Re: [ippm] Next version of the duplicate draft

On 2008-7-9, at 12:54, ext Henk Uijterwaal wrote:
> The text says that active measurement systems MUST NOT send packets  
> with
> identical information fields, in order to avoid that all packets are
> declared duplicates.   That is sufficient to use this metric in this
> case.  The metric will also work for passive methods where one is sure
> that each packet is different.  The metric won't work for cases where
> multiple identical packets are intentionally sent.
>
> In order to move forward:
>
> a) We accept this (and add text to make this clear to everybody).
>
> b) We don't accept this.  In that case, one will have to find  
> something
>   to distinguish two packets from one packet sent twice.
>
> My preference is for the former, but I like to hear other opinions.   
> I'm
> afraid that if we don't accept this and don't come up with good  
> suggestions
> to distinguish packets, we'll never end up with a duplication metric.

I'd personally prefer to come up with a definition of what a duplicate  
is that uses only IP header information, rather than having  
dependencies on higher layers. Source and destination IP address plus  
the IP ID field ought to do it? (For non-fragments, at least.) Also  
see http://tools.ietf.org/html/draft-touch-intarea-ipv4-unique-id

Lars

PS: This is a comment as an individual contributor. If there's WG  
consensus for (a), I'm OK with being on the rough side of it.
_______________________________________________
ippm mailing list
ippm@ietf.org
https://www.ietf.org/mailman/listinfo/ippm