RE: congrats!

Neal Cardwell (cardwell@cs.washington.edu)
Fri, 6 Nov 1998 00:44:55 -0800 (PST)

here are a few comments, mostly tiny nit-picky things that could all
dance on the head of a pin together:

o BTW, where does the estimate of 25M internet hosts come from? We could
use the estimate of 40M, from Huitema's page; by his methodology we passed
to 25M mark in Aug 97.

o I'm guessing the RAID paper is from 89, not 98. ;-) And the folks in
T.N. Team might appreciate us figuring out how to expand T and N. :^)
Anyone know how to do that? Does {} work?

o In section 2, the pair of phrases "we considered its round trip time to
be the average of the..., and its drop rate based on the number of these
samples..." is a bit awkward; a parallel structure might help.

o In section 2, we may want to be more careful with the phrase "such that
the round trip times or drop rates of AC + CB < AB", since drop rates do
not add quite like that.

o Section 3: it's a tad misleading to characterize this as a typical "Web
_page_ transfer"; it's just one of typically many objects in a page.

o Typo, section 3: "as our results in Section 2 shows,"

o In section 3, the nature of the example Internet path is a little
unclear. First it's characterized as a "completely unloaded cross country
link", but later on it is emphasized that losses occur, which implies some
sort of load.

o In figure 4, the vertical axis is mis-labled as latency; oughta be
bandwidth.

o In section 4.2: "six seconds in many implementations" is a bit
misleading. The spec says you're supposed to use 3 seconds, and Windows,
Linux, and FreeBSD comply.

o The paper does a good job of making clear that the overlay network is an
experimental mechanism until the last sentence, which muddles the issue a
bit again, leaving open the question of whether the virtual network is an
ultimate goal or just an experimental proving ground. Another possible
variation would have the flavor:

The University of Washington Detour project is attacking these issues by
deploying a virtual network as a testbed for investigating how to use
informed routing and transport to improve performance and availability in
the Internet.

in general, looks good! i'm afraid i have to say i like the new title
better, too... ;-)

neal

On Thu, 5 Nov 1998, Stefan Savage wrote:

> I haven't received any comments yet.
>
> The original paper is in http://www/homes/savage/detourorig.ps
>
> the latest version is the TR, and can be found in:
> http://www/homes/savage/papers/UW-CSE-98-10-05.PS
>
> - Stefan
>
> -----Original Message------
> From: tom@emigrant [mailto:tom@emigrant]
> Sent: Tuesday, November 03, 1998 12:18 PM
> To: syn@cs
> Subject: congrats!
>
>
>
> to everyone. These comments are with respect to the previous version.
> some of us (e.g., me) should try to make one last read of the latest
> version and get comments to Stefan by this Thursday? Friday? to give
> him time to take one last pass before the deadline.
>
> tom
>
>
>