Re: OSPF : Summarization of type-7 lsa's

"Pat Murphy - (650)329-4044" <pmurphy@omega7.wr.usgs.gov> Thu, 07 July 2005 14:47 UTC

MIME-version: 1.0
Content-type: TEXT/PLAIN; CHARSET="us-ascii"
Message-ID: <01LQC49ECAE2003LOZ@omega7.wr.usgs.gov>
Date: Thu, 07 Jul 2005 07:50:51 -0700
Reply-To: Mailing List <OSPF@PEACH.EASE.LSOFT.COM>
Sender: Mailing List <OSPF@PEACH.EASE.LSOFT.COM>
From: "Pat Murphy - (650)329-4044" <pmurphy@omega7.wr.usgs.gov>
Subject: Re: OSPF : Summarization of type-7 lsa's
To: OSPF@PEACH.EASE.LSOFT.COM
Precedence: list

Ajay,

>    1> If path type selected is external type 2 then type 5 metric 
>should be set to the largest type-7 metric subsumed by this net range + 
>1 ;

>****** I have doubt like.. why we add 1 in this cost ..Is there any 
>specific reason behind this?

The only hint comes from RFC 1587/3101 wording

     These metric and path type rules will avoid routing
     loops in the event that path type 1 and 2 are both
     used within the area.

I believe this was done in RFC 1587 to distinguish the Type-5 LSA 
originated via the Type-7 translation/summarization process from a 
Type-7 LSA that has the same prefix. However this is conjecture on my 
part and I don't ever remember verifying it in a discussion with Rob 
Coltun. A tight routing loop seems possible in this case when the tie 
breaker rules of RFC 1587's external route calculation Step 5 are 
applied:

    a. Any type 5 LSA.
    b. A type-7 LSA with the P-bit set and the forwarding
       address non-zero.
    c. Any other type-7 LSA.

An NSSA would need two ABRs and the translator would have to see an NSSA 
path to the Type-7 LSA's prefix through the second ABR for the loop to 
occur. 

The RFC 3101 tie breaker rules deal with this situation completely 
differently, and it is possible that adding +1 is no longer necessary. I 
realized this once, but adding +1 to the type 2 computed metric appeared 
harmless enough that a change seemed unwarranted. I do have a minor 
correction to make to RFC 3101 from an earlier disuccsion, so if the 
current WG feels this should be changed I have no objections.

>2> If the path type selected in external type 1 , the type -5 metric 
>should be set to the largest metric..

>***  In this case we don't add any cost to ASBR. why this is so?

The distance to the originator of the type-7 LSA plays a role in 
computing the translated Type 5's type 1 metric. Hence the second ABR 
above would compute a more preferred path via the Type-7 LSA that does 
not go through the translator ABR

Pat

Re: OSPF : Summarization of type-7 lsa's Pat Murphy - (650)329-4044
Re: OSPF : Summarization of type-7 lsa's tajay