Natterings about history files

[Forrest J. Cavalier III]

Is there an iterator other than HISexpire()?  It seems like
a perfectly good one, but just asking.


I think the HISexpire callback should be 

bool (*exists)(void *cookie, time_t arrived, time_t posted, time_t expires, const TOKEN *token);

(Having a cookie in any callback seems like a good idea,
but the other time_t items should go along because they
might be needed anyway in the callback, which should not
have to do a HISlookup())


---------------------------------------------

I don't suppose you would consider implementing a "cheap
lookup" which would only do something quick, and never
hit the disk.  Yes, it would be inaccurate, (false
negatives) but it can be used to speed up transfers
on average.  (Might need a companion function of
HISwrite_only_if_unique()....)

---------------------------------------------

Any chance of asynchronous API?  I know, it
makes your head spin.  

Imagine opening N fds to the history text file
(which may happen anyway because of multiple text
files) and setting them non-block.

---------------------------------------------

Tell us more about HISreplace()

---------------------------------------------

Is "key" a message-ID or a hash?  I vote for message Id
(do we get votes?), and please change the prototypes if so.

---------------------------------------------

What are the goals for this?  Speed?  Clean-ness?  Reducing
the size of the history file?

---------------------------------------------