[comp.protocols.misc] EGP3

jones@ingr.UUCP (Mark Jones) (09/16/88)

Is anyone out there doing any work with EGP3.  We have a copy of the
latest document, and I have heard rumors about publishing it, but the it
looks to me like it needs some major work.


One of the biggest problems is the sequence numbers, why are they so big
(32-bits), this is ridiculous!  The numbers are also sparse, which makes
for lots of overhead when you go to search your routing tables, you
can't use any form of subtraction or anything else, all you can do is
loop through the table and count them.  Also, what good does it do you
to even have them?


Another problem is the metrics, they are also to long.  32 bits is an
awful lot of bits to only use the values 1-255 and 0,1,2,255.  To quote
the document:

Metric Type 0			Hop Count: each gateway increments the
				hop count by 1
				>=255 indicates the Down State

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0   |N|         hopcount                                      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Metric Type 1			Reachability:
				0 primary
				1 secondary
				2 tertiary
				3-254 discard
				>= 255, no connection

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 1   |N|         reachability flag                             |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Metric Type > 1		Reserved for future use


Why are we wasting soooo many bits?  This looks like it was engineered
for the military( ie  definition of Elephant:  Mouse built to mil-specs).


The spec seems to almost require infinite memory, and instructions the
operate in zero time.


Where are the authors, if you have reasons for these seemingly mindless
acts, defend you spec.  I gave you the chance to respond via e-mail, now
you can tell everybody at once!



PS
Why are we using a minimum of 12 bytes for each entry?  Do you have some
pet machine you are trying to keep things on longword boundaries for?



Mark Jones