I am the assigned Gen-ART reviewer for this draft. For background on
Gen-ART, please see the FAQ at
<http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
Please resolve these comments along with any other Last Call comments
you may receive.
Document: draft-ietf-grow-filtering-threats-06
Reviewer: Robert Sparks
Review Date: 24-Jun-2015
IETF LC End Date: 2-Jul-2015
IESG Telechat date: not yet scheduled for a telechat
Summary: Ready with nits
From looking at the document history and list archives, this document's
been around for some time, and has had some editorial push already.
The unintended consequences it highlights are interesting, and it will
be useful to operators to know these possible causes of unexpected behavior.
I encourage another strong editing pass before publication.
This is being published as an IETF-stream document. When published it
reflects IETF consensus.
There are places in the text that I think are problematic given that
status. The issues are editorial, and I expect they will be easy to address.
The document uses "we" frequently. Originally, that meant the authors.
It's ambiguous what it means in an IETF-stream document. I suggest
editing out all occurrences. Try to avoid simply changing "we" to "the
authors" - find a way to reflect what the IETF is saying here.
Is the last paragraph of 4.1 an IETF consensus position on how operators
might charge one another? It would be good to find a way to word this
that look more like statements of fact and less like charging advice.
The document draws some conclusions that I think are unnecessary. For
instance, "Therefore, we conclude that the reactive approach is the more
reasonable recommendation to deal with unexpected flows." Why does the
IETF need to say that (and is it an IETF consensus statement)? It would
be enough, I think, to reduce the discussion in these sections to
calling out the issues with each approach.
Please simplify the sentences, and avoid passive construction. For
instance, "It can be considered problematic to be causing unexpected
traffic flows in other ASes." can be much shorter. After you do that, I
think you'll find it easier to identify and collapse sections of
redundant text.
RjS