>> now that I have a bit more experience with my own tool and the 'hard
>> cancellation' (coded with a CURLE_CANCELED error code passed to
>> curl_ftp_done, currently). this way seems rather acceptable, even
>> performance-wise. what I would like is then to progress first on this,
>> having something official and working. this thread is still interesting
>> because we viewed the main problems i think and we could keep in archive
>> for later improvements.
>>
>> so, from my point of view adding a simple test of the multi state wrt
>> COMPLETED and setting some error code would OK for that purpose (which
>> one? there is nothing very appealing in the list).
>
> Sure. I think Curl_done() needs a new argument named "premature" or
> something like that, so that protocols that care can get to know about it
> and act accordingly.
>
> Wouldn't that be good enough?
perfect, i'll download the latest version and propose a patch asap (it will
probably patch all curl_???_done functions I imagine, adding the "premature"
parameter, and do something cool for FTP only).

Armel
Received on 2006-12-23