On 22/Oct/11 23:43, Murray S. Kucherawy wrote:
> 
> Send three reports.  Consensus appears to be that this is easier
> than coming up with some crazy syntax to try to keep it all
> contained in one report, especially when these are meant to be
> machine-processed and not handled by humans; it's not much of a
> burden.

Agreed, even though some cases don't need a different syntax.

>> In addition, forwarding from a central host may be a technique worth
>> being mentioned, although it's quite expensive compared to syslog.  It
>> may be used for vetting and internal auditing, besides updating
>> delivery-results, reckoning ri=, and the like.  Worth?
>
> We haven't prevented that kind of design with the text that's there
> now, so I don't think any changes are needed there either.  It
> would be an odd thing to do, but the current text doesn't block
> it.

Perhaps it looks odd because I conflated a number of things.  Let's
postpone delivery-results, reassembling, and ri=.  Would it be helpful
to have some text like the following appended to Section 7.3?

  Alternatively, report generators can accept a configuration
  parameter that instructs them to generate reports only in certain
  cases and/or to send them to a mailbox different than the one
  discovered in the DNS.  The latter is supposedly an internal
  mailbox whose operator, either human or automatic, can examine what
  data is being sent to whom before allowing or denying the report to
  proceed.  The destination address discovered in the DNS is still
  available in the relevant header field saved in the second MIME
  part of the report, thus forwarding-after-vetting can be
  automatically assisted.

> I think we're ready for another WGLC, unless you have other cases
> you'd like to explore.

I'm looking forward to -04.

_______________________________________________
marf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/marf

Reply via email to