Hacker News Reader

Comment by Piskvorrr | Hacker News Reader

Aaand that's why we have DNS. Solved decades ago, next.

As someone who's used IPv6 at home as a consumer for probably at least two years, I can tell you I have and to type an IPv6 address into my browser exactly zero times. I think I've maybe had to type one out once when configuring my router (but probably didn't as I use DHCP for address assignment internally to my home network and to get an external address). I couldn't even tell you what the numeric address format is, because I literally never have to know it.

Not solved on LANs, ad-hoc local networks, virtual networks, or when DNS is down or broken.

This is not really an end-user issue but it is a serious usability issue for IT admins and developers. It's very very common to schlep around raw IPs constantly when messing with networks and I don't see that going away. It's also very important to be able to visually parse IPs when understanding the topology of a network, writing firewall rules or routes, etc.

This is a DX (developer experience) issue more than a UX (user experience) issue.

Edit: three specific problems with DNS:

(1) What happens when things are not configured yet?

(2) OSes are designed to have one DNS server but people belong to many networks either at once (local + VPN + virtual + ...) or serially via mobility. In reality you need many DNS servers, but then how do you deal with naming conflicts?

(3) DNS is dependent on IP so you can't use DNS to debug DNS issues. It's a circular dependency.

In practice IP schlep is very common.

Hostname discovery (mdns), etc solves it on local networks. If DNS is down or broken, it would be difficult for me to "use the internet", and I'd have to copy and paste a lot anyways.

So, run yet another service with a poorly thought out RFC and add another code base to my vulnerability monitoring?

At home/small business scale, yes. At enterprise scale, you shouldn't be relying on IP addresses for clients for anything important.

mDNS is slow and unreliable and on large networks it doesn't scale. I personally don't find it very useful since half the time it barely works even on wired LANs let alone big WiFi or distributed networks. It's also prone to naming conflicts (is this linux-1, linux-2, or linux-3?) and other OS configuration issues and is not secure.

On a network like that, wouldn't DDNS solve the problem?

I mean DDNS where the DHCP server tells the DNS server which IP has what hostname, not e.g. dyndns.

I don't think people trust DDNS enough in corporate networks and any security sensitive network.

> ad-hoc local networks

Actually that's not a problem for that. Since people who are actually using them have no idea about IPv4, too and just copy addresses. Actually you could create local address with everything prefixed by fd + 40 bit which could be just fdff:ffff:ffff:ffff::1 which isn't really hard to remember.

Or, you know, when your friend is hosting a StarCraft game over TCP and you have to write their address. People don't have DNS addresses and even if your ISP has a mapping by default (usually to addr-x-y-z-w.isp.com), the OS doesn't show it.

On the other hand, DNS could fix this - just have a TLD of ip6 and have it resolve all the examples in the article. It would require no changes to current software and will work transparently. I.e. you'd enter http://deadbeef.ip6:1234 and when the ip6 TLD servers receive a request for deadbeef.ip6, they will reply with dead:beef:0:0:0:0:0:0. Similarly with deadbeef.1.ip6 and so on. You could easily implement this in the OS too without much hassle and not even need servers on the internet to do it.

I see this idea like a really cool one.

http://deadbeed.ip6 would help to start using it.

If the local dns could be setup to autotranslate them to ip6 could help this notation to gain traction

Using a special TLD for IP6 addressing is the sanest proposal I've read here so far. Add the ip4 TLD while we're at it.

I just had this dystopian vision where a non-profit operated .ip6 to work as discussed then went defunct and a domain-grabber (named Network Solutions) bought it. Everybody scrambled to patch their recursive resolvers real fast :-)

If it catches on, I expect it to be built into every dns library so it does not even hit the network. The idea of grabbing the TLD and then redirecting to a middleman that inserts ads sounds quite profitable though...

Hosting Starcraft on a port open to the Internet sounds like a terrible idea. If you actually want to host a public server, then getting a dynamic domain name is not hard. If it's a private game, you're better off using a VPN, over which a broadcast discovery protocol can be used.

In any case, if you really need to, what's the problem with copy-pasting an address?

I don't even get why we have IP addresses anymore. Just use DNS duh.

DNS are centralized , IPs are not. When USA decides to block or take down you domain name, because they can, all you have left is your IP, they can't take your IP.

Meet https://www.arin.net/ and friends.

Well, one reason not to is limited hardware.

In an embedded system where memory is at a premium and you may have serious constraints on how long things take (e.g. for timing purposes), the ability to hard-code an address or have it entered in some way saves you from having to support an entire DNS layer in that system.

I think the joke is that DNS just hides IP addresses, it doesn't do away with them. No IP address = no DNS.

Finally, thank you. Someone with some sense around here.

Wow!

Reply on news.ycombinator.com