[NNTP] AUTHINFO diffs

[Jeffrey M. Vinocur]

On Jun 8, 2005, at 8:17 PM, Russ Allbery wrote:

> Jeffrey M Vinocur <jeff at litech.org> writes:
>> On Jun 8, 2005, at 1:09 PM, Ken Murchison wrote:
>
>>>       Future extensions may add additional arguments to this 
>>> capability.
>>> -     Unrecognized arguments SHOULD be ignored or brought to the
>>> -     attention of the user.
>>> +     Unrecognized arguments MUST be ignored by the client.
>
>> I'm not sure I agree with this change.  Perhaps the user would like 
>> to be
>> notified of new AUTHINFO methods.  Regardless, this is a user 
>> interface
>> issue and not a protocol issue, so I'm skeptical of the 
>> appropriateness of
>> dictating it.
>
> Isn't bringing something to the attention of the user still allowed 
> under
> the conventional interpretation of "ignored by the client"?

I...dunno.  Do we use "ignored" in this fashion elsewhere?  It doesn't 
seem intuitive to me.


> I would tend
> to read that as "ignored for protocol purposes," not implying that the
> client can't take informational notice of them.

Hrm.  Can the client use that information to trigger an automatic check 
for a new release and present that information to the user?  Can the 
client refuse to authenticate until the user acknowledges the notice?  
I see no reason we should prohibit either of these behaviors, but I'm 
hesitant to say they qualify as "ignoring" anything.

So, then, what's the actual meaning of the "Unrecognized..." sentence?  
Is there any protocol problem we're at risk here?  The "Future 
extensions..." sentence is useful, but what does the second sentence 
add?


-- 
Jeffrey M. Vinocur
jeff at litech.org