On 6/7/2012 9:21 AM, Rich Gray wrote:
> William Betts wrote:
> It seems to me that without a mechanism provide the caller with
> information about WHICH address(es) failed, it is appropriate to give
> up on any failure and abort the transaction.
I've been tossing that around, and though it makes sense from a certain
point of view, it is a very brittle way to program. If I were sending an
email with data on it to multiple recipients, I'd rather have it go
through to someone than to be discarded- considering how common it is
for mistyping of addresses.

> At best, one could set an option to Allow Partials which would permit
> the transaction to proceed as long as at least one address was
> accepted. A 'Some Sent' status would result in that case. This seems
> generally unsatisfactory though.
>
It seems better to have some sent rather than none- even if I didn't
have a great way to determine who it was sent to. I've been toying with
the idea of parsing the verbose debugging text for the addresses that
were invalid as a temporary workaround- though I'm not sure if that's
possible at all and I would be at the mercy of changes in the debugging
output in the future.

Another alternate (simple and rather crude) solution might be to mimic
mail servers in that curl could do "bookkeeping" of the unsent addresses
and send an "undeliverable mail" email to the return address (if it
isn't <>) for the unsent addresses. I'm not sure that is at all
desirable - since curl isn't a mail server that can't finish a delivery,
but rather a client who cannot start one.

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html
Received on 2012-06-07