beach@DDNUVAX.AF.MIL (darrel beach) (07/28/90)
Here's a query for the well informed GATEr on the internet. It has to
Mail Bridges and the IGP they use, and how they use EGP. First the drawing:
The addresses are of course bogus and I'm hoping the picture is worth a
thousand words because I'm not gonna describe it in detail:
+----------------------------------+
| |
Mailbridge | | Mailbridge
MB1 -------+ Milnet +------ MB2
26.10.0.100 | 26.0.0.0 | 26.2.0.2
| |
| |
Router G1 | | Router G2
to -------+ +------to 192.50.50.0
192.25.25.0 on | | on 26.1.0.50
26.1.0.25 +-----+-----------------------+----+
| |
| |
| |
Router R1 Router R2
Gateway to Gateway to
131.2.0.0 128.5.0.0 on
26.2.0.131 26.5.0.128
First let's assume the following:
R1 has only MB1 as an EGP neighbor
G1 has only MB1 as an EGP neighbor
R2 has only MB2 as an EGP neighbor
G2 has only MB2 as an EGP neighbor
MB1 and MB2 are using the IGP currently being used by the
Milnet Mailbridges.
The questions:
1. Is the following correct?
Its my understanding that MB1 does not provide the direct Milnet
address for EGP neighbors in its IGP exchange to MB2. This would
result in the following routing table for R1:
192.25.25.0 via 26.1.0.25 at 2 hops;
128.5.0.0 via 26.2.0.2 at 3 hops;
192.50.50.0 via 26.2.0.2 at 3 hops;
If my understanding of what's in the IGP packets is correct, then
the effect is obvious and leads to Mailbridge bashing when a direct
route is available.
2. The real question is yet to come and is moot if I'm wrong about 1.
Let's change the neighboring of R1 to use both MB1 and MB2. Let's
also say that R1 use a 10 minute poll and one mailbridge is polled
every 5 minutes, i.e. the polling is 180 degrees out of phase, wit
MB1 getting polled first. After the first update R1's routing
table would be as follows:
192.25.25.0 via 26.1.0.25 at 1 hops;
128.5.0.0 via 26.2.0.2 at 2 hops;
192.50.50.0 via 26.2.0.2 at 2 hops;
When the update from MB2 was received, it would include the
following infomation:
192.25.25.0 via 26.1.0.25 at 1 hops;
128.5.0.0 via 26.5.0.128 at 1 hops;
192.50.50.0 via 26.1.0.50 at 1 hops;
So, would R1 now overwrite the existing routes with the new ones,
or maintain both routes and use the lowest hopcount??
And the real question is what do various implementations do under
such conditions??? If the routes get overwritten, which seems
logical to me, then you have a table that changes with every
update. If you keep old routes when they're better than new ones,
you could never be sure the better route was still available. I
guess there are several philosophies for handling this situation,
and I'd really like to know which is implemented by whom.
It must be apparent that the real question is what I describe in 1
the real scoop, or real poop. I think it would be very dumb if
true, but I was actually told be someone (who shall now and always
remain nameless) I usually believe that 1 is the real scoop. I was
also told, however, that it would soon change such that the IGP
exchanges would carry the info necessary to preclude 1 and let
everyone have direct routes where possible.
As usual, any info is appreciated, including flames IF THEY INCLUDE USEFUL
INFO.
Darrel Beach
....still only an egg after all this time