ietf-dkim
[Top] [All Lists]

Re: [ietf-dkim] Issue 1576: Revise wildcard discussion

2008-07-07 17:50:21

On Jul 7, 2008, at 3:14 PM, Jim Fenton wrote:

John Levine wrote:

Well, maybe.  There's plenty of other ways a record can be ill- 
formed.

*.foo.example TXT "dkim=none dkim=discard"
*.foo.example TXT "dkim=a total crock nobody is gonna use"
Of course I didn't mean to imply that checking the first five bytes of
the ADSP record was going to guarantee validity of the rest of it.   
The
key word in my statement above was "unintended".


Section 4.1 will be problematic when ADSP records are concurrent with  
other TXT records also published at a wildcard domain.
,---
|A domain MUST NOT publish more than one ADSP record; the
|semantics of an ADSP lookup that returns multiple ADSP
|records for a single domain are undefined.
'---

The ADSP record is _any_ TXT record that exists below _adsp._domainkey  
domain prefixes.  While the "dkim=" tag is required, it could be  
anywhere within the record.

When a domain attempts to shield address records from being considered  
a valid source for unsigned messages, including the _domainkey prefix  
as well can be considered detrimental since this will increase zone  
size and administrative efforts.  In addition, for smaller domains,  
whenever the _domainkey zone is delegated to a third-party email  
provider, ADSP will also be controlled.  When separate domains below  
the _domainkey are delegated to enable use of multiple third-party  
providers, or to isolate ADSP control, then the third-party provider  
remains reliant upon the domain to publish the ADSP record, or to  
separately delegate the _adsp prefix as well.

ADSPs records below _adsp.<Author Domain> can be delegated, or  
controlled separately in the same manner as the _domainkey record.   
Placing the ADSP record below additional prefixes provides little  
benefit, especially when tag placement for the required dkim= tag is  
mandated as beginning the record.  Concerns related to potential TXT  
record confusion is reduced to an even greater degree than that  
provided by the additional prefixes since these prefixes offer no  
isolation for wildcard domains.  In addition, placing _adsp below  
_domainkey simply wastes an additional 11 bytes of payload.  Why  
attempt to optimize domain delegations for adsp records?  MX records  
are also likely involved when dealing with third-party providers.   
Unlike DKIM public key roll-over, there should be little reason for  
ongoing changes to ADSP records once established.

-Doug 
_______________________________________________
NOTE WELL: This list operates according to 
http://mipassoc.org/dkim/ietf-list-rules.html

<Prev in Thread] Current Thread [Next in Thread>