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

[John Stanley <stanley@xxxxxxxx>]

Harald Tveit Alvestrand <harald@xxxxxxxxxxxxx>:

>I haven't seen an answer to this question - would anyone else like to 
>answer?

Russ, as I recall, explained the problem quite sufficiently.  Charles'
question does not demonstrate the problem, and I suspect was chosen
specifically because it does not.

The issue is not systems that have no special relationship to any of the
problem components of the path header. They will do nothing differently.  
That's pretty obvious. If I see "dead" or "dead:beef" in the Path header
field, and I neither feed nor accept articles from either, then of
course there is no difference.

However, if I feed site "dead" and I am RFC1036 conformant, then no 
article that has passed through the site "dead:beef" will I feed to site 
"dead". Nor will I feed the site "dead" any article that has passed 
through the site using the IPv6 domain-literal 
"ff::dead:128.99.32.1".

Furthermore, if I feed site "dead:beef" and am RFC1036 conformant, then I 
will happily feed any article I get that has already passed through 
"dead:beef" back to "dead:beef", since I will not see the path identity 
"dead:beef" in the Path header field, only the two sites "dead" and 
"beef". 

This is not a complete meltdown of the news system, but then, nobody has 
ever said it was. It is a serious disruption of the news feed to the site 
"dead", however.