How would one quantify the following two statements:
Links that are slow/expensive/difficult to establish.
Links that are expensive/difficult to keep up.
One could take the position that these statements are true for links through an ad-hoc network by their very nature.
Regards,
Darrell Shane
-----Original Message-----
From: Mark Allman [SMTP:mallman@lerc.nasa.gov]
Sent: Tuesday, January 19, 1999 7:55 AM
To: pilc@lerc.nasa.gov
Cc: Vern Paxson; Aaron Falk; Spencer Dawkins
Subject: PILC: prioritization
We'd like to get a discussion going in an attempt to prioritize the
link characteristics discussed in the PILC BOF in Orlando. We'd
like to discuss which of the characteristics are important and which
the community feels are not so important. Also, we are interested
in attempting to determine which characteristics have well
understood mitigations that are generally agreed upon. The
following is a strawman classification of the characteristics.
Thanks,
allman
The following characteristics seem to be inherent in certain
channels. Therefore, mitigating some of the problems caused by
these characteristics should take place in the upper layers.
Significantly low bandwidth links.
Links with intermittent outages.
Links with asymmetric bandwidth.
Shared-channel broadcast to huge numbers of receivers.
Non-transitive reachability.
The following characteristics may be fixable at the link itself (at
least to a large degree). To the degree these are fixable at lower
levels, it seems desirable to do so. However, when the lower
layers do not fix the problem, how are the upper layers coping? Are
there ways to make the upper layers better cope with these problems?
Links with unusually high error rates (significant
non-congestive loss).
Small MTU links.
Links with inconsistent error rates.
The following characteristics may have an impact on the upper layer
protocols similar to cross-traffic (i.e., the delay and capacity is
dynamic, not static). Our initial proposal is that these
characteristics need not be addressed, because applications must
tolerate similar variations due to cross-traffic anyway. We'd be
interested in hearing opinions on this proposal.
Links with variable bandwidth-delay products.
Links with varying delay.
Links with link layer flow control.
Links with varying bandwidth.
The following two characteristics have been (or are being) handled
by the tcpsat WG and it is unclear that further effort needs to be
put into these.
Long delay links.
Links with high bandwidth-delay product.
Finally, there are a few items that seem like we should back
burner. However, if you think this is the wrong classification,
please let us know.
Links that reorder packets.
Unidirectional.
Links that are slow/expensive/difficult to establish.
Links that are expensive/difficult to keep up.
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 09:12:19 EST