It's amazing how many people will accuse an innocent man of something they are guilty of themselves, and incriminate themselves in the process.
Go read and educate yourself. "Bufferbloat because RAM is getting cheaper" is exactly the problem. Too many packets are buffered, and not enough are dropped.
Almost. I think the solution is to continue to invest in infrastructure where and when needed, while returning buffers to sizes that are appropriate (there have to be some buffers).
Really, if it were possible to convince vendors and users, all we need is software patches to make buffers smaller. So what if the RAM goes unused?
Hey, I know: for devices with the CPU to burn, use the extra RAM as HTTP cache for really popular sites. Then there'd be less traffic on the backbones.
Tomato's QoS works well. It was well worth the $40 for the ASUS router so I could play games while my neighbor was downloading TV with BitTorrent. After configuring it well, I tested it extensively and got no increase in game latency even when the connection was saturated with torrents.
The issue is always going to be there. Pegged connection == FIFO queuing, absent some sort of QoS scheme.
You don't understand. The point is not that FIFOs shouldn't exist, the point is that the FIFOs are too big. If you think of the buffer sizes in milliseconds rather than bytes (buffer size / connection speed), you'll see that big buffers == big latency. Appropriately-sized--smaller--buffers will also fill up, but being much shorter (in terms of how long they take to empty), latency won't skyrocket, because a new packet won't have to wait seconds to make it from the back to the front of the queue.
And it does, even with QoS. All you do with QoS is force the buffering to happen on equipment that you control rather than equipment your ISP controls. In this manner you can ensure that time sensitive packets (interactive VPNs, VoIP, etc.) don't sit in the queue behind someone's Windows Update download.
As was mentioned, you can only truly effect QoS on your upstream data. You can affect ACKs going upstream, but if the node at the other end wants to saturate your downstream connection, you can't stop it from doing so. And since you can't effect QoS on downstream data, your VOIP/etc. may have good upstream latency, but downstream latency and RTT will suffer, because those incoming VOIP packets will wait their turn in the queue just like the Windows Update download's packets.
You're right. But every new or updated device that has smaller, appropriate buffers will help to solve the problem. Over time, as devices are updated or replaced, the problem could be eradicated, almost like a vaccine gradually wiping out a virus.
I think it's a bit silly to add yet more layers of complex software to cover over a problem that could simply be fixed by using less in the first place. It's like over-engineering a solution to an over-engineered solution!
It's like building larger and more complex recycling plants to compensate for huge, overflowing landfills while ignoring the obvious solution of making less in the first place (smaller, more efficient, reusable packaging, etc). Then there wouldn't be as much of a need for recycling. "REDUCE, REUSE, recycle."
Do the simple, obvious fix--that's actually undoing what shouldn't have been done in the first place--before creating a complex workaround for the workaround!
or the total re-architecture of millions of consumer devices to remove buffering
Total re-architecture? A tiny software patch could shrink the buffers on any device, regardless of how much RAM a device has. The problem is getting vendors to make the patches and users to install them--it's that simple, and that difficult.
The problem here isn't that the buffers are big, it's that you don't have the bandwidth needed for your traffic load, period.
What would be enough bandwidth for a given load? If a user is downloading an ISO of a Linux distro, that can't be downloaded in mere seconds, period. That's going to saturate the connection at one end or the other, period. And thus, it's probably going to end up filling buffers at one end or the other, leading to cascading latency and failure of interactive apps and really frustrated people. This is the 21st century, after all. Even when I was using a 56k modem, I sometimes had less latency when downloading a file and browsing the web at the same time. Things should be getting better, not worse.
You seem to be suggesting that if we could just have enough bandwidth, buffers wouldn't matter. But we're not talking so much about backbones, we're talking about end-to-end paths with bottlenecks. We can't just add more bandwidth to everyone's connection--we have to fix the buffers.
QoS is not the solution to latency caused by bufferbloat--it's like putting a square peg into a round hole: it will either not fit or still leak.
If buffers were sized appropriately, latency for interactive apps wouldn't be ruined by a single HTTP download using all available bandwidth. QoS can only be controlled at each end separately, which means end users giving up some of their bandwidth for the sake of creating their own bottleneck to control the buffers. Users shouldn't have to artificially limit their bandwidth--properly-implemented networks can handle simultaneous connections without multi-second latency. TCP was designed and improved to handle the problem itself, but buffers have destroyed the very mechanism TCP uses to handle the problem.
QoS might be unnecessary for VOIP-type apps if it wasn't for bufferbloat. QoS should provide a small boost to interactive performance, at most, because latency shouldn't be that bad in the first place.
Think of it this way: we could either create additional, complicated mechanisms to manage packet queues and explicitly notify nodes of congestion...or we could just use smaller buffers and let TCP do what it used to do already, what it was engineered to do. We could add more layers of software to compensate for hardware problems...or we could remove the unnecessary hardware buffers and let the existing software work the way it used to. Which is simpler, and cheaper?
Arkansas is no different than any other place in the U.S. You can find rednecks, hicks, and hillbillies, as well as intelligent, educated, even wealthy people in every state in the Union. It's just silly to make "Arkansaw" jokes; not even funny. Urban Arkansas is like urban anywhere, and rural Arkansas is like rural anywhere.
Fords are so poorly designed inside and out, technically and ergonomically, I can't understand why anyone would buy one when there are alternatives like Honda and Toyota available. The quality of workmanship is also so poor compared to brands like Honda and Toyota--again, why? One can buy a good, used Honda for less money than a new Ford and come away with a better car, one that's more reliable, more durable, better performing, and more pleasant to operate.
Is it just a matter of poor taste? Not realizing that there are great alternatives? Masochism?
I would love to support an all-American auto company with my business, but I won't reward shoddy engineering, design, workmanship, and service with my money. Besides, Hondas have been built in the U.S. for a long time now, so they provide plenty of American jobs.
It can be likened to a government and it's people. If the government gives its people bread and circuses, the masses may not care whether they have a say in the government which rules over them. But the day may come when the party ends, and then the people will be powerless. It's important to exercise freedom before it is too late to acquire it.
You may think this sounds grandiose and silly, but I think much truth lies in these ideas, and much naiveté and ignorance in dismissing them. Time will tell.
I would argue that it does affect 99% of users (or whatever large majority you choose), because when someone takes that code and starts a new project (Netscape->Firefox, Konqueror->WebKit, WebKit->Chrome, Linux->Debian, Debian->Ubuntu, etc), it goes on to affect many, many people. Just because 99.9% of end users won't end up writing or compiling code is not the point at all. It's the people that end up using and benefiting from the software.
If we just gave up and used closed-source software, we'd be stuck with IE, maybe Opera (probably without extensions), and even the Mac OS might not be where it is today, based on BSD (free software often ends up benefiting proprietary software, depending on the license). Not to mention how by far the majority of the Internet runs on Linux--I shudder to think what the Internet would be like if we were stuck with IIS and the like!
No offense intended--but I think to say that to say that open-source software doesn't matter is incredibly short-sighted and naive. History and the present have shown otherwise. It saddens me whenever I hear people downplay its significance and blindly gulp down their proprietary punch. Sustainable, future-oriented, and freedom-oriented computing lies with free software. We need to support its development with, if nothing else, our "feet", lest it slip into obscurity due to the ignorance of the masses.
I think in the end the best policy may be to use a reliable webmail service, such as Gmail, that is regularly backed-up to local storage that's in your control. offlineimap can take care of that pretty well.
As for Yahoo, I wouldn't be surprised if their email offerings continue to dwindle and crustify, like all their other services have. Gmail seems solid for the foreseeable future, but I'm sure even it won't last forever. As long as we keep copies of our data, we can move on when services wither.
You have some valid points, but I completely disagree about the last one. A browser could do many things besides phone home in plain text. And all it'd take is one malicious employee or cracked Opera server to allow malicious code into the binary that might not be found for a long time.
If you want to keep your emails "out of the cloud," then put your email on any service that does POP3 and download it into Thunderbird, KMail, mutt, etc. Make sure you make regular backups, and forget about accessing your email from other computers without complex and fragile replication schemes.
If you want webmail, you can import to Gmail with IMAP or other utilities. The UI is a million times better, and you don't have to filter unbelievable animated ads, and tolerate text ads that you can't filter.
If you truly trust Yahoo to protect your privacy and keep your account secure more than you trust Google, well, I think you're...confused. No company is altruistic, but Google is far closer to it than Yahoo ever has been or ever will be. Google is far more user-focused than Yahoo. Google is also more competent than Yahoo on a technical level. Yahoo's starting to fall apart, while Google keeps flourishing.
Oh, and by the way...Yahoo will delete your account and all your email if something happens and you don't log in for 6 months.
Richard Stallman's mindset may be overkill at times, but he's basically right. I don't want to use a closed-source browser for the same reason I don't want to use a closed-source OS, among other reasons.
I think determining which browser is more secure is more complex than simply checking a vulnerabilities list. There are all sorts of other variables and factors that play into the number, which may or may not be representative of reality.
Opera seems like a good bunch of people, so I'm not especially worried about their doing anything malicious. But when there are excellent open-source alternatives, why even bother?
If you truly don't care about Free Software, then you have missed the point, and I wouldn't expect you to care about open-source browsers. I think that Free Software is better in principle, in theory, and in practice. It's more future-proof than being locked-in to any company's closed-source software. And as time goes on, it will become even more important, because the power of marketing and media and government conglomerates will continue to grow, and Free Software is nearly the only way to retain user freedoms and privacy, and control over one's own hardware, software, and data.
I don't understand why any knowledgeable user would want to use a closed-source browser nowadays. Any specific benefit Opera might have over another particular browser would be outweighed by the drawbacks, IMO.
Why even make a separate Gmail account? You can use nested labels.
But really, this is what sites like Evernote, Springpad, Diigo, etc. are for. They even have smartphone apps, and have web clippers to save straight from the browser.
"I haven't figured how to back up Yahoo emails because POP3 isn't free."...switch to Gmail?
Seriously, I cringe when I walk into a computer lab and see people using Yahoo mail. Why would anyone want to use it instead of Gmail? It should be self-evident--I shouldn't have to post screenshots.
Slashdot Mirror
It's amazing how many people will accuse an innocent man of something they are guilty of themselves, and incriminate themselves in the process.
Go read and educate yourself. "Bufferbloat because RAM is getting cheaper" is exactly the problem. Too many packets are buffered, and not enough are dropped.
Almost. I think the solution is to continue to invest in infrastructure where and when needed, while returning buffers to sizes that are appropriate (there have to be some buffers).
Really, if it were possible to convince vendors and users, all we need is software patches to make buffers smaller. So what if the RAM goes unused?
Hey, I know: for devices with the CPU to burn, use the extra RAM as HTTP cache for really popular sites. Then there'd be less traffic on the backbones.
Tomato's QoS works well. It was well worth the $40 for the ASUS router so I could play games while my neighbor was downloading TV with BitTorrent. After configuring it well, I tested it extensively and got no increase in game latency even when the connection was saturated with torrents.
The issue is always going to be there. Pegged connection == FIFO queuing, absent some sort of QoS scheme.
You don't understand. The point is not that FIFOs shouldn't exist, the point is that the FIFOs are too big. If you think of the buffer sizes in milliseconds rather than bytes (buffer size / connection speed), you'll see that big buffers == big latency. Appropriately-sized--smaller--buffers will also fill up, but being much shorter (in terms of how long they take to empty), latency won't skyrocket, because a new packet won't have to wait seconds to make it from the back to the front of the queue.
And it does, even with QoS. All you do with QoS is force the buffering to happen on equipment that you control rather than equipment your ISP controls. In this manner you can ensure that time sensitive packets (interactive VPNs, VoIP, etc.) don't sit in the queue behind someone's Windows Update download.
As was mentioned, you can only truly effect QoS on your upstream data. You can affect ACKs going upstream, but if the node at the other end wants to saturate your downstream connection, you can't stop it from doing so. And since you can't effect QoS on downstream data, your VOIP/etc. may have good upstream latency, but downstream latency and RTT will suffer, because those incoming VOIP packets will wait their turn in the queue just like the Windows Update download's packets.
I think your first line said it all. :)
Netalyzr does exactly this, among other things. It will tell you how big your perceived buffers are along your network path, upstream and downstream.
Good post, my friend. "All progress depends on the 'fool.'" "Fools" like you are the leaders we need. (I mean "fool" in a positive way; I'm with you.)
You're right. But every new or updated device that has smaller, appropriate buffers will help to solve the problem. Over time, as devices are updated or replaced, the problem could be eradicated, almost like a vaccine gradually wiping out a virus.
I think it's a bit silly to add yet more layers of complex software to cover over a problem that could simply be fixed by using less in the first place. It's like over-engineering a solution to an over-engineered solution!
It's like building larger and more complex recycling plants to compensate for huge, overflowing landfills while ignoring the obvious solution of making less in the first place (smaller, more efficient, reusable packaging, etc). Then there wouldn't be as much of a need for recycling. "REDUCE, REUSE, recycle."
Do the simple, obvious fix--that's actually undoing what shouldn't have been done in the first place--before creating a complex workaround for the workaround!
or the total re-architecture of millions of consumer devices to remove buffering
Total re-architecture? A tiny software patch could shrink the buffers on any device, regardless of how much RAM a device has. The problem is getting vendors to make the patches and users to install them--it's that simple, and that difficult.
I think "BanCan" sounds all right. :)
The problem here isn't that the buffers are big, it's that you don't have the bandwidth needed for your traffic load, period.
What would be enough bandwidth for a given load? If a user is downloading an ISO of a Linux distro, that can't be downloaded in mere seconds, period. That's going to saturate the connection at one end or the other, period. And thus, it's probably going to end up filling buffers at one end or the other, leading to cascading latency and failure of interactive apps and really frustrated people. This is the 21st century, after all. Even when I was using a 56k modem, I sometimes had less latency when downloading a file and browsing the web at the same time. Things should be getting better, not worse.
You seem to be suggesting that if we could just have enough bandwidth, buffers wouldn't matter. But we're not talking so much about backbones, we're talking about end-to-end paths with bottlenecks. We can't just add more bandwidth to everyone's connection--we have to fix the buffers.
QoS is not the solution to latency caused by bufferbloat--it's like putting a square peg into a round hole: it will either not fit or still leak.
If buffers were sized appropriately, latency for interactive apps wouldn't be ruined by a single HTTP download using all available bandwidth. QoS can only be controlled at each end separately, which means end users giving up some of their bandwidth for the sake of creating their own bottleneck to control the buffers. Users shouldn't have to artificially limit their bandwidth--properly-implemented networks can handle simultaneous connections without multi-second latency. TCP was designed and improved to handle the problem itself, but buffers have destroyed the very mechanism TCP uses to handle the problem.
QoS might be unnecessary for VOIP-type apps if it wasn't for bufferbloat. QoS should provide a small boost to interactive performance, at most, because latency shouldn't be that bad in the first place.
Think of it this way: we could either create additional, complicated mechanisms to manage packet queues and explicitly notify nodes of congestion...or we could just use smaller buffers and let TCP do what it used to do already, what it was engineered to do. We could add more layers of software to compensate for hardware problems...or we could remove the unnecessary hardware buffers and let the existing software work the way it used to. Which is simpler, and cheaper?
Arkansas is no different than any other place in the U.S. You can find rednecks, hicks, and hillbillies, as well as intelligent, educated, even wealthy people in every state in the Union. It's just silly to make "Arkansaw" jokes; not even funny. Urban Arkansas is like urban anywhere, and rural Arkansas is like rural anywhere.
At least try to be creative.
One question: Why?
Fords are so poorly designed inside and out, technically and ergonomically, I can't understand why anyone would buy one when there are alternatives like Honda and Toyota available. The quality of workmanship is also so poor compared to brands like Honda and Toyota--again, why? One can buy a good, used Honda for less money than a new Ford and come away with a better car, one that's more reliable, more durable, better performing, and more pleasant to operate.
Is it just a matter of poor taste? Not realizing that there are great alternatives? Masochism?
I would love to support an all-American auto company with my business, but I won't reward shoddy engineering, design, workmanship, and service with my money. Besides, Hondas have been built in the U.S. for a long time now, so they provide plenty of American jobs.
Your words are so generic they are meaningless.
It can be likened to a government and it's people. If the government gives its people bread and circuses, the masses may not care whether they have a say in the government which rules over them. But the day may come when the party ends, and then the people will be powerless. It's important to exercise freedom before it is too late to acquire it.
You may think this sounds grandiose and silly, but I think much truth lies in these ideas, and much naiveté and ignorance in dismissing them. Time will tell.
I would argue that it does affect 99% of users (or whatever large majority you choose), because when someone takes that code and starts a new project (Netscape->Firefox, Konqueror->WebKit, WebKit->Chrome, Linux->Debian, Debian->Ubuntu, etc), it goes on to affect many, many people. Just because 99.9% of end users won't end up writing or compiling code is not the point at all. It's the people that end up using and benefiting from the software.
If we just gave up and used closed-source software, we'd be stuck with IE, maybe Opera (probably without extensions), and even the Mac OS might not be where it is today, based on BSD (free software often ends up benefiting proprietary software, depending on the license). Not to mention how by far the majority of the Internet runs on Linux--I shudder to think what the Internet would be like if we were stuck with IIS and the like!
No offense intended--but I think to say that to say that open-source software doesn't matter is incredibly short-sighted and naive. History and the present have shown otherwise. It saddens me whenever I hear people downplay its significance and blindly gulp down their proprietary punch. Sustainable, future-oriented, and freedom-oriented computing lies with free software. We need to support its development with, if nothing else, our "feet", lest it slip into obscurity due to the ignorance of the masses.
An interesting post; thanks.
I think in the end the best policy may be to use a reliable webmail service, such as Gmail, that is regularly backed-up to local storage that's in your control. offlineimap can take care of that pretty well.
As for Yahoo, I wouldn't be surprised if their email offerings continue to dwindle and crustify, like all their other services have. Gmail seems solid for the foreseeable future, but I'm sure even it won't last forever. As long as we keep copies of our data, we can move on when services wither.
You have some valid points, but I completely disagree about the last one. A browser could do many things besides phone home in plain text. And all it'd take is one malicious employee or cracked Opera server to allow malicious code into the binary that might not be found for a long time.
Ok. Someday I think you will see good reasons. :)
I'm not an extremist, but someday you will be more thankful for free software than you are now.
(Though it seems to me there's already plenty of reasons to be thankful for it. Imagine if we had never had Firefox.)
Good point...but why bother to converse with an AC?
Hm. I don't quite understand your reasoning.
If you want to keep your emails "out of the cloud," then put your email on any service that does POP3 and download it into Thunderbird, KMail, mutt, etc. Make sure you make regular backups, and forget about accessing your email from other computers without complex and fragile replication schemes.
If you want webmail, you can import to Gmail with IMAP or other utilities. The UI is a million times better, and you don't have to filter unbelievable animated ads, and tolerate text ads that you can't filter.
If you truly trust Yahoo to protect your privacy and keep your account secure more than you trust Google, well, I think you're...confused. No company is altruistic, but Google is far closer to it than Yahoo ever has been or ever will be. Google is far more user-focused than Yahoo. Google is also more competent than Yahoo on a technical level. Yahoo's starting to fall apart, while Google keeps flourishing.
Oh, and by the way...Yahoo will delete your account and all your email if something happens and you don't log in for 6 months.
Hey, it's your call....
Richard Stallman's mindset may be overkill at times, but he's basically right. I don't want to use a closed-source browser for the same reason I don't want to use a closed-source OS, among other reasons.
I think determining which browser is more secure is more complex than simply checking a vulnerabilities list. There are all sorts of other variables and factors that play into the number, which may or may not be representative of reality.
Opera seems like a good bunch of people, so I'm not especially worried about their doing anything malicious. But when there are excellent open-source alternatives, why even bother?
If you truly don't care about Free Software, then you have missed the point, and I wouldn't expect you to care about open-source browsers. I think that Free Software is better in principle, in theory, and in practice. It's more future-proof than being locked-in to any company's closed-source software. And as time goes on, it will become even more important, because the power of marketing and media and government conglomerates will continue to grow, and Free Software is nearly the only way to retain user freedoms and privacy, and control over one's own hardware, software, and data.
I don't understand why any knowledgeable user would want to use a closed-source browser nowadays. Any specific benefit Opera might have over another particular browser would be outweighed by the drawbacks, IMO.
Why even make a separate Gmail account? You can use nested labels.
But really, this is what sites like Evernote, Springpad, Diigo, etc. are for. They even have smartphone apps, and have web clippers to save straight from the browser.
"I haven't figured how to back up Yahoo emails because POP3 isn't free." ...switch to Gmail?
Seriously, I cringe when I walk into a computer lab and see people using Yahoo mail. Why would anyone want to use it instead of Gmail? It should be self-evident--I shouldn't have to post screenshots.
Google doesn't hold your email hostage, either.