Re: #1416 Injection-Date: proposed diff

[Russ Allbery <rra@xxxxxxxxxxxx>]

"Charles Lindsey" <chl@xxxxxxxxxxxxxxxx> writes:

> Yes, I suppose he is technically doing reinjection, since he injected it
> into his own server, and his own server then reinjected when it sent it
> on.

There's nothing technical about it.  That's exactly what he's doing.

> But that is little more than hair splitting, since the article as thus
> injected the first time is still a valid proto-article, and contains no
> evidence of the event apart from an entry in the Path (and odd entries
> at the RH end of the Path are commonly put there for all sorts of odd
> reasons).

It's not hair-splitting -- it's the entire point.  This is why strict
definitions of agents and strict statements about flow between those
agents are important!

Someone who takes an existing news article and injects it again, rather
than taking the same proto-article and injecting it in multiple locations,
is reinjecting.  They therefore are responsible for getting the details
right as described in our draft.  It's possible that this will mean
specifically *not* doing some things, rather than anything they have to
do.  But any time you do that, no matter why you're doing that, you're
doing reinjection and have to pay attention to the relevant parts of the
protocol.

> But that 'technical' reinjection was not the cause of the problem. The
> man was totally convinced that he was not doing multi-injection (and so
> maybe saw no reason to provide an Injection-Date).

He's doing multiple serial injection, not multiple parallel injection.
Yeah, we definitely do need to clarify terms here; I can see how it's hard
to have a conversation about this with the current terms.  I think we
should probably go back to the terms we were using before and use:

    multiple injection: sending the same proto-article to multiple
    injecting agents.

    reinjection: taking a netnews article, converting it back into a
    proto-article, and sending it to an injecting agent.

By those definitions, he's not doing multiple injection because they're
mutually exclusive.  (Well, unless his conversion process recovers exactly
the same proto-article as he had originally, which one can do with care in
a local news environment that's under one's complete control, but which is
not horribly interesting in the general case.)

> But, in fact he was multi-injecting,

He was injecting once and then reinjecting, thereby leaving himself open
to all of the possible problems created by running the article through an
injecting agent more than once.  He SHOULD be doing multiple injection
instead (as defined above), for precisely that reason.

Of course, he certainly MAY add Injection-Date headers regardless, to be
more robust against people (including possibly himself) violating that
SHOULD.

> And I did not necessarily mean to imply that that he had written that
> script himself - it could have been obtained from all sorts of sources.

This still isn't creating a situation where no one has responsibility for
knowing that reinjection is happening.

> And he would probably argue, also, that his whole setup - newsreader,
> CNEWS and all - could legitimately be regarded as one large MUA, since
> he intended it to appear that way to the world outside.

Posting agent, you mean, I assume.

Sure, if he wants to argue that, he can analyze it on that basis, but that
means he can't do IHAVE to any other site.  As soon as his supposed
posting agent starts also acting like an injecting or relaying agent,
doing things that posting agents don't do, it's clear that he can't treat
it as a single monolithic entity and has to analyze the components for
compliance.

-- 
Russ Allbery (rra@xxxxxxxxxxxx)             <http://www.eyrie.org/~eagle/>