spf-discuss
[Top] [All Lists]

Re: Standard Authentication Query

2005-03-30 11:36:57
At 09:07 PM 3/29/2005 -0800, William Leibzon wrote:
On Tue, 29 Mar 2005, David MacQuigg wrote:

I'm just going by what I have read on this list. As I understand it, the objections to SPF at the vote on the last draft had to do with DNS loading.

It has to do with problems in design and higher dns load is just a visible end-result of those. Particularly various IETF people and those involved in dns have problems with:
 1. (Ab)use of of TXT record
 2. Use of dns for what amounts to general-purpose email policy database
 3. Structure of SPF is not properly layered and technically wrong.
    What may seem like a convinience to some of yoy as far as being
    able to define both hosts with "a" and "mx" operators and ip
    addresses directly with "ip4" is not correct way to do such design
    (and that is obvious to almost every serious protocol engineer).
    It should first be layer listing hosts/networks authorized to act
    on behalf of given domain and then those hosts ip addresses may be
    further looked (if this is needed, which is not necessary if host
    is already verified in some other way) in dns. Basicly RMX
    or Paul Vixie's MAIL-FROM is a lot more cleaner design...
 4. This structure of many spf operators and macros may cause high dns
    load and this load is not always easy to predict and because of
    the incorrect layered design there is possibility of loops and
    security issues that otherwise could be taken care of easily.

As I understand it, these all relate to DNS loading. In other words, if loading were not a concern, these flaws in design would be a minor issue.

If we can address those concerns by something as simple as adding a modifier that doesn't break anything, I would say go for it.

You can't address major concerns with simple modifier.

If we could limit DNS queries to one per email (even just in a special "defense mode") would that not address the major concerns? My expectation is that mail system admins would invoke this "defense mode" for only a few hours, until the DoS storm subsided, and 90% of the legitimate mail would get through, even during the most intense parts of the storm.

Think positive!! We have a problem with the potential abuse of DNS. How would you solve this problem, if you really wanted to?

I don't usually like "adding features" in a situation like this, but I think of masks as a counter to some features that were added without much thought as to abuse. The purpose is not to add new conveniences, but limit abuse.
...
I think Radu addressed this point yesterday. As I understand it, we are free to define any syntax we want within the string on the right side of m=. Maybe to avoid confusion we should write the compact notation as 24~24.30.203.

The purpose of spf-classic draft is to document current use of SPF and you
can't go walking around with major change to syntax that would not be
undertood by current SPF clients as many people pointed before.

I don't know the history, but it is my understanding that we need to address the DNS loading issue in SPF1. Maybe the council could give us some guidance on that. I don't see the m= proposal as a "major change in syntax". Using the "unknown modifier" is extending the syntax in a way that was anticipated. Current SPF clients will simply ignore it.

If you are interested in contributions and changes to syntax, those would need to go to next step in development of SPF (likely SPF3 as SPF2 has been "contaminated" by MARID experience and further anauthorized use for so-called SenderID drafts). There are many things to talk about for next SPF and I was hoping the formation of the council would have helped to quickly resolve SPF1 issues and move those along and that now we'd be well on the way of working on UnifiedSPF and next version of SPF record. Unfortunetly SPF council has been slow and UnifiedSPF has not been mentioned in almost 4-6 months...

I expect at some point there will be a unified standard allowing interoperability of the most prevalent methods. This could include methods as different as SenderID/SPF and DomainKeys. I'll be happy to participate in that development.

-- Dave
************************************************************     *
* David MacQuigg, PhD      email:  dmquigg-spf at yahoo.com      *  *
* IC Design Engineer            phone:  USA 520-721-4583      *  *  *
* Analog Design Methodologies                                 *  *  *
*                                   9320 East Mikelyn Lane     * * *
* VRS Consulting, P.C.              Tucson, Arizona 85710        *
************************************************************ *


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