Re: Operations requirement: no hop-by-hop

On Mon, Feb 16, 2004 at 12:47:01PM -0500, Hector Santos wrote:
> 
> > We are trying to design a *next-generation* protocol, not a temporary
> > patch. It is not useful to say "I have a requirement that spammers
> > will probably get around after I implement it."
> 
> 
> Why can't a system has some "basic fundamental requirements" ones that makes
> "authentication" process a requirement.  This will certainly solve and
> refine all kinds of requirement/design issues.
> 
> Yet as you know, this does not exclude the idea of anonymous access.


I strongly disagree. First of all, "anonymous authentication" is 
as simple as it looks like. Please elaborate how exactly this should
be possible and what this would mean.

Second, authentication can't usually be done without any kind of 
credential, token, proof, trusted third party, or anything like that.

Sometimes you need to be able to send a message when you are "naked". 
And a lot of people strongly deny to accept a protocol which requires
everyone to authenticate. This might not even be possible, since
different legislations have different requirements. Others might
outlaw the use of cryptography.

So we can not make authentication a general requirement. 

What we have to do is to accept that there might be different 
methods and levels of authentication/authorization/whatever.
With or without. Strong or weak. Authentication of sender or
authentication of server. Plenty of options.

So what we need is negotiation. Sender and receiver have to 
declare what they are willing to do/accept, and then they can 
see whether the message is to be transported or not.

By the way, the requirement of authentication in general is 
unfulfillable. Most people won't accept authentication of the 
MTAs only. End-to-End (i.e. sender to recipient) authentication 
requires their presence and direct communication between sender and
recipient. Since we are talking about message transfer with time delay
where sender and receiver may not be online at the same time, 
authentication is not as simple as just claiming the requirement 
of "authentication".

regards
Hadmut