
March 28, 2000
8 p.m.
"TW" == Thomas Wouters <thomas@xs4all.net> writes:
TW> But, assides from delivery, it might be useful to store
TW> messages which failed elsewhere in the pipeline too; messages
TW> in the archive pipe, for instance, or the usenet pipe. It can
TW> currently happen, for instance because of a deadlock, that
TW> messages just get lost. I haven't looked at the new code yet,
TW> but imho it shouldn't be too hard to push messages back into
TW> those pipelines, assuming they fail 'cleanly' (and not with
TW> files half-written or some such.)
That's actually a good idea. I think a wrapper around the pipeline loop, perhaps using a bare try/except (hmm...) is the way to go. What you'd probably have to do is have a checklist of delivery modules so you know 1) which ones you wanted to send the message through; 2) which ones failed. And then to know what the disposal is for a message that failed at a particular step. Definitely more complicated, but worth thinking about. Robustifying message delivery should be very high on the list, but for 2.0 final we'll have to find a happy compromise.
TW> (Then again, I haven't seen failures at all, yet, so I'm not
TW> too worried for myself.)
Me neither! :)
-Barry