Re: Fwd: New Version Notification for draft-cdn-loop-prevention-00.txt

On 02/07/18 20:47, Poul-Henning Kamp wrote:
> --------
> In message <F1950AC9-CA4C-4F17-9F1A-CADA18679FE6@mnot.net>, Mark Nottingham wri
> tes:
> 
>> For interest / discussion. This is a proposal for a minimal mechanism to 
>> avoid loop attacks and misconfigurations against CDNs. Feedback 
>> appreciated.
> 
> This problem is as old as packet networks, and why IP have the TTL field.
> 
> I think it would be better and more robust to define a "max-hops"
> header with a single numerical field, which each (conforming) proxy
> decrements and if it becomes zero, 50x error is returned.
> 
> CDNs can create a max-hops header if there is none, and even if
> nobody else implements the max-hops header, it will eventually count
> down to zero if there is a loop.
> 
> In difference from the proposed draft, this doesn't reveal the
> architecture to the client.
> 

Max-Forwards goes almost there. Would just need to revise this part of
RFC 7231 section 5.1.2 to make all recipients SHOULD decrement instead:

"
 A recipient MAY ignore a Max-Forwards header field received with any
 other request methods.
"

... and nothing is preventing us middleware and CND people decrementing
on any method right now.

Amos

Received on Monday, 2 July 2018 09:57:06 UTC