Routing and Filtering

From DXSpider Documentation Wiki
Revision as of 12:59, 11 March 2008 by G0vgs (Talk | contribs)

Jump to: navigation, search

Introduction

From DXSpider version 1.48, major changes were introduced to the way node connections are treated. This is part of an ongoing process to remove problems with loops and to enable talk and other functions to propagate across the whole of the worldwide cluster network. In fact, in a Spider network, it would be useful, perhaps even necessary to have loops. This would give real resilience to the network, meaning that if a link dropped, the information flow would simply come in and go out via a different route. Of course, we do not have a complete network of Spider nodes, there are other programs out there. Some of these do not have any protection from loops. Certainly AK1A does not handle loops well at all. It is therefore necessary to have some form of protection for these nodes.

In fact DXSpider has had a simple system for some time which is called isolation. This is similar to what in other systems such as clx, is passive mode. A more detailed explanation of isolation is given further below. This system is still available and, for simple networks, is probably all that you need.

The new functionality introduced in version 1.48 allows filtering the node and user protocol frames on a "per interface" basis. We call this route filtering. This is used instead of isolation.

What this really means is that you can control more or less completely which user and node management PC protocol frames pass to each of your partner nodes. You can also limit what comes into your node from your partners. It is even possible to control the settings that your partner node has for the routing information that it sends to you (using the 'cmd' command).

Route Filters

Initially when route filters were being tested we generated a "default" filter. Unfortunately it quickly became apparent that this might suit the UK cluster network but didn't really fit anybody else. However using a default filter is an appropriate thing to do. How, is explained further on.

The first thing that you must do is determine whether you need to use route filtering �[1mat all�[22m. If you are a "normal" node with two or three partners and you arranged in an "official" non-looping tree type network, then you do not need to do route filtering and you will feel a lot better for not getting involved. If you are successfully using isolation then you also probably don't need to use route filtering.

To put it simply, you should not mix Isolation and Route Filtering. It will work, of sorts, but you will not get the expected results. If you are using Isolation sucessfully at the moment, do not get involved in Route Filtering unless you have a good supply of aspirin! Once you have started down the road of Route Filtering, do not use Isolation either. Use one or the other, not both.

You will only require this functionality if you are "well-connected". What that means is that you are connected to several different parts of (say) the EU cluster and, at the same time, also connected to two or three places in the US which, in turn are connected back to the EU. This is called a "loop" and if you are seriously looped then you need filtering.

I should at this stage give a little bit of background on filters. All the filters in Spider work in basically the same way. You can either accept or reject various options in order to create the filter rules you wish to achieve. Some filters are user settable, others can only be altered by the sysop. Route filtering can only be done by the sysop.

Anyway, without further discouragement, let me start the process of explanation.

The node_default filter

All normal systems should have a default routing filter and it should usually be set to send only the normal, unlooped, view of your "national" network. Here in the UK that means nodes from the UK and Eire, in EU it is more complex as the networks there grew up in a more intertwined way.

The generic commands are:-


reject/route node_default <filter_option>

or

accept/route node_default <filter_option>


where filter_option is one of the following ...


call <prefixes>
call_dxcc <numbers>
call_itu <numbers>
call_zone <numbers>
channel <prefixes>
channel_dxcc <numbers>
channel_itu <numbers>
channel_zone <numbers>

Please be careful if you alter this setting, it will affect ALL your links! Remember, this is a default filter for node connections, not a per link default.

For the default routing filter then you have two real choices: either a "national" view or the "safe" option of only your own callsign. Examples of each (for my node: GB7DJK) are:-

acc/route node_default call_dxcc 61,38
acc/route node_default call gb7djk

GB7DJK uses the first of these. The DXCC countries can be obtained from the show/prefix command.

The example filters shown control output TO all your partner nodes unless they have a specific filter applied to them (see next section).

It is also possible to control the incoming routing information that you are prepared to accept FROM your partner nodes. The reason this is necessary is to make sure that stuff like mail, pings and similar commands a) go down the correct links and b) don't loop around excessively. Again using GB7DJK as an example a typical default input filter would be something like:

rej/route node_default input call_dxcc 61,38 and not channel_dxcc 61,38

What this does is accept node and user information for our national network from nodes that are in our national network, but rejects such information from anyone else. Although it doesn't explicitly say so, by implication, any other node information (not from the UK and Eire) is accepted.

As I imagine it will take a little while to get one's head around all of this you can study the effect of any rules that you try by watching the debug output after having done:-

set/debug filter

After you have got tired of that, to put it back the way it was:-

unset/debug filter


General route filtering

Exactly the same rules apply for general route filtering. You would use either an accept filter or a reject filter like this ...

reject/route <node_call> <filter_option>

or

accept/route <node_call> <filter_option>

Here are some examples of route filters ...

rej/route gb7djk call_dxcc 61,38 (send everything except UK+EIRE nodes)
rej/route all                    (equiv to [very] restricted mode)
acc/route gb7djk call_dxcc 61,38 (send only UK+EIRE nodes)
acc/route gb7djk call gb7djk     (equiv to SET/ISOLATE)

In practice you will either be opening the default filter out for a partner by defining a specific filter for that callsign:-

acc/route gb7baa all
acc/route gb7baa input all

or restricting it quite a lot, in fact making it very nearly like an isolated node, like this:-

acc/route pi4ehv-8 call gb7djk
rej/route pi4ehv-8 input call_dxcc 61,38

This last example takes everything except UK and Eire from PI4EHV-8 but only sends him my local configuration (just a PC19 for GB7DJK and PC16s for my local users).

It is possible to write much more complex rules, there are up to 10 accept/reject pairs per callsign per filter. For more information see the next section.

General filter rules

Upto v1.44 it was not possible for the user to set their own filters.
From v1.45 though that has all changed.  It is now possible to set
filters for just about anything you wish.  If you have just updated
from an older version of DXSpider you will need to update your new
filters.  You do not need to do anything with your old filters, they
will be renamed as you update.


There are 3 basic commands involved in setting and manipulating
filters.  These are �[4maccept�[24m, �[4mreject�[24m and �[4mclear�[24m.  First we will look
generally at filtering. There are a number of things you can filter in
the DXSpider system. They all use the same general mechanism.


In general terms you can create a "reject" or an "accept" filter which
can have up to 10 lines in it. You do this using, for example ...


     accept/spots .....
     reject/spots .....


where ..... are the specific commands for that type of filter. There
are filters for spots, wwv, announce, wcy and (for sysops) connects.
See each different accept or reject command reference for more
details.
There is also a command to clear out one or more lines in a filter.
They are ...


     clear/spots 1
     clear/spots all


There is clear/xxxx command for each type of filter.


and you can check that your filters have worked by the command ...


     show/filter


For now we are going to use spots for the examples, but you can apply
the same principles to all types of filter.


�[1m1.6.  Types of filter�[0m
There are two main types of filter, �[4maccept�[24m or �[4mreject�[24m.  You can use
either to achieve the result you want dependent on your own preference
and which is more simple to do.  It is pointless writing 8 lines of
reject filters when 1 accept filter would do the same thing!  Each
filter has 10 lines (of any length) which are tried in order.  If a
line matches then the action you have specified is taken (ie reject
means ignore it and accept means take it)


If you specify reject filters, then any lines that arrive that match
the filter will be dumped but all else will be accepted.  If you use
an accept filter, then ONLY the lines in the filter will be accepted
and all else will be dumped.  For example if you have a single line
�[4maccept�[24m filter ...


     accept/spots on vhf and (by_zone 14,15,16 or call_zone 14,15,16)


then you will �[4mONLY�[24m get VHF spots �[4mfrom�[24m or �[4mto�[24m CQ zones 14, 15 and 16.


If you set a reject filter like this ...


     reject/spots on hf/cw


Then you will get everything �[4mEXCEPT�[24m HF CW spots.  You could make this
single filter even more flexible.  For example, if you are interested
in IOTA and will work it even on CW even though normally you are not
interested in CW, then you could say ...


     reject/spots on hf/cw and not info iota


But in that case you might only be interested in iota and say:-


     accept/spots not on hf/cw or info iota


which achieves exactly the same thing. You should choose one or the
other until you are comfortable with the way it works. You can mix
them if you wish (actually you can have an accept AND a reject on the
same line) but don't attempt this until you are sure you know what you
are doing!


You can arrange your filter lines into logical units, either for your
own understanding or simply convenience. Here is an example ...


     reject/spots 1 on hf/cw
     reject/spots 2 on 50000/1400000 not (by_zone 14,15,16 or call_zone 14,15,16)


What this does is to ignore all HF CW spots and also rejects any spots
on VHF which don't either originate or spot someone in Europe.


This is an example where you would use a line number (1 and 2 in this
case), if you leave the digit out, the system assumes '1'. Digits
'0'-'9' are available.  This make it easier to see just what filters
you have set.  It also makes it more simple to remove individual
filters, during a contest for example.


You will notice in the above example that the second line has
brackets.  Look at the line logically.  You can see there are 2
separate sections to it.  We are saying reject spots that are VHF or
above �[4mAPART�[24m from those in zones 14, 15 and 16 (either spotted there or
originated there).  If you did not have the brackets to separate the 2
sections, then Spider would read it logically from the front and see a
different expression entirely ...


     (on 50000/1400000 and by_zone 14,15,16) or call_zone 14,15,16


The simple way to remember this is, if you use OR - use brackets.
Whilst we are here CASE is not important. 'And BY_Zone' is just the
same as 'and by_zone'.
As mentioned earlier, setting several filters can be more flexible
than simply setting one complex one.  Doing it in this way means that
if you want to alter your filter you can just redefine or remove one
or more lines of it or one line. For example ...


     reject/spots 1 on hf/ssb


would redefine our earlier example, or


     clear/spots 1


To remove all the filter lines in the spot filter ...


     clear/spots all


�[1m1.7.  Filter options�[0m
You can filter in several different ways.  The options are listed in
the various helpfiles for accept, reject and filter.


�[1m1.8.  Default filters�[0m
Sometimes all that is needed is a general rule for node connects.
This can be done with a node_default filter.  This rule will always be
followed, even if the link is isolated, unless another filter is set
specifically.  Default rules can be set for nodes and users.  They can
be set for spots, announces, WWV and WCY.  They can also be used for
hops.  An example might look like this ...


     accept/spot node_default by_zone 14,15,16,20,33
     set/hops node_default spot 50


This filter is for spots only, you could set others for announce, WWV
and WCY.  This filter would work for ALL nodes unless a specific
filter is written to override it for a particular node.  You can also
set a user_default should you require.  It is important to note that
default filters should be considered to be "connected".  By this I
mean that should you override the default filter for spots, you need
to add a rule for the hops for spots also.


�[1m1.9.  Advanced filtering�[0m
Once you are happy with the results you get, you may like to
experiment.


The previous example that filters hf/cw spots and accepts vhf/uhf
spots from EU can be written with a mixed filter, for example ...


     rej/spot on hf/cw
     acc/spot on 0/30000
     acc/spot 2 on 50000/1400000 and (by_zone 14,15,16 or call_zone 14,15,16)


Note that the first filter has not been specified with a number.  This
will automatically be assumed to be number 1.  In this case, we have
said �[4mreject�[24m �[4mall�[24m �[4mHF�[24m �[4mspots�[24m �[4min�[24m �[4mthe�[24m �[4mCW�[24m �[4msection�[24m �[4mof�[24m �[4mthe�[24m �[4mbands�[24m �[4mbut�[24m �[4maccept�[24m �[4mall�[0m
�[4mothers�[24m �[4mat�[24m �[4mHF.�[24m  �[4mAlso�[24m �[4maccept�[24m �[4manything�[24m �[4min�[24m �[4mVHF�[24m �[4mand�[24m �[4mabove�[24m �[4mspotted�[24m �[4min�[24m �[4mor�[24m �[4mby�[0m
�[4moperators�[24m �[4min�[24m �[4mthe�[24m �[4mzones�[24m �[4m14,�[24m �[4m15�[24m �[4mand�[24m �[4m16�[24m.  Each filter slot actually has a
'reject' slot and an 'accept' slot. The reject slot is executed BEFORE
the accept slot.


It was mentioned earlier that after a reject test that doesn't match,
the default for following tests is 'accept', the reverse is true for
'accept'. In the example what happens is that the reject is executed
first, any non hf/cw spot is passed to the accept line, which lets
through everything else on HF.  The next filter line lets through just
VHF/UHF spots from EU.


�[1m1.10.  Basic hop control�[0m
In /spider/data you will find a file called hop_table.pl.  This is the
file that controls your hop count settings.  It has a set of default
hops on the various PC frames and also a set for each node you want to
alter the hops for.  You may be happy with the default settings of
course, but this powerful tool can help to protect and improve the
network.  The file will look something like this ...


     #
     # hop table construction
     #
     package DXProt;
     # default hopcount to use
     $def_hopcount = 5;
     # some variable hop counts based on message type
     %hopcount =
     (
      11 => 10,
      16 => 10,
      17 => 10,
      19 => 10,
      21 => 10,
     );


     # the per node hop control thingy


     %nodehops =
     (
      GB7ADX => {            11 => 8,
                             12 => 8,
                             16 => 8,
                             17 => 8,
                             19 => 8,
                             21 => 8,
                        },
      GB7UDX => {            11 => 8,
                             12 => 8,
                             16 => 8,
                             17 => 8,
                             19 => 8,
                             21 => 8,
                        },
      GB7BAA => {
                             11 => 5,
                             12 => 8,
                             16 => 8,
                             17 => 8,
                             19 => 8,
                             21 => 8,
                        },
     );


Each set of hops is contained within a pair of curly braces and
contains a series of PC frame types.  PC11 for example is a DX spot.
The figures here are not exhaustive but should give you a good idea of
how the file works.


SHould any of the nodecalls include an ssid, it is important to wrap
the whole call in single quotes, like this ...


      'DB0FHF-15' => {
                             11 => 5,
                             12 => 8,
                             16 => 8,
                             17 => 8,
                             19 => 8,
                             21 => 8,
                        },


If you do not do this, you will get errors and the file will not work
as expected.


You can alter this file at any time, including whilst the cluster is
running.  If you alter the file during runtime, the command �[4mload/hops�[0m
will bring your changes into effect.


�[1m1.11.  Hop Control on Specific Nodes�[0m
You can set a callsign specific hop count for any of the standard
filter options so:-


     set/hops gb7djk spot 4
     set/hops node_default route 10
     set/hops gb7baa wcy 5


all work on their specific area of the protocol.


The �[4mset/hops�[24m command overrides any hops that you have set otherwise.


You can show what hops have been set using the �[4mshow/hops�[24m command.


�[1m1.12.  Isolating networks�[0m
It is possible to isolate networks from each other on a "gateway" node
using the �[4mset/isolate�[24m �[4m<node_call>�[24m command.


The effect of this is to partition an isolated network completely from
another node connected to your node. Your node will appear on and
otherwise behave normally on every network to which you are connected,
but data from an isolated network will not cross onto any other
network or vice versa. However all the spot, announce and WWV traffic
and personal messages will still be handled locally (because you are a
real node on all connected networks), that is locally connected users
will appear on all networks and will be able to access and receive
information from all networks transparently.  All routed messages will
be sent as normal, so if a user on one network knows that you are a
gateway for another network, he can still still send a talk/announce
etc message via your node and it will be routed across.


If you use isolate on a node connection you will continue to receive
all information from the isolated partner, however you will not pass
any information back to the isolated node.  There are times when you
would like to forward only spots across a link (maybe during a contest
for example).  To do this, isolate the node in the normal way and use
an �[4macc/spot�[24m �[4m>call<�[24m �[4mall�[24m filter to override the isolate.


�[1m1.13.  A DXSpider Filtering Tutorial�[0m
There is now an excellent primer/tutorial on filtering written by Jim
Samuels, W3BG with an introduction by Dave Hawes N3RD that I strongly
suggest you read.  You can read it here The DXSpider User Filtering
Primer