[Dxspider-support] installing dx-spider

[Dirk Koopman]

On Wednesday 14 March 2001 10:38, you wrote:
> Hi
> As a coming user of spider  I feel free to adress  some comments /questions
>  to this board.
>
> For the moment I am running ak1a for 10 years ( ON0DXK)
> I was one of the guys asking for a CD to install spider on a more easy way
> but I managed to do it without althought I am still interested to receive
> the CD when it comes out. The installation was a lot easyer I thought hi
> .My linux knowledge was almost zero and still is very poor .Reading and
> reading again the "RedHat" manuals and the howto's did help and when in big
> troubles having someone to relie on was all I needed.
> It is also a must to write down all the steps you make and also the most
> linux-commands you regulary need .
> The documentation delivered with the spider-distribution  is ok for me,no
> critics. Many thanks to Ian and of course to Dirk for writing all this code

Thank you for that Roland, hopefully the full user documentation will be on 
the website RSN.

> ! The ax25 kernel and utils are in my opinion a difficult task but with 
> the help of a written procedure made by Gert,ON1BLU  I could set it up
> without many difficulties. On the linuxbox I am running also the node (
> linuxnet ) connected to the ax25 rf-network via a kiss port and via
> ethernet.
> If all goes well I am planning to install the the linuxbox  very soon at
> the definitive site.
>
> But now my questions :
>
> 1.Database chaining : QSL databases f.e.
>
> The database-chaining is ok but the output  is not the way it is  now on
> on0dxk (ak1a) for many years.
> I have 5 databases having each there own properties.
> The output for "key" ( if  found ) is  displayed  for each database that
> way the user can interprete  what info is the correct one at that time.
> If "key" not found  no display .
> Whit spider once the "key" is found the other databases are not searched
> any more. I  miss also the .pre  and  .nof   files and the  upload facility
> ( f.e. upload/qsl)  from  the ak1a  system .
> A few time ago I think it was Dirk who you asked  for wanted new features .
> Maybe the ak1a system for the databases is one hi !

The database facilities are a bit basic. They are due to be over hauled after 
I have finished the current tranche of work on the basic internals of the 
software.

> The qsl-database on on0dxk is very popular and much used specialy because
> there is very recent info in it .F.e the qslinfo taken out the dxspots by
> f6cyv every 2-3 days. He uploads the new info's ( file ) and during
> maintenance the file is added and dupes are cleaned, all written in
> dbase(clipper).
> The qslnew-database updated by users is very interresting to.
> All this make it unnecessary for users to give all this ennoying announces
> (pc12) we see now on most clusters.

Try: sh/dx ey8xx qsl

(for instance) you don't really need the qsl_new database because it is built 
in to the sh/dx command (you can search for iota and qra locators as well, 
see help sh/dx)

>
> 2. I heard abt a problem when more then +- 50 users are connected .
>
> It was not sure yet spider was the cause .
> Because On0dxk has often more then 50 users and during contests the users
> are at their maximum ( 62 ) I am very interrested to have follow up of the
> problem and for this reason I will wait to shut down the ak1a cluster for
> the moment ..

The current development version in CVS should be less cpu and memory hungry 
(and it is lack of RAM that is usually the problem). However this version is 
not for beginners and, currently, cannot do OUTGOING ax25 connects. 

In general terms you should remember that this program is written in an 
interpreter and will run out of CPU quicker than a program written in a 
compiled language like C or C++. Having said that, on a celeron 500 with 
128Mb RAM I have seen upwards of 50 users + 6 nodes and had no problems (and 
room to spare for more). 

Having enough RAM is the important bit, if the machine starts to swop then 
the performance will drastically reduce and it won't work. There are those 
who are running DXSpider + clx + xfbb all on the same 200Mhz pentium 1 with 
only 128Mb. As we have found out the hard way. this will definitely not 
support 50+ users! (although you stand a MUCH better chance with the absolute 
latest CVS using the internal listeners [hint, hint]).

There are some that have questioned the use of perl (as opposed to C or C++). 
The reason I chose it was that:

1) I needed to learn it for www type work. 

2) I wanted something that was quick to develop and modify rather than 
necessarily super efficient.  

3) following on from that I am not at all certain that I like the concept of 
large 'super-nodes', I only have a 64Kb line here and I would not like 60+ 
users on-line permanently as it will rob me of too much of my bandwidth.

4) I, initially, started with wanting to make a 'router' that fixes the 
protocol problems in the PC protocol rather than a full blown 'user' node. 
Unfortunately, it turned out that I had to do a large part of the 'user' 
interface in order to get accepted so that I can now (finally) get on with 
sorting out the PC protocol!

Dirk G1TLH