>> upon unexpected hangup
>> can't some kind of signal be automatically sent to all the
>> fetching stuff...
ab> On receipt of a signal the process that is doing the fetching will
ab> need to exit cleanly. The problem is that it is difficult to do this
ab> reliably since the process could be doing anything. It could be
ab> difficult to find an appropriate course of action to handle the open
ab> sockets and cache files that works for all cases.
>> hmmm, at least my ip-down runs, and inwhich i have
>>
>> wwwoffle -offline
>> fetchmail --quit
>> killall -HUP allsigs fetchnews
>>
>> perhaps i should also do
>>
>> killall -HUP wwwoffles
>>
>> there too?
ab> You could except that there are no processes called wwwoffles. This
ab> is a historical hangover from version 1 of WWWOFFLE where the child
ab> processes were a different program. Now that it forks off a copy of
ab> the same program they are all called wwwoffled, but wwwoffles is
ab> retained for the syslog output.
odd... then this needs to be made truthful and accurate for bug
hunters to not be looking for wwwoffles.
Anyway, I have the ability with killall from ip-down to send any certian/all the
wwwoffle subprocesses a kill -HUP signal ... which I hope they would
handle as "don't keep trying to read from the gone server, instead
just backoff and leave the item on the outgoing list [or whatever,
in case of partial downloads].
Anyways, I'm hoping every wwwoffle subprocess could be fitted with
procedures for what to do incase the line was hungup [or i send that
signal] at every state.
What I expect is that in the case the line has hungup, but the
wwwoffle subprocesses don't know about it yet, they are certainly in
some kind of "trying to read" state that will timeout in a minute... I
would hope when they get my signal they would think "oh he's telling
me to backoff and put the item back into the outgoing list, instead of
waiting a minute and then assuming the remote party is unreachable and
putting a 503 or whatever message into the lasttime list."
By the way, it would be neat if along with
[Outgoing] [Monitored] [Last Time] [Last Outgoing] [http] [ftp] [finger]
[Unsorted] [Modification Time] [Access Time] [Date Changed] [Alphabetical]
[Domain Name] [File Type] [Random]
there could also be a [Screwups]
which would be, say like the lasttime list, but instead with
categories
Pages that failed with error 503 [time out]:
Pages that failed with error 404 [bla bla]:
etc.
By the way, "[Date Changed]" makes mysterious bars that ought to have their
date meaning mentioned inside them: --- 3 days old ----
>> darn, with lynx's ^E i could get to the index fast, but now it's not
>> last on the page
>> [Delete|Refresh:Options|Monitor|Index|Configure] - WWWOFFLE
>>
>> hmmm...
ab> You can always edit the file that contains the message that is added
ab> to the bottom of the page.
well, i was thinking that others might like index last too.
>>>>> "Felix" == Felix Karpfen <[EMAIL PROTECTED]> writes:
Felix> Is the need to add `-c /var/spool/wwwoffle/wwwoffle.conf' a new
Felix> requirement needed for WWWOFFLE 2.7? Or have I just been lucky that the
Felix> routines worked without this addition?
this old notation should be swept out of all WWWOFFLE docs and
programs now that WWWOFFLE has finally gotten with the /etc config
file home default standard.
--
http://www.geocities.com/jidanni/ Tel+886-4-25854780
