> >From there... I remembered the partition status return code patch.
>
> > Anyway, I'll have to patch the byterange filter in Apache to account for
> > the change, and I'm looking for input on what the correct action is when a
> > failure actually occurs. Right now, Apache doesn't test for a failure to
> > partition (which would have been indicated by a NULL return value) at all.
> > So in the code below, what would be the right thing to stick in the "XXX:
> > handle error" spot?
>
> Dunno. Some kind of magic error bucket? Recovering from errors during filter
> processing is totally undefined. If you've sent data to the client, then
> there isn't much that you *can* do except to:
Recovering from errors during filter processing is completely defined. It
is VERY easy.
1) If you have never passed data to a lower filter, then you can pass an
eror bucket.
2) If you have passed data to a lower filter, then you log an error, and
return to the higher filter. This means that as far as the client is
concerned, the connection closed prematurly, but that is basically what
would happen in 1.3 if there was a fatal error during request processing.
Ryan
_______________________________________________________________________________
Ryan Bloom [EMAIL PROTECTED]
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------
