[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [nm-wg] treatment of lost packets when measuring delay



One short reply and a point about the "Hierarchy" memo.

shalunov@internet2.edu said:
> What is externally observed loss, if not non-arrival before a timeout?

You point out the reason why loss -is not- an infinite delay! My impression is
that it is a sort of artifact that has a lot to do with network management, and
does not help in predicting the network delay. In addition there are
applications that need small/regular delays and tolerate high loss rates, and 
other
that require the opposite to work properly.

My point about the memo is that losses should be addressed separately, as in
the current version. I also found appropriate the presence of the "Loss
Patterns (Statistical Properties)" sub-section.

As for section "9-Delay Characteristics", the interesting discussion that is
taking place encourages the inclusion of a similar "Delay Patterns" subsection.

As for the representation of such "Patterns" (in both the Loss and Delay
section), one should take into account that a request may include parameters
(for instance, the gain of a SRTT defined in RFC793). Flexibility should be
mandatory, since there is still active research on the field.

Hope this is of help

Augusto

-- 

The "application/pgp-signature" attachment is my PGP signature.
If your mail client is unable to use it, you can safely ignore 
it.

Augusto Ciuffoletti
Dipartimento di Informatica
Università degli Studi di Pisa
Corso Buonarroti
I-56125, Pisa (ITALY)
http://www.di.unipi.it/~augusto

Attachment: pgp00001.pgp
Description: PGP signature