[Wylug-help] Fetchmail setup

[A J Cole]

I am replying to this out of order and it may well have already been
answered on the list (my phone has been ringing at lot!).

>At Thu, 25 Nov 2010 11:29:47 +0000,
>Simon Brown wrote:
>> So to avoid this annoyance and as I don't really use postfix anymore I
>> decided to look at eliminating it. I'm now using procmail as the
>> fetchmail mda and so far it seems to be working. 
>
>Cancel that, mail is still getting dropped into /var/spool/ This is
>rubbish.

Well I suppose you can use procmail as the MDA for fetchmail (have you
seen that described somewhere? What about asynchronous errors?) but in
that case you will probably get caught by procmail's default delivery
behaviour which is to deliver to the user mailbox (unless a rule has
"taken" the message) or signal its MTA (fetchmail) on a failure to n
the message.

Rather then dropping the mail on the floor it should perforce go to
/var/spool/mail on error or to the user mailbox on success (unless a
rule intervenes).

>So any bright ideas as to how to ensure my mail is properly delivered
>or left on the POP server greatly appreciated. I don't really want to
>go back to running a rather redundant local smtpd though.

Haven't you burnt your boats - as far as a (unaugmented) POP server
is concerned if fetchmail determines it has the (filestore) resources
to download?  I assume fetchmail checks for space for it to spool and
normally runs in delete message from server mode?  There might be more
chance with IMAP since fetchmail might then only deletes the upstream
message when the MTA signals it is safe to do so.

But you still have the problem that the MDA (for several reasons) can
soft-error in which case its the MTA responsibilty to not drop anything
on the floor (thats mostly what /var/spool/mail is for).  On hard error
does fetchmail have even vestigial SMTP capability to properly handle
asynchronous errors (I doubt it)?

Traditionally many people (well I did) ran sendmail non-daemon on
workstation (for outgoing mail) BUT you always needed to cause the
occasional sendmail sweep of the local /var/spool/mail (using cron)
in case something was "dropped" because of an upstream reject (from
the equivalent of a "smarthost").

I don't feel that fetchmail (as MTA) with procmail as (NDS) is viable.

Andrew