spf-discuss
[Top] [All Lists]

Implementation questions

2003-11-13 03:55:35
I have a couple of implementation questions arising from the spec:

Q1.
It is not immediately clear what the use of the '?' modifier on individual 
mechanisms may be, since it does not 'short-circuit' evaluation.

I assume that if no subsequent mechanisms match, then the result of the prior 
'?'-modified match would be returned. Would this be the most-recent or the 
least-recent such match? This is relevant for calculating the result's cache 
TTL using the various DNS record TTLs.

Would it be fair to describe a '?'-modified match as replacing the default 
modifier for subsequent evaluation? In which case the most-recent such match 
would be the most logical interpretation.

If this interpretation is correct, how is the new default modifier interpreted 
for subsequent mechanisms which have no explicit modifier? Does the following 
from section 3.0 of the spec also apply here:

   If "default=deny",     matching mechanisms return "allow" by default.
   If "default=softdeny", matching mechanisms return "allow" by default.
   If "default=unknown",  matching mechanisms return "allow" by default.
   If "default=allow",    matching mechanisms return "deny"  by default.

If so then can we say that any not-explicitly-modified match subsequent to a 
'?'-match will be interpreted as 'allow', regardless of what was in your 
original default modifier?

Q2.
The PTR section is really confusing. Am I supposed to do a PTR lookup of the 
IP at in-addr.arpa, or somewhere else? I thought the in-addr.arpa mappings 
were supposed to be one-to-one? Are there any examples of one-to-many 
mappings there at present? Could a malicious domain do this?

Additional comment:
One of the things I like about the original SMTP spec (rfc821) is the 
inclusion of implementation guides in the form of state diagrams. With 
something more complex like SPF it would be wise to include such things.

In particular I am worried about the complexity of getting result caching 
calculations right, given the number of records brought into play. This 
requires some analysis to get right and at worst might leave individual sites 
or perhaps the wider internet open to DoS and DDoS attacks if a poorly 
considered implementation becomes widespread (eg if any of sendmail, qmail, 
exim, exchange etc get it wrong)

Could we have some discussion of and notes on result/partial result caching 
please?

- Dan

-------
Sender Permitted From: http://spf.pobox.com/
Archives at http://archives.listbox.com/spf-discuss/current/
Latest draft at http://spf.pobox.com/draft-mengwong-spf-02.txt
To unsubscribe, change your address, or temporarily deactivate your 
subscription, 
please go to 
http://v2.listbox.com/member/?listname(_at_)©#«Mo\¯HÝÜîU;±¤Ö¤Íµø?¡