Text 11184, 65 rader
Skriven 2005-07-20 18:55:54 av Matt Bedynek (1:106/1)
  Kommentar till text 11176 av Ross Cassell (1:123/456)
Ärende: Fidonews Echo Rules
===========================
Hello Ross.

20 Jul 05 18:04, you wrote to me:

 RC> .. such high end methods are passe if not overkill.

It only becomes overkill the moment it requires intervention.

 RC> .. the flow of echomail in our zone is not anything to fret over ..

Michiel brings up an interesting point.

 RC> Bitching about dupes here and there is a waste of time, ..

If someone notifies another operator about looping conditions and those other
operators do nothing to correct the problem, that is being lazy.  I do not
demand others run the software or implement the methods here, but if I send you
(example) a notification of a problem and nothing is done that is plain sloppy.
Maybe I outta whip something up that will post notifications in fn_sysop to get
some egg in the face of those are involved.  :-)

 RC> .. forcedpath checking is strongarm operations because if the
 RC> conventional wisdom is that we dont have the time to keep a ERN list
 RC> then we certainly dont have time to keep a similiar database ..

Database?  I have five links in the forcepath setup.  That hardly qualifies as
a "database".  My last modification to the configuration was when Grant
unlinked with your node.  I probaly should have never placed his link there in
the first place since he is one of those that plays musical links.

Though, I have thought about implementing a stateful database, one that would
track the number of messages from each node.  It would monitor and learn input
trends and temporarily rate limit if it exceeded.  While you consider such
things unimportant, imagine if a system were to suddenly dump 100K+ messages on
distribution (greater than 1yr in age) easily surpassing in one swoop the
database capacity for all nodes (even hubs).  Not only would this make those
systems explode it would clear (expire) their current dupebase.

Such methods could even be used for attack on Fidonet itself, similar in
method, to a DOS or DDOS attack.


 RC> .. for echomail paths for forcedpath checking to really work and with
 RC> echomail flow as it is, I dont want a message languishing on my system
 RC> sidelined until I get up in the mornings or until I return home from
 RC> work, because it came in all of a sudden from a different path. This
 RC> method worked in busier times, I dont feel it is warranted any longer.

There are times when I do not login to the node for weeks at a time and nobody
has ever complained about a late or missing message.  They have, however,
complained elsewhere, where such methods of checking are not even used.

I have trapped about 10K messages in the last few months using the methods
described about.  Obviously, the benefit exists, IMO.

 RC> The majority of my dupes appear to originate on 123/140 ..

If the message body is the same, hpt will detect those.

Matt

e-mail: matt [at] thunderdome.us | icq: 16568532 | yahoo: mbedynek

---
 * Origin: ..::[ high speed feeds - http://fido.thunderdome.us/ ]::.. (1:106/1)