[NNTP] [2503] Allow mode-switching servers?

[Ken Murchison]

Clive D.W. Feather wrote:

> Ken Murchison said:
> 
>>I'm in favor of B to clean up the protocol, but from an implementation 
>>standpoint does it really matter?  Isn't the cat already out of the bag 
>>on this one?
>>
>>I think its already been determined that given its current architecture, 
>>making INN a full-time general purpose server is a non-starter, at least 
>>for the short term.
> 
> 
> I thought Mark Crispin was going to take a six-pack of beer over to Russ
> and fix that code? Or did I misunderstand?

I also volunteered to fix the code (or at least help drink the beer).


> How are we going to resolve the stuff about the ordering of AUTHINFO/TLS
> and MODE READER?

Good question.  I don't see any *easy* way of keeping TLS or SASL 
security layers alive across a MODE READER if it spawns a separate 
process.  If we put the TLS and/or SASL context in shared memory, then 
it would survive the fork()/exec() and could be retained, but this 
probably isn't portable on all platforms that INN runs on.  Or we could 
make innd multi-threaded.


> Question: if mode-switching stands, are people happy with the way I present
> it in pre4? In particular, Ken, are you happy that this better addresses
> the stuff about using extensions before and after MODE READER?

Can you direct me to the text to which you refer?  I don't recall seeing 
any reference to MODE READER and extensions.

-- 
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