how-to block ads
|
|
Share Topic
 |
 |
|
|
davePremium,MVM
join:2000-05-04
not in ohio
kudos:7 Reviews:
 ·Verizon FiOS
 ·Verizon Online DSL
2 edits | reply to Skippy25
Re: Who Cares? said by Skippy25:Again I do not see how this "breaks" DNS. You don't get the page you wanted because it does not exist. You keep talking about "pages". There's no such thing in DNS.
If I write an application, and the user configures a non-existent name, then I expect gethostbyname() to return "no such name".
I do not expect it to return the address of some random system that does not implement the expected protocol.
These are the rules I've understood to be operative in the last 20 years or so.
Correct me if I am wrong but DNS is to resolve the name you type to an IP address. That address does not exist and DNS responds accordingly. You're wrong.
What is supposed to happen is that DNS should return "no such name".
Instead, it pretends the name exists, and returns the address of some system I've never heard of.
If the application then connects via HTTP, the HTTP server at that address will serve up some sort of web page. Very nice for HTTP.
If the application then connects via SMTP, the mail server will (I presume) say "no such user" at the appropriate juncture in the protocol, and mail sort of works, but differently from what it did before.
Maybe there's a similar thing for the ftp protocol.
But for any other protocol -- let's say the mind file system protocol, since I guarantee no-one has implemented that -- then the "bogus address" will not have a server and cannot deliver an appropriate response. They don't even know the goddamn encoding of a mind file system response, so they can't respond to it.
Looks to me like DNS name resolution worked perfectly. Just because another party took that error and turned it into something more meaningful to a vast majority of the internet users makes it no less of a negative DNS response. No. The correct response for a non-existent name is to return a "nonexistent name" error.
A name-to-address service that lies to its client is broken.
A service definition not only describes what happens in the 'success' case (for DNS, looking up a known name), it describes what happens in 'failure' causes (like looking up a name that is not known). Returning an address for an unknown name is just as much a violation as not returning an address for a known name.
Sounds like a mail application problem to me and the application should be fixed or the administrator of the mail server should change their method for resolving addresses. Yes. Every piece of application code written in the last 20 years should be reworked because some idiot fails to return "unknown name" in response to an unknown name, but instead pretends that the name is valid.
This is precisely the objection.
Now lets discuss the... it's always been that way and we can't change it argument. This is called innovation and evolution of technology. So find something else petty to complain about. No, let's discuss the fact that the internet holds together because the implementors agree to implement standard protocols in a standard way so that systems can interoperate in a predictable way.
Innovators are free to implement any damn service they want, no matter how twisted. What they're not free to do is to implement a variation of DNS that violates the protocol definition, and then call it "DNS" and foist it on people that are connecting to a DNS service.
You've got some balls to say that the people who put the Internet together (which doesn't include me) are afraid of "innovation" !
Since we're fond of car analogies at this site -- why don't you unilaterally start driving on the left-hand side of the road? I know the standard in the USA is to drive on the right, but hell, why not innovate? There might be a few interoperability problems to start with, but everyone else will have to adapt. | | |
|
| Wow you sure had a lot to say there, but what is stopping you from redirecting your DNS queries to another server?
Does a redirection even affect applications outside of HTTP request (mail appears to work one user said)? What is stopping innovation from going to only HTTP being affected? Oh that's right.... it's been like that for the 20 years you know of so why move on.
DNS is not perfect, maybe determining the type of name request would be an improvement. | |
davePremium,MVM
join:2000-05-04
not in ohio
kudos:7 Reviews:
·Verizon FiOS
·Verizon Online DSL
1 edit | said by Skippy25:Does a redirection even affect applications outside of HTTP request (mail appears to work one user said)? What is stopping innovation from going to only HTTP being affected? Oh that's right.... it's been like that for the 20 years you know of so why move on. It's DNS. DNS has nothing to do with HTTP. You can't tell from a DNS query that the application intends to talk HTTP. That's the whole problem.
You're framing this as me being 'against innovation', but my argument is that they shouldn't break a basic Internet protocol just to benefit web users. If it could be confined to web users, that would be more-or-less ok. But the protocol does not carry enough information for anyone to make that distinction.
said by Skippy25:DNS is not perfect, maybe determining the type of name request would be an improvement. Yes, that's the sort of innovation that would actually make some sense.
As it stands, DNS mostly resolves 'network layer' names - you get an IP address (MX would seem to be an exception).
What a service like this needs is to handle 'transport layer' names (mapping to IP address, protocol, port number). Given such an arrangement, you could reply with the address of a web server if it was a name lookup from a web app, or with 'no such name' if it was one of these protocols that I'm more worried about.
The trouble is, that requires client code. As such, it's hard to unilaterally declare you've got a new service. You need to either write code for all clients, or persuade client implementors that you're on to a good thing. (This last is what the Internet RFC mechanism is all about). | |
 winkyTurn Left At The Moon
join:2001-02-11
Saint Louis, MO | reply to dave
Oh Dave, save your breath. This guy probably learned all he needed to know about economics in seventh grade civics class just like he already knows all about the interweb now. 
--
From this point forward Hoedown, from the ballet RODEO, by Aaron Copeland will not be reffered to as "The Beef Song". Thank You | |
|
