On Jun 11, 2008, at 1:48 AM, Charles Lindsey wrote: > On Tue, 10 Jun 2008 18:34:57 +0100, Douglas Otis <[EMAIL PROTECTED] > abuse.org> > wrote: > >> On Jun 9, 2008, at 9:21 PM, Jim Fenton wrote: > >>> Since it apparently isn't clear: I am proposing retaining the >>> NXDOMAIN domain validity check as a MUST. It is only the MX and A/ >>> AAAA check that I'm proposing be changed from a SHOULD to a MAY. >> >> The situation created by MS Exchange creates a problem where just >> an NXDOMAIN check is still problematic. While NXDOMAIN might occur >> for any leaked X.400 address or typical >> "[EMAIL PROTECTED]", NXDOMAIN results might also occur >> with any proxy SMTP addresses assigned by MS Exchange. This occurs >> since MS Exchange assignments and routing do not depending upon DNS >> records. Such an NXDOMAIN test would disrupt messages created by >> the company where I work, for example. In addition, unless the >> test goes one step further to determine whether a domain appears to >> support SMTP, this would offer far less utility in preventing >> address spoofing. Nor could just an NXDOMAIN test offer protection >> for non-SMTP domains. > > But you have repeatedly failed to explain how a verifier could > recognise and handle this case in a manner that did not leave a > loophole for all the scammers and spoofers to walk through. If some > message arrives with a From that includes a proxy SMTP address > assigned by MS Exchange (which will surely result in NXDOMAIN), what > do you want the Verifier to do? Is there some way that is can > recognise this as a proxy address and let it through whilst still > rejecting things apparently from the domain funny.ebay.com?
It is impossible for a verifier to determine the nature of an NXDOMAIN result when validating From email-addresses. Validating that the From domain might support SMTP (or that it exists) is not limited to domains using DKIM or those that may publish ADSP at some domain level. While a domain validation effort might be seen as necessary in the elimination of address spoofing (and to limit where publishing practices might be required), the impact of domain validation is global. As such, domain validation should be seen as a systemic change to SMTP, and not simply as an aspect of ADSP. Once SMTP interoperability is changed to require the From (or all originating) email-address domains to publish records supporting SMTP, some messages will not be compliant. When imposing this requirement, provisions should be available that allow receiving hosts to make exceptions based upon either the domain or IP address of the SMTP client. A means to make exceptions ensures crucial systems remain functional even when DNS is inoperable. Exceptions also provide a solution while waiting for non-complaint domains to become compliant. > If some companies using MS Exchange allow such messages to escape, > then I am afraid that is just Tough! It is a stupid behaviour. I > might accept that domains whose TLD clearly did not exist could be > exempted from the NXDOMAIN check in ADSP. Which TLDs should be ignored? Imposing SMTP domain requirements will likely reveal a need to make many exceptions. Do you agree there should be a means for making exceptions? Whether making address assignments exclusively within MS Exchange is considered stupid (and you'll find agreement there), imposing a requirement that email- addresses must be valid (in some manner) changes SMTP interoperability. As it is now, recipients will normally see these messages (which may not expect a response), and might even be considered an alternative to the use of "do-not-reply@" local-parts. > And what do you mean by a "non-SMTP domain. AKAIK the phrase is > meaningless. A domain publishing SMTP discovery records could be defined as a possible SMTP domain. With even greater certainty, those that don't can be described as a non-SMTP domain. IMHO, a draft defining what might be an SMTP domain should exclude AAAA from a list that provides confirmation of SMTP support. AAAA records for SMTP discovery supports a case where local host definitions are needed, especially when DNS is not available. Just as with a local host definition, exceptions made for publishing MX records also are needed for crucial systems. In practice, SMTP already requires extensive client evaluation. An option for requiring or allowing exceptions for SMTP domain support by receiving hosts seems an appropriate means for imposing the requirement. Judging by the growing impatience, this strategy should be published in a separate draft from that of ADSP. -Doug _______________________________________________ NOTE WELL: This list operates according to http://mipassoc.org/dkim/ietf-list-rules.html
