Re: packet selection: open issues

[Maurizio Molina <molina@ccrle.nec.de>]

Tanja Zseby wrote:

>  Hi,
>
> here are the remaining open issues for the packet selection (PS) 
> draft. Please comment on them until December 5.
>
> PS-1: Configuration of non-uniform probabilistic and flow-state 
> sampling: should we specify some functions exemplarily ?
> Problem: For both methods theoretical proposals exist in research, but 
> they are not yet mature. Furthermore the paramters for the schemes 
> differ for different proposed methods. So we can not define parameters 
> for the general scheme. Only give exemplarily parameters for specific 
> schemes (e.g. sample and hold). But this may also lead  to the 
> interpretation that psamp prefers certain methods.
> Proposed Solution: Only define the method and leaf the specification 
> of explicite schemes (and parametes) up to vendors (e.g. as extension 
> of the info model)
>
> PS-2: Terminology: approximate selection
> Problem: Should hash based sampling be described as "Random sampling 
> emulation" or as "random sampling approximation" or is there a better 
> term for this ?
> Proposed Solution: find a better term
>
> PS-3: Operating time
> Problem: The start and stop time of the sampling process is currently 
> part of the Info Model. But the start of a new scheme could be also 
> configured by the application by simply installing a new scheme.
> Proposed solution: remove it from info model
>
> PS-4: Integration of hash-functions:
> Problem: Should some selected hash functions (from the hash-function 
> draft) become part of the PS draft ?
> Proposed Solution: select 2-3 functions and put description into PS draft 

My opinion is that the detailed description of some hash functions 
should be part of the PS draft, otherwise we risk to leave it to vague 
and do not clarify how coherent packet selection at multiple points can 
be achieved.
On the other hand, Andy expressed the concern during last meeting that 
there is no established consensus on the fact that the hash functions 
that we can put in there now ( those of 
http://www.ietf.org/internet-drafts/draft-niccolini-hash-descr-00.txt) 
are really the best ones for doing the job. And I share it, even being a 
co-author of the draft.
So, is there a way out of it, in which we can put a formal hash 
descriprion  now in the PS draft but leave it open to the insertion of 
other functions in the future?
Maurizio

> Regards
> Tanja



--
to unsubscribe send a message to psamp-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/psamp/>