cks@ziebmef.uucp (Chris Siebenmann) (12/22/88)
[A good illustration of why agressive rerouting is BAD.]
The unicus<->ziebmef link (over which pathalias loves routing things)
will be dead by the end of the week due to unicus's departure from the
net. Until the new maps come out, it would be greatly appreciated if
people could mark the link as dead. Active rerouters are especially
asked to do this.
--
"You're a prisoner of the dark sky/The propeller blades are still
And the evil eye of the hurricane's/Coming in for the kill"
Chris Siebenmann uunet!utgpu!{ontmoh!moore,ncrcan}!ziebmef!cks
cks@ziebmef.UUCP or .....!utgpu!{,ontmoh!,ncrcan!brambo!}cksjmm@ecijmm.UUCP (John Macdonald) (12/24/88)
In article <1988Dec22.005827.14253@ziebmef.uucp> cks@ziebmef.uucp (Chris Siebenmann) writes: >[A good illustration of why agressive rerouting is BAD.] > > The unicus<->ziebmef link (over which pathalias loves routing things) >will be dead by the end of the week due to unicus's departure from the >net. Until the new maps come out, it would be greatly appreciated if >people could mark the link as dead. Active rerouters are especially >asked to do this. > Much as I agree that active rerouters tend to be a bad thing at least as often as a good thing, the above is not a good illustration of it. It has exactly the same good/bad trade-off as any other mail where there is no confusion about multiple sites with the same apparent name - the active rerouting is good if the maps at the active rerouting site are more up-to-date than the maps at the originating site and bad if the they are more out-of-date. If you assume that active rerouters accept a certain responsibility for their maps, updating them rapidly and constantly as postings like the above one are posted, then it is more likely that active rerouters are a good thing - it is certain that many of the sites that pass mail through them do not do any more than use new postings of the official maps (and many don't even do that). In fact, I would argue that this is one of the few occassions where active rerouting is justified. If mail came into site 'fred' with a path of '...!fred!ziebmef!unicus!...' during the next few weeks, rewriting this path as '...!fred!other!way!unicus!...' makes good sense. It makes sense because the rerouter: 1) knows the entire path that it is rewriting 2) knows a good alternate path. 3) knows that the specified path is bad The place where active rerouters break down is in 1 - they reroute where they think they know the destination, without knowing that the specified path is really intended to reach the destination they have chosen. The shameful thing about active rerouting is that it arbitrarily assumes ignorance on the part of the originator. It is far better to only do the active rerouting in the case of proven ignorance (like the situation above), and only do passive rerouting otherwise. This allows the originator to choose to have the rerouting done, without forcing it. (For example, our maps tend to be fairly complete for local and regional systems, but everything else gets punted to uunet for them to passively reroute.) -- -- John Macdonald