Re: #1047 Path field delimiters - feed behaviour? (fwd)

[Bruce Lilly <blilly@xxxxxxxxx>]

On Wed June 29 2005 07:56, Harald Tveit Alvestrand wrote:
> 
> I haven't seen an answer to this question - would anyone else like to 
> answer?
[...]
> --On 23. juni 2005 17:05 -0700 John Stanley <stanley@xxxxxxxx> wrote:
> 
> > It would seem to be unwise to use it in IPV6 addresses as well, since
> > ::FFFF:129.144.52.38 and ::FFFF:129.145.23.23 are clearly two different
> > addresses but the same failure that would make the dead:beef admin an
> > idiot would make both of these people idiots. (I.e., non-conforming
> > servers would match the "path-identity" of FFFF for both systems and
> > feeds  would be disrupted.)
> 
> If, in a message arriving at the old system "here", there is a Path: header
> that says
> 
> Path: there!::FFFF:129.144.52.38!not-for-mail
> 
> and "here" does not know about any system named "FFFF", will any change in
> handling happen?

Change relative to what?  Does "here" "know" about a system "named"
"129.144.52.38?   The root of the issue is that colon always has been
and remains today a delimiter.

> Conversely, if a message arrives with a Path: header that says
> 
> Path: there!::FFFF:129.144.52.38!::FFFF:129.145.23.23!not-for-mail
> 
> will the two occurences of "FFFF" in the Path: cause anything bad to happen?

Not that I know of.  Some statistics gathering scripts might interpret
that as a loop of some sort.  No problem for propagation. (assuming of
course that the conditions for the previous question still apply)
 
> Or is there some third circumstance that I haven't guessed at yet that will
> cause something bad to happen?

Yes,

Path: there!FEDC:BA98:7654:3210:FEDC:BA98:7654:3210!foo

(example IPv6 address literal from RFC 3513 section 2)  Will cause the article
not to be propagated to any sites "known" to "here" "named" "FEDC", "BA98",
"7654", and/or "3210" (a less repetitive IPv6 example would of course provide
more opportunity for damage).

The RFC 3513 example and description of the problem has been previously
described in http://www.imc.org/ietf-usefor/2004/Jul/0192.html
which was part of a discussion also referenced in recent discussion.

An example of a highly problematic IPv6 literal incorporating observations
made by Frank and Russ would be DEAD:BEEF:FEE:BEE:CAFE:ABBE:ED:BE.