[NNTP] Consensus?

Peter Robinson pmrobinson at gmx.net
Sat Aug 20 08:09:26 PDT 2005


Clive D.W. Feather <clive at demon.net> wrote:

> Where are we on reaching consensus on the following items?

I was away when the article number issue came up most recently, but here
is my opinion for the record.

> (1) Article number limits. In particular, the possibilities of:
>   (a) changing the limit from 2^32-1 to 2^31-1;

The main benefit of the current work is to present a clear, accurate and
well specified account of the current state of NNTP protocol.  New
software authors should be able to read this spec and have a good idea
of what they will find out there.  This is not the time for significant
new work.

RFC 977 did not specify a limit for article numbers.  That isn't because
it expected implementations to cope with arbitrarily large numbers, it's
because it is a product of its time.  This is just one of the many ill
specified parts of NNTP that we are cleaning up.  Changing the de-facto
limit from 32/1 to 64bits is very far from an insignificant change, and
one that we certainly cannot make at this stage.

Current clients and servers do not cope with 64 bit article numbers.
That will remain the case whatever this document says.  I strongly
believe that the way forward is via a future extension along the lines
of the BIGNUM proposal, where the capacities of both sides are exchanged
as necessary.  We should not be declaring 32 bit implementations
(conditionally) uncompliant with SHOULD or MUST language for 64bit
support.

As for the specific 31/2 bit question, I suggest we should face reality
and change to 2^31-1, but I have no strong feeling either way.

>   (b) adding wording saying the present situation is unsatisfactory but
>       is being worked on ready for the next version of this document.

That is probably a good idea.

> (2) Limits on the length of command name and capability labels.

Remove the limits.

Regards,

Peter



More information about the ietf-nntp mailing list