Re: #1416

[Russ Allbery <rra@xxxxxxxxxxxx>]

Harald Alvestrand <harald@xxxxxxxxxxxxx> writes:

>> -            <t>The injecting agent MUST then add an Injection-Date header
>> -            field containing the current date and time.</t>
>> +            <t>If the proto-article already had an Injection-Date header
>> +            field, it MUST NOT be modified or replaced.  If the
>> +            proto-article had both a Message-ID header field and a Date
>> +            header field, an Injection-Date header field MUST NOT be
>> +            added, since the proto-article may have been multiply injected
>> +            by a posting agent that predates this standard.  Otherwise,
>> +            the injecting agent MUST add an Injection-Date header field
>> +            containing the current date and time.</t>

> This ("if the prot-article had both a Message-ID header field and a Date
> header field") is the place that bothers me most, since it will mean
> that singly injected articles produced by "modern" UAs that generate
> their own message-IDs will never get an Injection-Date header field.

Yes, this is the point of disagreement.

> It seems to sacrifice a general goal (that messages should have
> Injection-Date) for a narrow benefit (that multiple copies of multiply
> injected articles that are injected over a large span of time won't be
> detected as duplicates by servers that check Injection-Date and ignore
> Date).

> I'm not sure that's the right tradeoff.

This is the topic that we've gone back and forth on at some length.  To
summarize my opinion:

* I'm not sure that having all messages have Injection-Date header fields
  should be a general goal of this work.  The header is primarily useful
  when Date is well in the past and not that useful otherwise.  It's
  certainly fine if we get there, but I don't think getting there is a
  high priority.

* The Injection-Date feature will not realistically be able to be used to
  allow for articles with very old Date headers for the forseeable future,
  possibly ever.  It will require upgrading all servers to use
  Injection-Date rather than Date, which I think is very unlikely to
  happen.  Accordingly, the benefit of servers adding Injection-Date is
  fairly minor for the forseeable future; the only effect will be somewhat
  more correct behavior with articles with borderline Dates.

* Given the very slow development pace of netnews software, I think
  backward compatibility should be our highest goal.

I'm unlikely to change my mind in this area, but I'm happy to be
overridden by the working group if other people just don't agree with me.
If that happens, I'll gladly prepare a document that implements Charles's
approach.

I think a working group chair needs to decide if that's happened.

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