ietf-nntp Backfill

[markd at mira.net.au]

> > Yet several people at the BOF thought that backfilling was reasonable.
> 
> I did not hear any server people state that backfilling was reasonable
> behavior.  On the contrary, most server implementors thought this was a 
> bug in some implementation.
> 
> I'd like to know exactly what server supports backfilling today.
> If nobody responds, I propose we close this issue and move on.
> We have bigger fish to fry.

I'd be curious to know why too. Does backfilling simply reflect an
underlying allocation scheme that happens to find the first free
number within a range?

Further, backfill most cause confusion for a reader, what if:

	Server				Reader
	------				------
T+0	new article arrives
	allocated # 10
T+1					Reader reads article 10
					notes that #10 is read
T+2	article 10 cancelled
	
T+3	new article arrives
	re-allocated # 10
T+4					Reader never sees the new #10
					as it is marked as read


Or does the reader have to consult the message ids of the article and
track message ids. Ie, a reader history database - ug.


Regards.