While the BGP protocol could, in theory, be routed across the internet, my understanding is that in practice it never is.
Quite the opposite. iBGP is almost always routed (within a single AS, not across the public Internet, though), and eBGP is routed whenever two neighbouring ASes don't have routers on the same link, which is fairly common.
I may be wrong, but I was under the impression that the exterior protocol in the OSI suite is IDRP (Inter-Domain Routing Protocol). ES-IS (End-System to Intermediate-System) is the equivalent of ARP, but I've never heard of an ES-ES protocol.
(ES is the OSI term for host, while IS is the OSI term for router.)
It is not trivial to get BGP peering, or to keep it if you are doing bad things.
It's fairly easy to get if you're willing to pay for it. (Keeping it after you do bad things is a different matter, but you can just move to a different provider.)
- You need to be able to offer a better path from Point A to Point B than the existing Internet topology
No, you just need to advertise a more specific prefix. Be careful, though -- too specific, and you'll get filtered.
- Unless you are Dr. Evil and can afford infinite bandwidth, this better path had better not also apply to a large chunk of the Internet, or you will get hosed with a lot of bandwidth
Yes, you need to balance carefully the prefixes you advertise. Too specific and you get filtered, not specific enough, and either your path won't get selected, or you'll get too much traffic.
A nuclear plant could produce twice that on about ten acres.
When nuclear power was first developed, we knew we'd solve all its problems except for refuse. At that time, the problem of refuse was at least 20 years in the future, and we thought we'd have it solved by then.
Today, 50 years later, we still don't have the faintest idea about what to do with nuclear refuse. Until this problem is solved, suggesting nuclear as the one solution to every energy problem is at best short-sighted.
A few weeks ago, I had almost managed to eliminate all the non-Free software that I use. I then tried out a beta of FF 3, and realised that it doesn't fix any of the annoyanced I had with FF 2, and adds new ones.
I've now switched to using Opera. After a couple days of customisation (including learning to design my own buttons), I'm quite happy with it, except for a few minor nits:
it's not Free software. You may not care, but I do.
Unfortunately, it doesn't work very well, since POSIX doesn't (AFAIK) specify the largest write that is guaranteed to be atomic. Hence, unless you're careful, you may end up with log entries from two processes being interleaved.
The entire file is the commit unit. For most files, you either want the entire file written correctly or you don't want anything written. When nothing is written, the previous version, if any, should remain intact.
You don't need any extra kernel support for that. You just write the new version under a temporary name, and then atomically rename it over the old file. Fsync before renaming for extra credit.
Good text editors have been doing that for as long anyone can remember.
Log files. Files are only extended [...] This is UNIX "open for append" mode.
Unfortunately, it doesn't work very well, since POSIX doesn't (AFAIK) specify the largest write that is guaranteed to be atomic. Hence, unless you're careful, you may end up with log entries from two processes being interleaved.
I still don't get why every coffe place doesn't have free, unencumbered wifi access to everyone. It's a great way to get more customers. I always check if there is a free wifi before getting coffee some place.
You're assuming that they have their own ADSL link to every shop. They don't. They decided to let T-Mobile and AT&T control their networking infrastructure, and the operators are understandably less than thrilled by the free competition.
In other words, they decided to give control of their network to a potentially hostile company, and they're getting what they bargained for.
I think you meant to say "packet switched" as opposed to "switched": the latter could also apply to circuit-switched TDM networks, which are engineered differently.
No, the same applies to circuit switching. While there are probably a few million telephone subscribers in the Warsaw area, the telephone network does not have the capacity to carry millions of conversations between Warsaw and New York. We assume that not everyone in Warsaw is going to call their relatives in New York at the same time.
ISPs engineered service under certain assumptions, and those facts on which those assumptions were based are changing
How did this get modded insightful? Oversubscription it the IP supply space is absolutely common. I do not know of a single ISP at any tier that does not sell more bandwidth than they have.
How does that contradict what I said?
Oversubscription is good. It allows us to have more throughput available than what we are each paying for. It is one of the two reasons why switched networks are a goodness (the other being the ability to access peers that we are not directly connected to).
Oversubscription works well when the available capacity is large enough to ensure that everyone gets enough capacity for their needs even at peak times. If you need to artrificially throttle your users in order to provide a useful service, you're probably running with insufficient capacity.
DNSSEC has gone through three (3) mutually incompatible specifications. The DNSSEC people are claiming that the last revision really really works, honest, gov, and that all that remains to be done is deploying it.
Oversubscription is a very, very normal thing in service provider networks.
I think you're confusing oversubscription and unsufficient capacity. Oversubscription is a good thing, it's the very reason we have switched networks in the first place.
The point is that a properly designed and sufficiently provisioned network should not suffer from congestion even if it is oversubscribed. If they've got congestion in their network core, then either they're doing their routing and scheduling all wrong, or they're underprovisioning their network.
Which is fine, as long as they explicitly sell it as ``underprovisioned service''.
The computer can also send alerts to a public transport control centre via a variety of wireless connections, including mobile radio systems, wifi or wimax networks, and UMTS (3G).
Does anyone know if it's IP? And what they're using for routing?
It'd be fun to design a mesh routing protocol for mobile stations with no less than four radio links with very different characteristics...
Web 2.0 is about a thousand layers above hardware, it does not in any manner, approach.
If you're running a highly redundant and completely pointless application, then you want to optimise your hardware differently than if you're running a monolithic and mission-critical one. Which is what the article is about.
Slashdot Mirror
While the BGP protocol could, in theory, be routed across the internet, my understanding is that in practice it never is.
Quite the opposite. iBGP is almost always routed (within a single AS, not across the public Internet, though), and eBGP is routed whenever two neighbouring ASes don't have routers on the same link, which is fairly common.
ESES is the exterior gateway version of ISIS.
I may be wrong, but I was under the impression that the exterior protocol in the OSI suite is IDRP (Inter-Domain Routing Protocol). ES-IS (End-System to Intermediate-System) is the equivalent of ARP, but I've never heard of an ES-ES protocol.
(ES is the OSI term for host, while IS is the OSI term for router.)
I came up with something called "Gateway Database Protocol"
Did you write it down? Any chance you can provide a link?
It is not trivial to get BGP peering, or to keep it if you are doing bad things.
It's fairly easy to get if you're willing to pay for it. (Keeping it after you do bad things is a different matter, but you can just move to a different provider.)
- You need to be able to offer a better path from Point A to Point B than the existing Internet topology
No, you just need to advertise a more specific prefix. Be careful, though -- too specific, and you'll get filtered.
- Unless you are Dr. Evil and can afford infinite bandwidth, this better path had better not also apply to a large chunk of the Internet, or you will get hosed with a lot of bandwidth
Yes, you need to balance carefully the prefixes you advertise. Too specific and you get filtered, not specific enough, and either your path won't get selected, or you'll get too much traffic.
He starts by saying he'll violate the Observer style guide, and then he doesn't.
A nuclear plant could produce twice that on about ten acres.
When nuclear power was first developed, we knew we'd solve all its problems except for refuse. At that time, the problem of refuse was at least 20 years in the future, and we thought we'd have it solved by then.
Today, 50 years later, we still don't have the faintest idea about what to do with nuclear refuse. Until this problem is solved, suggesting nuclear as the one solution to every energy problem is at best short-sighted.
Why did they include a mono family? What's wrong with IBM Courier, which has been included with every distribution of X11 since X11R5?
A few weeks ago, I had almost managed to eliminate all the non-Free software that I use. I then tried out a beta of FF 3, and realised that it doesn't fix any of the annoyanced I had with FF 2, and adds new ones.
I've now switched to using Opera. After a couple days of customisation (including learning to design my own buttons), I'm quite happy with it, except for a few minor nits:
Huh? I've been lending books for as long as I can remember.
Unfortunately, it doesn't work very well, since POSIX doesn't (AFAIK) specify the largest write that is guaranteed to be atomic. Hence, unless you're careful, you may end up with log entries from two processes being interleaved.
You are wrong, any size buffer passed to write() is guaranteed to be written atomically: http://www.opengroup.org/onlinepubs/000095399/functions/write.html Look for O_APPEND.Indeed. Thanks for the info.
The operation finished successfully [OK].
In Soviet Russia...
In Soviet Russia, the Government educates You.
You don't need any extra kernel support for that. You just write the new version under a temporary name, and then atomically rename it over the old file. Fsync before renaming for extra credit.
Good text editors have been doing that for as long anyone can remember.
Log files. Files are only extended [...] This is UNIX "open for append" mode.Unfortunately, it doesn't work very well, since POSIX doesn't (AFAIK) specify the largest write that is guaranteed to be atomic. Hence, unless you're careful, you may end up with log entries from two processes being interleaved.
You're assuming that they have their own ADSL link to every shop. They don't. They decided to let T-Mobile and AT&T control their networking infrastructure, and the operators are understandably less than thrilled by the free competition.
In other words, they decided to give control of their network to a potentially hostile company, and they're getting what they bargained for.
From the Universal Declaration of Human Rights:
No, the same applies to circuit switching. While there are probably a few million telephone subscribers in the Warsaw area, the telephone network does not have the capacity to carry millions of conversations between Warsaw and New York. We assume that not everyone in Warsaw is going to call their relatives in New York at the same time.
ISPs engineered service under certain assumptions, and those facts on which those assumptions were based are changingVery well put.
Which email client brigns up an error message for a self-signed POP3 server certificate?
(Try it, you'll be surprised how many don't.)
How does that contradict what I said?
Oversubscription is good. It allows us to have more throughput available than what we are each paying for. It is one of the two reasons why switched networks are a goodness (the other being the ability to access peers that we are not directly connected to).
Oversubscription works well when the available capacity is large enough to ensure that everyone gets enough capacity for their needs even at peak times. If you need to artrificially throttle your users in order to provide a useful service, you're probably running with insufficient capacity.
But they don't appear to be deploying it on their own servers.
I've just checked -- and the ISC do sign their zone. Sorry for the mis-information.
Only after we switch to IPv6.
DNSSEC has gone through three (3) mutually incompatible specifications. The DNSSEC people are claiming that the last revision really really works, honest, gov, and that all that remains to be done is deploying it.
But they don't appear to be deploying it on their own servers.
Sorry for the lecture in Central-European geography, but you were thinking of Polish notation.
Hungarian is something different altogether. You don't want to know.
I think you're confusing oversubscription and unsufficient capacity. Oversubscription is a good thing, it's the very reason we have switched networks in the first place.
The point is that a properly designed and sufficiently provisioned network should not suffer from congestion even if it is oversubscribed. If they've got congestion in their network core, then either they're doing their routing and scheduling all wrong, or they're underprovisioning their network.
Which is fine, as long as they explicitly sell it as ``underprovisioned service''.
FTA:
The computer can also send alerts to a public transport control centre via a variety of wireless connections, including mobile radio systems, wifi or wimax networks, and UMTS (3G).Does anyone know if it's IP? And what they're using for routing?
It'd be fun to design a mesh routing protocol for mobile stations with no less than four radio links with very different characteristics...
If you're running a highly redundant and completely pointless application, then you want to optimise your hardware differently than if you're running a monolithic and mission-critical one. Which is what the article is about.