[c-nsp] ISIS MTU Ignoration good/bad???

[Adam Vitkovsky]

I'm not sure when did Cisco drop the ball on this one though it's obvious that things gone wild since.  
For example there's a bug in ME3600 platforms where the CLNS MTU is computed out of the blue not using interface MTU.  
Even though this can be manually corrected with "clns mtu" ME will still magically deduct one byte when actually generating an IIH thus what's shown as mtu in "sh clns int" is not what's actually used to pad the IIHs sent to a neighbor. 
The bug has been reported as early as 12.2(52.1)EY though it's still present on 15.3.1.S2 a safe harbor version. 
Seems it was easier to start ignoring MTU silently. Also IOS XR ignores MTU during ISIS adjacency formation.     

adam
-----Original Message-----
From: Peter Rathlev [mailto:peter at rathlev.dk] 
Sent: Wednesday, June 26, 2013 9:04 PM
To: Adam Vitkovsky
Cc: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] ISIS MTU Ignoration good/bad???

On Wed, 2013-06-26 at 17:07 +0200, Adam Vitkovsky wrote:
> I have just recently noticed that newer ISIS implementations actually 
> ignore interface MTU discrepancies.
...
> I'd like to know the opinion of the community on this matter please. 

I'm very much in favour of the MTU comparison and no adjacencies with different MTUs. As long as the knobs exist to disable it when needed.

>From where is this missing?

--
Peter