Comment by feelin_googley | Hacker News Reader

Comment by feelin_googley | Hacker News Reader

Parent Full thread

I have scripts and compiled utilities that transform HOSTS file to tinydns format. I use nsd as well which uses BIND format. I prefer the tinydns format.

Zone files are more flexible than HOSTS files, but I still use HOSTS as well. I have never had a concern about the speed of using HOSTS. It is certainly faster than DNS.

There is a comment in this thread where someone asserts that HOSTS was never designed to handle "millions of entries".

I would be interested in reading about a user who visits "millions" of websites or otherwise needs to do lookups on "millions" of domains.

I maintain a database of every IP address I am likely to use on a repeated basis, in several formats, with kdb+ as the "master" format. I believe most users will never come close to intentionally accessing "millions" of IP addresses in their entire lifetime.FN1 I could be mistaken and it would be interesting to learn of a user who can dispel this belief.

FN1. If you think about this, it may cause you to question the necessity of DNS for such users. Or not. Times have changed since the advent of HOSTS. They have also changed since the advent of DNS. For example, using "consumer" hardware, I can fit all the major ICANN TLD zones on external storage and the entire com zone (IMO, by far the most important) in memory. This is many, many more domains than I will ever need to look up. Assuming at best I will not live much longer than 100 years, I could not and will not explore them all or even a significant fraction.

If you know which DNS names you will need to know, then yes, there's no need for more than a hosts file.

Until DNS changes.

If I move a server from one IP to another, I change DNS, and in $TTL time everyone's pointing at the new server. Apart from you with a hosts file. How does that work if everyone has a hosts file?

If I say "check out this interesting story on blahblah.com", you don't have it in your hosts file, how do you get it?

I maintain a list of every phone number I am likely to use on a repeated basis, but sometimes I need to look up a phone number I don't know (in the old days this was a phone book locally, and directory inquiries further afield. Now it's ddg and assume they have a website. Which isn't in my hosts file or dns cache, and I've never visited before)

I maintain DNS entries for my home network of a dozen devices -- I host it on my mikrotik, but it's handy to have, when I type "ssh laptop" rather than remembering if it's on .71 or .73. It's one step better than a plain text file, as there's a standard based way to remotely query it. At work I maintain a DNS server with 2000 entries on my network, which is actually hosts file powered, but again I use dnsmasq for the DNS server rather than rsyncing that hosts file to 2000 machines.

"How does that work if everyone has a hosts files?"

In your particular case, I dont know. You have to do what best suits your needs, whatever they are.

Here is how someone else solves the problem of changing IP addresses. For my needs, I actually like this method.

The entire ICANN DNS used to be bootstrapped to a small text file called "root.hints", db.cache, named.root, named.cache, or something else. As far as I know, it still is.

How does one know the IP address from which to retrieve this text file?

Maybe they have it memorized, or written down somewhere, or perhaps it is written into some DNS software default configuration. In all cases, they have this address stored locally.

No remote lookup.

What happens when the administrator of the server that publishes the text file wants to change IP addresses?

This does not happen very often, but it does happen. What do they do? Considering that the entire ICANN DNS was bootstrapped to this one file, and assuming this is truly meant to be a dynamic system, then this is arguably the most important IP address on the internet.

They notify users in advance that the IP address is going to change.

Thats it.

As a www user, of course I would have to do a DNS lookup for blahblah.com. However I do not do lookups for the server with db.cache, for the .com nameservers, and in most cases not for the nameserver for blahblah.com either, and I do not do lookups using recursive caches. If blahblah.com changes its IP address I do not have to wait for changes to propagate through the system via TTLs. I am querying the authoritative nameserver, RD bit unset. If an IP address changes from the one I have stored, I know immediately when I try to access it. (I like being aware of these changes.) If I was relying a recursive cache I would probably not notice that the IP address had changed.

IME, IP address changes happen less frequently than people writing about DNS on the web would have one believe. Hence this system works well for me. Most domainnames I encounter are keeping the same IP address for long periods.

Ideally, if blahblah.com is not changing IP addresses frequently or unexpectedly but needs to make a change, she could publish a notice somewhere on her web server informing users she will be moving to a new IP address, just like the server that serves db.cache.

Reply on news.ycombinator.com