let me know. It would be nice to have the information not scroll off /.'s horizon.)
For anyont who RTFA'd, 610,000 shares at 1Y were offered, not bought. The error so far has cost about $224 million, and may eventually cost $250 million. That's a huge cost for a trader error, but it's not $3 billion.
And I don't think this qualifies as a typo. How about "data entry error"? Or how about software bug, since the number of shares sold was more than the number of available shares.
If the US had signed the Koyto treaty this wouldn't have happened!
I really doubt that they can produce them for $100. Let's take a look at the BOM:
1. 500 MHz CPU w/ mainboard (even some generic Via or clone): $35 (and that sounds low to me)
2. 1 GB Memory (not sure what kind of memory we're talking about here, but 1GB flash costs about $50. In '06 that will drop to $25, but with flash storage doubling or quadrupling every year, who will want just 1GB in '06?): $50 (wear-leveling like Intel's Strataflash)
3. Wifi barebones chipset: $10
4. 1 Megapixel LCD: $35 (for whatever Mr. Negroponte has in mind)
5. 4 USB Ports (attached to motherboard): $0
6. Battery and cool hand crank: $20
7. Linux w/ special development: $0 (assuming their development costs are donated)
Things they didn't mention:
8. Fancy silvery yet rubber enclosure with custom hinge: $15
9. Modular keyboard (to support int'l character sets): $5
10. Touchpad-like device: $5
11. Modem? GSM Modem? Maybe it's integrated with the motherboard?: $0
13. Misc. (power-cord/strap, stylus, manual?, buttons by screen, etc.): $10
14. Assembly: $10 (if they can find a sweatshop to do it somewhere cheap like say Cambodia)
These are pretty low component costs. Maybe they're looking for manufacturers to sell these components at no profit. But who is going to handle the operations, QA, warranties (?), support, software updates/patches rollouts? These things all cost money!
My subtotal costs about $195, and that still seems pretty cheap.
I have a feeling Mr. Negroponte is "throwing it out there" to see what the response will be. Maybe he'll get some donations of materials/time/R&D/etc. Maybe he'll have to pare down the features. In the end the computer that he can build for $100 will be much less than what he's showing us right now.
http://health.discovery.com/tvlistings/series.jsp? series=109445&gid=0&channel=DHC
Description:
Witness groundbreaking surgery to remove the largest facial tumors ever recorded, giving a five-year-old boy a chance for a normal life. The tumors, caused by a rare congenital disorder called fibrous dysplasia, stem from abnormal bone growth.
Some of the French youth are extremely happy about getting their hands on Firefox.
This article has recently been linked from Slashdot.
Please keep an eye on the page history for errors or vandalism.
Who came up with the "2 terabytes uploaded to Usenet every day" figure that the article throws around? Is that one of those astroturf'd numbers? Whose servers are storing 730TB/year for free?!
"we wrote at least one new entry for 190 straight days, including ones when one or both of us was tired, on vacation or not particularly inspired."
Heh. Boy, does this refrain ever sound familiar.
Oh yeah, sure. With the exception of Taco's diatribe against Blizzard last week, how much content do Slashdot editors* write in a week? Maybe 50 words?
* for all values of editors = submission moderators
FWIW, this article was discussed for a week on the Ars writer's forum before it was posted, so some pretty good eyeballs got a chance to check it over before it was posted on /.
Maybe the sun is getting hotter.