[Cluster-tech] PC93s
Dirk Koopman
djk at tobit.co.uk
Mon Sep 17 12:53:24 CEST 2007
Lee Sawkins wrote:
> Dirk
>
> Do Spider users see these duplicated talk messages, or does Spider have
> a dupe checker for PC10s? CC Cluster doesn't, as there is no reliable
> way to do it as far as I can see.
Well there is. The way I deal with all sentence dupes (and I am talking
about anything other than announce, chat and spots) is I have every PCxx
sentence go through an "ephemeral dupe" hash table. This simply stores a
copy of every sentence for a number of seconds (usually 60 secs, but
there are other defaults for PC15, 34 and 41). This stops the storms
that sweep the network otherwise.
However, this should only really occur if there the user isn't (yet)
visible as a direct route.
Also, despite my pleas, not everyone has upgraded even from the early
versions of the code yet. There were errors in my handling of talks.
I may have to force the issue by upissuing the version number. That
usually works...
>
> The problem is the conversion of PC93 to PC10. If the Spider nodes
> would send PC93s when possible instead of converting them to PC10s,
> then many of these dupes would go away as they could be dupe checked in
> the destination node before delivery.
>
> Another way is to simply not send user lists from CC Clusters to
> Spider. Then the PC93s would not be changed to PC10s.
>
> Lee
>
More information about the Cluster-tech
mailing list