On Nov 1, 2007, at 4:40 PM, Brian E Carpenter wrote:
Leslie asked for comments from uninvolved parties and I'm giving my
personal opinion that I would not find this work useful. If others
do, we should go charter it.
I think it will be useful, if it succeeds in rigorously defining
metrics for upper layer protocols, given their dependencies on
the metrics for lower layer protocols. The context is how to
write meaningful service level agreements for application level
services, which require well-defined and reproducible metrics.
For the services I work on in my day job (e.g. multicast/unicast real-
time video streaming) the metrics I am concerned with are either:
a) RTP/RTCP transport metrics, for which AVT has done a very good job
with and hence PMOL is redundant, or
b) session protocol setup metrics (e.g. RTSP), which are mostly useful
for debugging as opposed to service measurement
c) application-level SLA metrics, which are pretty abstract and not
amenable to direct computation (e.g. visible artifacts per hour), or
d) encoding/perceptual metrics, such as PSNR, R-factor, etc. which are
in generally defined on bodies other than the IETF and which have been
the subject of much controversy both in the industry and in the IETF
when people have tried to get them blessed.
As such, PMOL would be valuable to me if:
1. it succeeded in developing useful algorithms/methods for
correlating these metrics to lower level metrics like route
convergence time or congestive loss rates, or
2. it succeeded in developing application metrics which are less
susceptible to industry controversy, vendor wars, service-provider
"count everything that moves" behavior, and found wide adoption.
My personal assessment is:
a) the likelihood of PMOL succeeding at either of these is quite low,
but I'm willing to be pleasantly surprised, and
b) if PMOL starts encroaching on the work in AVT I will be unhappy as
I think AVT does a very good job without needing a new venue for
developing useful transport level metrics, at least for real time flows
In summary, my general level of interest in PMOL is low, but that's a
moot point since the WG has been chartered and I wish the group the
best of luck!
Dave Oran (IAB hat off).
Brian
_______________________________________________
PMOL mailing list
PMOL@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/pmol
_______________________________________________
Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf