@yoursunny said:
The future belongs in a purely name based system, in which every packet carries a name (of any length), and routers deliver packets directly by name. You'll never run out of addresses again because we don't need any to begin with. https://named-data.net/project/execsummary/
But it all depends on what tag your packet identifies itself as..
Yes, you can name your data packet anything you want. However, routing announcement is restricted by certificates. To announce /com/yoursunny , you need a certificate of that name prefix. Without it, you can't pretend to be me, and requests to my website cannot get to your node (in a global network, see next paragraph for local area network).
Assignment of name prefixes at top few levels are arranged similarly as .com domain registration. We'll never run out of name prefixes just like we'll never run out of domain names.
In local area network, NDN has "self-learning" that is a form of multicast forwarding. In that case, data receivers will validate the signatures carried in each data packet, and reject the packet if it isn't signed by a trusted key that matches the name prefix.
@yoursunny said:
In local area network, NDN has "self-learning" that is a form of multicast forwarding. In that case, data receivers will validate the signatures carried in each data packet, and reject the packet if it isn't signed by a trusted key that matches the name prefix.
@yoursunny said:
In local area network, NDN has "self-learning" that is a form of multicast forwarding. In that case, data receivers will validate the signatures carried in each data packet, and reject the packet if it isn't signed by a trusted key that matches the name prefix.
@WSS said:
..because everyone's poorly maintained 10 year old garage door opener that barely supports WPA2 in 2030 should be exposed to the world at large.
There is this great device called firewall.
...which are generally glorified NAT boxes in the majority of residences, except when IPv6 is involved with consumer hardware.
@yoursunny said: The future belongs in a purely name based system, in which every packet carries a name (of any length)
"Of any length" has been a showstopper in the past as far as I have read, because header of unknown length is not optimasible with ASICs. So a header of "any length" is going to be a massive performance hit.
@yoursunny said: The future belongs in a purely name based system, in which every packet carries a name (of any length)
"Of any length" has been a showstopper in the past as far as I have read, because header of unknown length is not optimasible with ASICs. So a header of "any length" is going to be a massive performance hit.
Our collaborator is working on FPGA acceleration.
The basic idea is, hardware can compute a digest of the name prefix, and query forwarding tables with the digest(s).
Digest computation is possible in hardware because its computation complexity is bounded by packet length.
@yoursunny said: The future belongs in a purely name based system, in which every packet carries a name (of any length)
"Of any length" has been a showstopper in the past as far as I have read, because header of unknown length is not optimasible with ASICs. So a header of "any length" is going to be a massive performance hit.
Our collaborator is working on FPGA acceleration.
The basic idea is, hardware can compute a digest of the name prefix, and query forwarding tables with the digest(s).
Digest computation is possible in hardware because its computation complexity is bounded by packet length.
@comi said:
I might be stretching it here, but wouldn't digest length be the same as address length then?
Yes, but digest is for acceleration and does not have to be unique. Even if two names hash to the same digest, the forwarder can still operate correctly. It's just like collision resolution in a hash table.
In IP, if two nodes have the same routable address, disaster happens.
My home ISP started rolling out DS IPv6 after literally years of testing and delays (saying it was an unnecessary extra). But I finally have IPv6 at home now. With a /48(!) subnet. I can’t even calculate how much addresses are in a subnet.
IPv6 networking is still black magic for me as well. So I just delegated a /64 subnet to my router and let it take care of it. It works, so that’s a plus. I only see one downside: 75% of traffic is routed through HE. That’s not down to my ISP, they just peer with HE on a public exchange. Looks like the best part of hosting providers still see HE as their go to network for V6.
Comments
Yes, you can name your data packet anything you want. However, routing announcement is restricted by certificates. To announce
/com/yoursunny, you need a certificate of that name prefix. Without it, you can't pretend to be me, and requests to my website cannot get to your node (in a global network, see next paragraph for local area network).Assignment of name prefixes at top few levels are arranged similarly as .com domain registration. We'll never run out of name prefixes just like we'll never run out of domain names.
In local area network, NDN has "self-learning" that is a form of multicast forwarding. In that case, data receivers will validate the signatures carried in each data packet, and reject the packet if it isn't signed by a trusted key that matches the name prefix.
Accepting submissions for IPv6 less than /64 Hall of Incompetence.
But is this Lantastic compatible?
My pronouns are like/subscribe.
No, NDN self-learning is incompatible with LANtastic. Read how it works in my dissertation: http://hdl.handle.net/10150/625652
Accepting submissions for IPv6 less than /64 Hall of Incompetence.
wait wait wait.. You're fucking serious, using domains under tld groups and shit? Or, are you just trying to pull me in with hardcore shitposts?
My pronouns are like/subscribe.
"Of any length" has been a showstopper in the past as far as I have read, because header of unknown length is not optimasible with ASICs. So a header of "any length" is going to be a massive performance hit.
That's not enough, i need more, we need more! More!!
"How much?"
- MORE!!!
* foams at the mouth, rolls eyes inside brain *
Our collaborator is working on FPGA acceleration.
The basic idea is, hardware can compute a digest of the name prefix, and query forwarding tables with the digest(s).
Digest computation is possible in hardware because its computation complexity is bounded by packet length.
Accepting submissions for IPv6 less than /64 Hall of Incompetence.
I might be stretching it here, but wouldn't digest length be the same as address length then?
Yes, but digest is for acceleration and does not have to be unique. Even if two names hash to the same digest, the forwarder can still operate correctly. It's just like collision resolution in a hash table.
In IP, if two nodes have the same routable address, disaster happens.
Accepting submissions for IPv6 less than /64 Hall of Incompetence.
My home ISP started rolling out DS IPv6 after literally years of testing and delays (saying it was an unnecessary extra). But I finally have IPv6 at home now. With a /48(!) subnet. I can’t even calculate how much addresses are in a subnet.
IPv6 networking is still black magic for me as well. So I just delegated a /64 subnet to my router and let it take care of it. It works, so that’s a plus. I only see one downside: 75% of traffic is routed through HE. That’s not down to my ISP, they just peer with HE on a public exchange. Looks like the best part of hosting providers still see HE as their go to network for V6.