History API docs

Alex Kiernan alexk at demon.net
Tue Mar 6 11:55:21 UTC 2001


Russ Allbery <rra at stanford.edu> writes:

> Looking this over, this looks good, except that I think we do want to add
> another interface besides HISwrite to write entries that are just a
> message ID (since some history mechanisms may only implement that
> interface and not HISwrite).  Something like HISremember?  Not sure of a
> good name.
> 

Agreed, its awkward the way it is at the moment, plus it has something
of a feel of knowledge of the internals about it.

> >     HISexpire expires the history database associated with
> >     *history*. Entries in the database with an arrival time less than
> >     *threshold* are automatically deleted from the database, messages
> >     with an arrival time greater than or equal to threshold are passed
> >     to the discrimination function *exists*.
> 
> See my other message for how I think this should work... I think that the
> automatic deletion should only apply to entries without an associated
> token (the rest should be passed to exists).  I'm not sure if entries with
> arrival times newer than the threshold should be passed to exists or not.
> 

That makes sense - only after going and looking at again was it
obvious that if you don't pass all the tokens through, you'll never
actually remove articles from the store, for those storage mechanisms
which need explicit removal.

> This all looks *really* good.  Thank you!
> 

I'm hoping to get a patch done this week - I'm to the point where I'm
fiddling around the edges & need to get proper feedback on it all.

-- 
Alex Kiernan, Principal Engineer, Development, Thus PLC


More information about the inn-workers mailing list