[Dxspider-support] A new version...

[Frank Hurlbut]

1.54 Dirk and thanks for the hard work
73
Frank KL7FH
----- Original Message ----- 
From: "Dirk Koopman" <djk at tobit.co.uk>
To: "The DXSpider Support list" <dxspider-support at dxcluster.org>
Sent: Wednesday, June 20, 2007 1:17 PM
Subject: [Dxspider-support] A new version...


> Well, it's been a long time coming, but I am going to release a "new"
> version. My problem is: which one do I release? I have a choice of two.
>
> The first one is essentially what is currently in CVS. If you have
> updated from CVSlatest.tgz in the last week, that is the one you will be
> running. It is very nearly the same as the one in SF. However, it's
> version number is 1.53 and the repository where it is stored has changed
> (I shall come onto that later).
>
> The second one has some fairly radical changes in protocol. It is also
> version 1.54. Essentially I have commandeered the PC9x space and I am
> (currently) doing routing and ann/talk/chat completely differently. This
>  system has been running at GB7DJK, GB7BAA, GB7MBC and WR3D for several
> months now and seems to be stable. I would dearly like to move over to
> this from now on, so that I can nail both routing loops + talk/chat to
> work over all DXSpider nodes + directly connected other cluster node,
> once and for all. It also gives me a better vehicle for other
> improvements (such as a decent traceroute and message/bulletin handling).
>
> There is one big "feature" though. In order to stop loops it is
> necessary to be completely ruthless about what a new style node is
> prepared to believe in terms of routes. This means that only non-new
> nodes that are *directly* connected will have their *local*
> configurations propagated. Anything beyond the next hop will not be
> propagated (although it is still available to that node locally - for
> what that is worth).
>
> The big difference for new style nodes is that a) all routing broadcasts
> contain their origin + the time they were issued b) only contain their
> *local* configuration (ie users and locally connected nodes) c) the same
> information for any of those directly connected nodes that are using
> old-style protocol and d) each node resends this information periodically.
>
> The other (smaller) difference is that you cannot mess with these
> routing broadcasts (using acc/route et al) because they are no longer
> what the supplying node *thinks* is the information (after it and any
> intervening node has applied route filters to it). It *is* the node
> information - from the horse's mouth.
>
> There is another big difference. I have moved the repository over to a
> machine that I control. I have become disillusioned with the delays that
> SF have for CVS, as well as with CVS itself. I am now using git
> (git.or.cz) for version control. Having said that, there is a basic CVS
> :pserver: style anonymous CVS access method (very similar to what most
> of you are using - just pointing to a different place) available as
> well. I imagine we start gently and carry on with that for now. When the
> time comes, and a decision is made, then everyone who wants to upgrade
> will need to overwrite their spider tree with the issued tar ball.
>
> So, poll time: which version should I release next? 1.53 or 1.54?
>
>
> Dirk G1TLH
>
> _______________________________________________
> Dxspider-support mailing list
> Dxspider-support at dxcluster.org
> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support