[Bloat] [Ecn-sane] abc congestion control on time varying wireless links

Dave Taht dave at taht.net
Wed Dec 18 14:43:58 EST 2019


"David P. Reed" <dpreed at deepplum.com> writes:

> I should explain that my motivation in writing the previous review is twofold:
>
> 1) I think the authors are very capable of doing great and valuable work. And they have in the past done so.
>
> 2) Intellectual honesty, professional honesty, and rigor are values
> that seem to have been declining over the years, in favor of a
> pragmatic sense that anything that can get published is important. (I
> think it has been reinforced by the idea that counting publications in
> good journals is the only metric of scientific contribution that
> counts, even at MIT). To me, this is a VERY important value. I learned
> that from my mentors, Jerry Saltzer and F.J. Corbato, who were
> relentless in making sure publications were worthy.


I think the core-est problem is the major universities cannot get source
licenses to the 5g and wifi stuff at the lowest level. Nor can they get
sufficient hardware IP to make a meaningful contribution anymore.

As a result, the next generation of engineers and scientists have to
make do with interpreting elephant entrails, and building cargo cult
papers for submission to cargo cult journals with cargo cult reviewers
who've not seen real research in the professional lifetime.

I can't help but note that several of the core-est papers in the
bufferbloat effort, couldn't be published - and reviewers still
currently poo-poo packet scheduling ideas (like pacing, fq-ing) in favor
of aqm idea of the month.

I was kind of hopeful, in that Huewei as one of their remediation ideas
for making their 5G stuff more trustable was a promise to make code
available to all. THAT would be breath of fresh air and level the
playing field once again so valid research could resume.

And for all I know, they are doing that... for chinese universities.

About the only thing going on in the computing world that's encouraging
is the rapidly accellerating work on the risc-v.

>
> _______________________________________________
> Bloat mailing list
> Bloat at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat



More information about the Bloat mailing list