I'm not quite sure where you get your information from, but I seriously doubt any providers are specifically rewriting spoofed udp packets into legitimate addresses. Routers simply do not perform tasks like that.
Providers are not dropping/rewriting just spoofed udp. They're either dropping all spoofed traffic, or letting all traffic through. Why would they apply this rewriting only to udp if they were indeed doing it? The simple answer is that they are not doing it.
This is incorrect. Low speed serial interfaces do tend to use a start bit and a stop bit, but higher speed interfaces generally do not.
I'm not very familiar with 10gige technology yet, but my brief research shows that it uses 64B/66B coding (e.g., 2 overhead bits out of every 66). Running at a clock rate of 10.3125GHz, that gives you a full 10Gbps of throughput, or 1.25 GB/sec.
100baseT uses 5B/4B coding, which does result in 2 overhead bits out of every 10 just like your serial line example. However, 100baseT actually runs at 125MHz so you do get a real 12.5 MB/sec out of it.
Of course, if you really want to be picky about "LoC/sec" or whatever pointless measure the popular media has latched onto this week, you need to consider the overhead of TCP headers, whether or not you want to allow jumbo frames in your calculations, and so on.
[] global users.. 61438 max.. 80102 (58085 invisible, 67 operators)
[] local users... 32351 max.. 38435 (52.7%)
From twisted.ma.us.dal.net. We've been holding > 25k users on a consistent basis the past few weeks since many other servers are missing. Our record is 38435 simultaneous users.
Don't forget to get the XF ree 86 4.0 drivers from NVIDIA, even if you're not using 3d; they're faster for 2d stuff than the open-source drivers included with XFree86.
Some cautions in doing it yourself
on
Homebrew S/ADSL
·
· Score: 5
I've been using an SDSL connection between my office and house for the past two months. Paid $800 for a pair of Flowpoint 2200 routers, and BA takes $30/mo for the copper loop ($100 install).
There are a few issues with doing it yourself. If there are are problems, you don't have an ISP to complain to. It can be difficult to track down if a problem is hardware related, in the copper loop, or in your own inside wiring. For the past two weeks, I had major problems (connection would develop massive packet loss and drop out at times), which I eventually figured out were the routers; just got them replaced yesterday after spending hours trying various possible fixes from Flowpoint. There are still some minor problems (had the routers randomly drop the link and reset on me once earlier tonight), so my next step is to redo all the inside wiring on both ends.
As far as hardware recommendations go, despite the problems I would still recommend the Flowpoints. They can run in routed or bridged mode and you can do fun stuff like set up multiple ATM PVCs with traffic shaping, so one user's traffic doesn't disturb another's. (Here's another issue; with the line saturated, the latency jumps from 3ms to 140ms; I set up one PVC for my roommates and one for myself, and set them both to 95% traffic max to get around this...)
At about 8500 feet, my SDSL runs at 1744kbit/sec; the max the Flowpoints will do is 2320 at shorter distances. Another option I looked at was the Netopia R7171; it wasn't out at the time (it should be now), but it'll do two bonded 1.5mbit/sec channels over a 2 pair loop, as opposed to a single pair for normal DSL.
So basically, if you're not prepared to have to fix problems yourself, go with a commercial DSL company. Otherwise, if you have somewhere to stick the other end (office or whatever), this is a pretty cheap alternative to going through an ISP; I doubt I could find an ISP to give me a large IP block and controllable revdns for so cheap.
Every time I've seen a domain go on hold, the updated time gets changed. This domain was updated on the 23rd, and as other people have noted, you can see that there is $35 due on Network Solutions' online payment page. This makes it seem fairly certain that Microsoft just forgot to pay.
The root server updates aren't in synch with the whois updates, which explains why it didn't disappear from the root servers until later.
Registrant: Microsoft Corporation (PASSPORT6-DOM) One Microsoft Way Redmond, WA 98052-6399 US
Domain Name: PASSPORT.COM ... Record last updated on 23-Dec-1999. Record created on 02-Oct-1996.
And from nslookup: > server a.root-servers.net > passport.com. Server: a.root-servers.net Address: 198.41.0.4
If you are using Debian, you can 'whois -C ' to see the full info. This has the added benefit of looking up info for domains registered with other registrars.
-C First query www.crsnic.net to find the registrar for a GTLD subdomain. This option overrides -h.
(Other whois or rwhois clients may have a similar option. The whois client on a Redhat 6.0 box doesn't seem to have any way to do this though.)
I'm not quite sure where you get your information from, but I seriously doubt any providers are specifically rewriting spoofed udp packets into legitimate addresses. Routers simply do not perform tasks like that.
Providers are not dropping/rewriting just spoofed udp. They're either dropping all spoofed traffic, or letting all traffic through. Why would they apply this rewriting only to udp if they were indeed doing it? The simple answer is that they are not doing it.
This is incorrect. Low speed serial interfaces do tend to use a start bit and a stop bit, but higher speed interfaces generally do not.
I'm not very familiar with 10gige technology yet, but my brief research shows that it uses 64B/66B coding (e.g., 2 overhead bits out of every 66). Running at a clock rate of 10.3125GHz, that gives you a full 10Gbps of throughput, or 1.25 GB/sec.
100baseT uses 5B/4B coding, which does result in 2 overhead bits out of every 10 just like your serial line example. However, 100baseT actually runs at 125MHz so you do get a real 12.5 MB/sec out of it.
Of course, if you really want to be picky about "LoC/sec" or whatever pointless measure the popular media has latched onto this week, you need to consider the overhead of TCP headers, whether or not you want to allow jumbo frames in your calculations, and so on.
This is a fast mirror (the one in the parent message is horribly saturated).
Well actually, you're wrong.
[] global users.. 61438 max.. 80102 (58085 invisible, 67 operators)
[] local users... 32351 max.. 38435 (52.7%)
From twisted.ma.us.dal.net. We've been holding > 25k users on a consistent basis the past few weeks since many other servers are missing. Our record is 38435 simultaneous users.
Also see the article about us on linux.com.
Don't forget to get the XF ree 86 4.0 drivers from NVIDIA, even if you're not using 3d; they're faster for 2d stuff than the open-source drivers included with XFree86.
I've been using an SDSL connection between my office and house for the past two months. Paid $800 for a pair of Flowpoint 2200 routers, and BA takes $30/mo for the copper loop ($100 install).
There are a few issues with doing it yourself. If there are are problems, you don't have an ISP to complain to. It can be difficult to track down if a problem is hardware related, in the copper loop, or in your own inside wiring. For the past two weeks, I had major problems (connection would develop massive packet loss and drop out at times), which I eventually figured out were the routers; just got them replaced yesterday after spending hours trying various possible fixes from Flowpoint. There are still some minor problems (had the routers randomly drop the link and reset on me once earlier tonight), so my next step is to redo all the inside wiring on both ends.
As far as hardware recommendations go, despite the problems I would still recommend the Flowpoints. They can run in routed or bridged mode and you can do fun stuff like set up multiple ATM PVCs with traffic shaping, so one user's traffic doesn't disturb another's. (Here's another issue; with the line saturated, the latency jumps from 3ms to 140ms; I set up one PVC for my roommates and one for myself, and set them both to 95% traffic max to get around this...)
At about 8500 feet, my SDSL runs at 1744kbit/sec; the max the Flowpoints will do is 2320 at shorter distances. Another option I looked at was the Netopia R7171; it wasn't out at the time (it should be now), but it'll do two bonded 1.5mbit/sec channels over a 2 pair loop, as opposed to a single pair for normal DSL.
So basically, if you're not prepared to have to fix problems yourself, go with a commercial DSL company. Otherwise, if you have somewhere to stick the other end (office or whatever), this is a pretty cheap alternative to going through an ISP; I doubt I could find an ISP to give me a large IP block and controllable revdns for so cheap.
Every time I've seen a domain go on hold, the updated time gets changed. This domain was updated on the 23rd, and as other people have noted, you can see that there is $35 due on Network Solutions' online payment page. This makes it seem fairly certain that Microsoft just forgot to pay.
The root server updates aren't in synch with the whois updates, which explains why it didn't disappear from the root servers until later.
Registrant:
Microsoft Corporation (PASSPORT6-DOM)
One Microsoft Way
Redmond, WA 98052-6399
US
Domain Name: PASSPORT.COM
...
Record last updated on 23-Dec-1999.
Record created on 02-Oct-1996.
And from nslookup:
> server a.root-servers.net
> passport.com.
Server: a.root-servers.net
Address: 198.41.0.4
*** a.root-servers.net can't find passport.com.: Non-existent host/domain
If you are using Debian, you can 'whois -C ' to see the full info. This has the added benefit of looking up info for domains registered with other registrars.
-C First query www.crsnic.net to find the registrar for a GTLD subdomain. This option overrides -h.
(Other whois or rwhois clients may have a similar option. The whois client on a Redhat 6.0 box doesn't seem to have any way to do this though.)
I got annoyed with having to use NVidia's 3.3.3.1 X server, and their glx.so module wouldn't work in 3.3.5, so I compiled their source against 3.3.5.
The binaries are available here.
Doesn't seem any faster than their 3.3.3.1 server, but at least it lets you use 3.3.5.