[HECnet] Circuit costs (again!)

[Paul_Koning at Dell.com]

> On Jan 25, 2016, at 12:33 PM, Johnny Billquist <bqt at softjar.se> wrote:
> 
> On 2016-01-25 18:17, Robert Armstrong wrote:
>>>  I will participate.
>> 
>>   Oh, and sure - I'll go along provided I can get the routing results that I want.  Sorry to be selfish, but I'm not going to put up with some nonsensical routing path just so that we can have standardized costs.
>> 
>>   I already put forth my requirements for the routing a while ago - there are only two.
>> 
>>   a) If two nodes have a direct point to point link, then that link should be the preferred route.
>>   b) I really want symmetrical routing.
>> 
>> I already know that there are those who believe that symmetrical routing is impossible to achieve, so maybe I'll give up on (b).  I still want (a) though.
> 
> Symmetrical routing will be impossible. DECnet is not designed in a way that makes it possible.

Correct.

More precisely, symmetric routing is not a design goal of DECnet.  (Nor is it in other networks I can think of.)

If you set link costs the same at both endpoints, then routes will be symmetric if there are no equal cost paths between the given source and destination.  If equal cost paths exist, the outcome will depend on the node addresses, and in general you won't get symmetric routes.

Nor does it matter.  There is no end user benefit to symmetric routes, and the debugging benefits, if any, are small at best.

	paul