Re: [Lsr] Flooding Path Direction

[David Allan I <[e-mail]>]

Hi Gents:

The algorithm in draft-allan actually has the notion of upstream, downstream
and both upstream and downstream FT adjacencies. However as a generalized
form is still a WIP and has yet to demonstrate merit against any of the
other approaches on the table, I'd not be looking to suggest a specific
encoding. 

If at some point it is decided that different classes of FT adjacency are
required, simply using additional types that share the format of the
flooding path TLV would appear to be sufficient....(?)

Cheers
Dave

-----Original Message-----
From: Lsr <[e-mail]> On Behalf Of Peter Psenak
Sent: Thursday, April 4, 2019 6:19 AM
To: Jakob Heitz (jheitz) <[e-mail]>; [e-mail]
Cc: [e-mail]
Subject: Re: [Lsr] Flooding Path Direction

Jakob,

On 04/04/2019 10:55 , Jakob Heitz (jheitz) wrote:
> How is it impossible that A may flood to B, but B does not flood to A ?

every node in the area must be able to flood the data and they must reach
all other nodes.

Adjacency bringup involves database exchange which is bidirectional, so is
the flooding. Flooding in one direction often serves as an ACK from the
other direction.

> How does it follow that if A floods to B, then B must also flood to A ?

does not have to be, but the LS protocols have been designed for
bidirectional flooding. Making this unidirectional would require significant
changes to the flooding procedures. Not something we want to go into.

>
> Uni-directional flooding allows a more efficient flooding topology to be
built.
> Suppose you had to build a flooding topology with the rule: 2 incoming and
2 outgoing floods.

why 2 incoming and 2 outgoing? We want each node to do minimal flooding,
which is 2 connections to flooding topology, not 4.

regards,
Peter


> With bidirectional flooding, the incoming are the same as the 
> outgoing, so any incoming flood can only be propagated on one interface.
> With uni-directional, any incoming can go out two interfaces.
> To achieve 2 out with bidirectional, you need 3 in, 3 out, because the ins
and outs are the same.
>
> Regards,
> Jakob.
>
> -----Original Message-----
> From: Peter Psenak <[e-mail]>
> Sent: Thursday, April 4, 2019 12:28 AM
> To: Jakob Heitz (jheitz) <[e-mail]>; [e-mail]
> Cc: [e-mail]
> Subject: Re: [Lsr] Flooding Path Direction
>
> Jakob,
>
> given that there is a single flooding topology calculated for an area, 
> I don't see how this can be unidirectional, considering that flooding 
> topology is used to flood in any direction.
>
> thanks,
> Peter
>
>
> On 04/04/2019 08:26 , Jakob Heitz (jheitz) wrote:
>> I mean flooding in one direction only, not unidirectional forwarding.
>>
>>
>>
>> Regards,
>>
>> Jakob.
>>
>>
>>
>> *From:* Tony Li <[e-mail]> *On Behalf Of 
>> *[e-mail]
>> *Sent:* Wednesday, April 3, 2019 11:19 PM
>> *To:* Jakob Heitz (jheitz) <[e-mail]>
>> *Cc:* [e-mail]
>> *Subject:* Re: [Lsr] Flooding Path Direction
>>
>>
>>
>>
>>
>>     I think this is too restrictive.
>>
>>     We should not exclude algorithms that can build a flooding topology
>>     with unidirectional links.
>>
>>
>>
>>
>>
>> Well, right now, the protocol doesn't really support unidirectional 
>> links. At all.
>>
>>
>>
>> Tony
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Lsr mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/lsr
>>
>
> .
>

_______________________________________________
Lsr mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/lsr