[Dxspider-support] Problems with command " dx by "

[Dirk Koopman]

On Mon, 2003-12-29 at 13:50, Ian J Maude wrote:

> OK, the question here, it seems to me is...
> 
> Do we continue with PC protocol or not?  I might be wrong (and I usually
> am ;-)) but the difficulty has been back compatibility in the past.
> So, making that assumption, do we scrap the whole idea of
> inter-operability with non-spider clusters?  Back to the old debate I
> guess.  From my point of view, I would go for a completely new protocol
> and AK1A protocol be damned.  As long as we keep the compatibility, we
> are potentially holding back the future development of cluster software
> and there is no real need for AK1A clusters to make the jump.  There are
> a LOT of spider clusters out there and the vast majority keep updated I
> think.

The issue for me is the 'state' information, routing and message (as in
mail) handling. Trying to maintain a map of users on nodes has (and
continues) to drive me mad. This is the fundamental reason why we cannot
happily 'talk' to each other. 

'Broadcast' things are not a particular problem to maintain
compatibility with (ie DX, Ann, WWV etc). Having said that, they are
limiting in scope and don't really allow for easy expansion for other
information. 

The PC50 is dead. Long live the PC50!

> 
> The next question is as difficult.  If we go down this track, how is the
> new code controlled.  By that I mean, how do we ensure that people
> upgrade?  Leaving it to them may cause more problems than anything
> else.  We have seen problems in the past due to sysops not updating
> their software to fix bugs etc, indeed we have seen questions on here
> from sysops seeing bugs that have already been fixed in CVS.  I guess
> the real question is, should there be a control of any kind?  That one
> is down to you Dirk I guess.

I will have to split off that development into a different project or
branch.

> So, are we back to Spinet again here?  It would be perfectly possible to
> set up a separate network to test new code with new protocol using a
> connection/hook into the old Spider code I guess.  A lot of people have
> moved from Windows to Linux now, although I have no numbers for sysops
> using Spider under Windows.  I remember a certain software developer
> saying that Spider would never be ported for Windows ;-)
> 

Not Spinet, at least not as a separate program. I discovered fairly
quickly that that was a non-starter. But yes, we are talking about a
separate network talking a new protocol and with most of the PC protocol
stripped out except for the absolute minimum needed for external
connection.

But the crucial difference is that, currently, all nodes are equal, in a
new version all DXSpider nodes would be equal and all other nodes just a
an input feed with limited output capability (just the actual DX,WWV,WX
etc info). I would like that external nodes and their users would be be
invisible (and vice versa), at least for any non-local nodes/users (so I
will live with directly connected nodes and their local users). 

Dirk G1TLH