[NNTP] Adding A Quota Command to NNTP (internet-draft)

Ade Lovett ade at lovett.com
Thu Feb 22 02:57:30 PST 2007


On Feb 05, 2007, at 20:27 , Russ Allbery wrote:
> Interesting.  I can't evaluate the general usefulness of this,  
> since I'm
> not at all involved in the binary or commercial provider world  
> where this
> is more common, but from a protocol perspective it seems like a  
> reasonable
> extension.

My primary concern over this would be whether the proposal would be  
expecting real-time status of quota usage.  In the commercial world,  
it is not uncommon to have a reasonable number of individual front  
end reader boxes behind an L4 load balancer, further, until sessions  
are actually closed and then some time after that, when the  
accounting data has been processed back into a central repository,  
would the usage limits be updated.

With these two issues in mind, it is therefore entirely possible for  
out-of-date (ie: incorrect) information to be passed back to the  
client (think 2 connections to news.example.com, via an L4 balancer,  
going to two completely different machines, the only commonality  
being the same authentication tokens have been passed).  Thus I'm not  
sure that the extra workload of "hey, your quota thing sucks! it told  
me I had <x> GB left, when I really only had <x-dx>!" on a support  
desk is worth the effort.

> I would use full keywords rather than single letters for the types of
> quota; it's more extensible.

Absolutely agreed.

> This does raise a different question of whether there should be  
> some way
> for the server to return timeout and maximum article size  
> configuration,
> although that's not quite the same problem and probably shouldn't  
> use the
> same command.

Definitely needs to be another extension, if implemented at all.  I  
guess the main question would be what kind of "server information"  
could possibly be made available to the client beyond timeout/ 
maxartsize.

-aDe



More information about the ietf-nntp mailing list