ietf-nntp "Common NNTP Extensions" document updated

[Clive D.W. Feather]

In message <E0xdccd-0002tg-00 at gorgon.lovett.com>, Ade Lovett
<ade at demon.net> writes
>>> We don't use INN in this part of the system, and I doubt our code would be
>>> of much use to anyone else.
>>> I think the chief developer lurks on this list, and he may well comment
>>> himself.
>>I thought I did already, but anyway.... I think NEWNEWS is useful. INN will
>>get a better implementation, but not in time for Stan's documents.
>I think Clive was actually talking about me, James.

Correct.

>our transport
>system does (amongst other things) provide for an optimised NEWNEWS
>method, but the code has digressed so much from stock-standard INN
>(read, it's a complete ground-up rewrite), that it's not going to
>be easy to provide a patch.
>I'm pretty sure I sent you the concepts though in a private email
>a while back whilst we were discussing other aspects of INN performance,
>I can resend if necessary.

I might as well publish them (thanks to Ade for the description). The
essential step is to maintain, for each group, a file containing:
   <timestamp> <message-ID>
NEWNEWS on a group can then by done quite effectively by a linear scan
down the file. Even though it's still linear searching, because we don't
have lots of clients accessing the same single monolithic history file,
and we're only going to groups that the client was interested in, then
it makes things a lot less stressful on the system.

-- 
Clive D.W. Feather    | Director of Software Development  | Home email:
Tel: +44 181 371 1138 | Demon Internet Ltd.               | <clive at davros.org>
Fax: +44 181 371 1037 | <clive at demon.net>                 |
Written on my laptop; please observe the Reply-To address |