ietf-nntp New wording on article numbers - draft 3

[Tom Hughes]

In message <Pine.NEB.3.95.970111174158.1799w-100000 at apb.iafrica.com>
          Alan Barrett <apb at iafrica.com> wrote:

> I don't see why clients "can't" deal with out-of-order arrival.  A few
> days ago I explained how a client can easily deal with out-of-order
> arrival.

I think it's not so much a question of can't as don't and/or don't
want to.

I'm a customer of an ISP that tried using a server that did it and
had to stop because of the trouble it caused. At least one popular
PC client now has the so-called "Demon option" that was added to
make it record enough history to cope with backfill.

Another customer did investigate a large number of clients on various
platforms and found that very few could cope with it - even the main
Unix clients which keep a newsrc that you would expect to have enough
history can apparently still fail to handle backfill properly in some
circumstances.

The question is, is the benefit gained from allowing backfill sufficient
given that servers probably won't be able to make use of it until client
authors have been persuaded to support it?

Tom

-- 
Tom Hughes (tom at compton.demon.co.uk)
http://www.compton.demon.co.uk/
...Nanu nanu!