[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: #1416: USEPRO 3.9: Reinjection and Injection-Date
Forrest J Cavalier <forrest@xxxxxxxxxxxxxxx> writes:
> Russ Allbery wrote:
>> Forrest has proposed only allowing reinjection between disjoint networks.
> Well, no. I was coerced. My original text said little. Then I upped
> it to handle leaf nodes. Then I was coerced to consider the disjoint
> network multi-injection. This WG is a broadening experience, doncha
*heh*. Sorry if I mis-stated there! :)
> If you are a private leaf node, you know you want to reinject articles
> posted locally and no others. In this case, you KNOW articles did not
> pass through A nor B, because they originated at your node and you
> haven't relayed them yet, since you don't relay. QED.
> You can implement this many different ways. I don't think we need to
> standardize that implementation.
That makes sense. Basically, if we say that you're only allowed to do
reinjection that changes the identity of the article (dropping
Injection-Date) when you *know* that the networks are disjoint, and leave
the problem of knowing that to the implementation, it handles the simple
leaf-node case while leaving the question of how to determine this in more
complicated cases to the individual.
We can still get problems if that knowledge is wrong, of course.
> Let those with other cases read Duties of Gateways. They don't get a
> standard formula, because this WG doesn't have a consensus about their
> needs and prevalance, and the Injection-Date scheme that was proposed is
> now consensed to be faulty (although different WG members think it is
> flawed for different reasons.)
There is that -- the Gateways section allows for Netnews-to-Netnews
gateways with more general descriptions of what sorts of care one has to
use, and is available as a fallback description of what has to be done for
weird and difficult cases.
Russ Allbery (rra@xxxxxxxxxxxx) <http://www.eyrie.org/~eagle/>