Re: static versus dynamic nature of DNS: rate limiting

On Sun, Feb 22, 2004 at 09:10:37PM +0100, Patrik Fältström wrote:
> 
> People which think HTTP is a good solution should first understand the 
> HTTP spec (RFC 2616 and RFC 2817) and then read RFC 3205. Especially as 
> the email architecture today doesn't use the HTTP protocol. Do people 
> really think it is a good idea to be dependent on one more protocol, 
> when we already need SMTP and DNS?

It is not as complicated as it looks like, because my proposal does
not use HTTP as a "substrate" for other protocols (like IRC over HTTP
or something like that) as described in RFC3205.

My proposal uses HTTP just exactly for what it has been designed for: 
Download a file, nothing else. No special tricks, no substrates, just
plain HTTP usage as since HTTP was invented.

Implementing this is much easier than implementing a DNS based
approach, because there are plenty of HTTP libraries. In most script
languages it will be a one line command. 


And I do believe it is better to depend on a robust and well proven
protocol instead of trying to teach the old cow fly. After all, 
SMTP, DNS and HTTP are by far the most used protocols of the
Internet. So it is not a disadvantage to depend on HTTP. 

It is better to use HTTP for what it has been designed for than using
DNS for what it has not been designed for. Several proposal try to 
stitch together data stored in different TXT records and need to use 
odd tricks to count the lines. Any way to cause a DNS server to
refetch the DNS records before expiry if one TXT line is missing or if 
TXT lines belong to different versions of the zone file? 

Using HTTP has a large list of advantages. And I don't think that it
is really a disadvantage that it is a "third" protocol.

regards
Hadmut