how-to block ads
|
|
Share Topic
 |
 |
|
|
| reply to Anon2
Re: Some more info on Goggle DNS & some speed tests said by Anon2 :
In particular, they moved away from that when hosts files began to take up megabytes of space (when megabytes was still a significant fraction of available space) and more importantly when searching began to take forever. A hosts file containing the entire internet would be more storage than most workstations have and take hours if not days to run a query on.
In this day and age when a computer can be had that contains many GB's of hard drive space for dirt cheap - when the processors are in the quad fashion - when extra storage is a matter of a few hundred bucks - again, a possibility.
Keep in mind that I as a user might browse to 1% of all of the servers on the WWW. That means I don't really need to have all of the name resolutions local. Maybe a cache that stores them as they're requested, and updates them on a schedule. We already have the technology to do that - I'm more against the concept of a central "DNS server" hosted by someone else. That said, I'd probably be more confident if I knew that DNS servers were a decentralized commodity.
I don't know, I'm throwing wild ideas out. I just don't like the idea of a server whose sole purpose in life is to put a friendly name around the true address of the destination server so that a user doesn't have to remember that true address. It feels archaic.
Let's take IPv6. At its core it looks like a hex string - why not cater that hex string to the individual computer name and domain? So that if you type in said name, your workstation simply translates the name into hex in order to locate it? NO server, just reading a string of characters and turning it into something. | |
| No offense, but maybe you should spend a few weeks thoroughly learning the technology you are trying to improve before throwing "wild ideas out".
The biggest thing you are missing are the dynamics of the current DNS structure. The way the current DNS structure works is actually quite brilliant considering the way it has to scale.
Another thing you are missing is that it does cache. Your computer already cache the address, your company's network DNS cache it, your isp DNS cache it, and so on until it has to reach the root. Which is the reason a major issue will take hours to replicate out across the net.
If you want local caching for the sites you frequent then modify your local lmhost file, then your problem is solved. Just don't come here whining when the company changes IP addresses and you can't find your stuff anymore and have no clue why. | |
NormanSPremium,MVM
join:2001-02-14
San Jose, CA
kudos:4 Reviews:
 ·SONIC.NET
·Pacific Bell - SBC
| reply to ReVeLaTeD
said by ReVeLaTeD:Keep in mind that I as a user might browse to 1% of all of the servers on the WWW. That means I don't really need to have all of the name resolutions local. Maybe a cache that stores them as they're requested, and updates them on a schedule. We already have the technology to do that - I'm more against the concept of a central "DNS server" hosted by someone else. That said, I'd probably be more confident if I knew that DNS servers were a decentralized commodity. It is "decentralized". I use my ISP DNS servers. They are not the same ones used by Comcast customers, or Verzion customer, or Road Runner customers, or Qwest customers, or Charter customers.
They are also not the only ones I can use. There are, of course, a number of free DNS servers available, and, if I chose, pay DNS servers, as well. Hardly "centralized".
In addition, you do have the option of running a recursive, caching resolver locally, right on your computer.
I don't know, I'm throwing wild ideas out. I just don't like the idea of a server whose sole purpose in life is to put a friendly name around the true address of the destination server so that a user doesn't have to remember that true address. It feels archaic.
I sometimes drive a 1997 Plymouth Voyager. Compared with what is available today, it probably is archaic. But what is available today, commencing with the latest year models, is archaic, when compared with the first Otto internal combustion powered vehicles appearing on the road 130 years ago.
I was swapping war stories with a nephew returned from Iraq. The weapons he was issued are much newer models than I carried; but they still feel "archaic". Considering that they are based on ideas patented, and invented in the latter part of the 19th Century, they are archaic.
Just saying that "archaic" isn't always bad.
Let's take IPv6. At its core it looks like a hex string - why not cater that hex string to the individual computer name and domain? So that if you type in said name, your workstation simply translates the name into hex in order to locate it? NO server, just reading a string of characters and turning it into something.
Maybe I am missing something important here. Isn't that exactly what the 'hosts' file does? The reason the 'hosts' file was abandoned was because it didn't scale well, and who takes care of the updates?
With the DNS service, there is a set of authoritative DNS servers for each domain, and all the others are caching servers, who get the results, ultimately, from the authoritative servers. You've got to trust, at the least, the authoritative DNS servers.
--
Norman
~Oh Lord, why have you come
~To Konnyu, with the Lion and the Drum | |
 espaethDigital PlumberPremium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
 ·Clear Wireless
1 edit | reply to ReVeLaTeD
said by ReVeLaTeD:We already have the technology to do that - I'm more against the concept of a central "DNS server" hosted by someone else. That said, I'd probably be more confident if I knew that DNS servers were a decentralized commodity. DNS servers ARE decentralized. You have to perform 4 queries against completely different clusters of servers just to get the final IP resolution of an address like www.dslreports.com. DNS is decentralized from the core -- starting from the root servers that are operated by a combination of government and private companies across several different countries.
said by ReVeLaTeD: I just don't like the idea of a server whose sole purpose in life is to put a friendly name around the true address of the destination server so that a user doesn't have to remember that true address. It feels archaic. I think anyone who works in the industry would argue exactly the opposite. That's like saying we shouldn't publish phone books or offer directory services, you should just remember the phone numbers of anyone you'd want to dial. Hard-coding IP addresses into applications and services has been the biggest headache of any IT organization's operations -- name-based abstraction allows the IP address, which is tied to physical location, to change without having your method of accessing the site change.
They could pick up the dslreports.com servers from NAC.net on the east coast and drop them in a DC in the west coast and with DNS abstraction you would never have to change anything about how you access the system.
IP addressing is location centric, just like postal ZIP codes. That's why name-based abstraction makes all kinds of sense. | |
|
