[NNTP] Extension snapshots 2

[Ken Murchison]

Clive D.W. Feather wrote:

> Russ Allbery said:
> 
>>I do agree that the MODE_READER capability
>>MUST NOT be advertised (since that invites clients to use it),
> 
> 
> Oh hell, I overlooked that. I need to change some of the pre-6 wording.

I have other pre-6 comments that I'm accumulating, but since you've 
singled out this text, I'll comment here.


> 3.4.2 now says:
> 
>    A mode-switching server has three modes:
>    o Initial mode, after the initial connection:
>       * it MUST advertise the MODE-READER capability;
>       * it MUST advertise the IHAVE capability;
>       * it MUST advertise the READER capability with the -MODE-READER
>         modifier;

Since you state elsewhere that capability modifiers aren't required to 
be used by a server, I object to the MUST above.  Either change the MUST 
to SHOULD, or rewrite it to something like:

"* the READER capability MUST NOT be advertised unless prefixed with the 
-MODE-READER modifier;"


>       * it MAY advertise other capabilities with the -MODE-READER modifier;
>         if these have other modifiers as well, -MODE-READER should be first.
>    o Reading mode, after a successful MODE READER command:
>       * it MUST advertise the READER capability;
>       * it MAY advertise the IHAVE capability;
>       * it SHOULD advertise the MODE-READER capability with the -- modifier;

Same objection as above.  Perhaps:

"* it SHOULD NOT advertise the MODE-READER capability unless prefixed 
with the -- modifier;"


>       * it SHOULD advertise, without the -MODE-READER modifier, any other
>         capabilities it was advertising with that modifier.
>    o Transit mode, after any command except CAPABILITIES or a successful
>      MODE READER command:  
>       * it MUST advertise the IHAVE capability;
>       * it MAY advertise the MODE-READER capability; there are
>         circumstances where it SHOULD NOT do so.
> 
> and MODE READER now has added:
> 
>     The server MAY cease to advertise the MODE-READER capability once
>     the client is no longer permitted to use the command, and SHOULD do
>     so if the command would have unexpected effects.
> 
> I can easily be convinced to change the last two SHOULDs to MUST.

I'd in favor of MUST unless there is good reason not to.

-- 
Kenneth Murchison     Oceana Matrix Ltd.
Software Engineer     21 Princeton Place
716-662-8973 x26      Orchard Park, NY 14127
--PGP Public Key--    http://www.oceana.com/~ken/ksm.pgp