[e2e] double bland reviewing

[Jon Crowcroft]

so certain conferences have a tradition of double blind
submission/review process to minimse the effect of
"oh its by so and so so it must be {brilliant|rubbish}"

some people have been known to post TRs to this (and similar lists)
just before the submission date to achieve certain effecs
i) familiarity amongst possible subset of reviewer pool
ii) feedback about paper to improve it before submitting....

these are not necessarily bad things, except for the 
iii) "brand recognition effect"

how about we setup a parallel list to e2e, which is closed
member-only submission, but anonimyzed.

so then people could "safely" post drafts to get community feedback
without breaching etiquette

I have here (affecting a Tom Lehrer voice/accent) a modest example I
prepared earier of just such a possible submission - this is, of
course, not by me, but by my evil twin at Some Other Cambridge
University Computer Laboratory

---------------------------------------------------------------
Self Similarity in the Arrival Process of 
Network Systems Architectures considered Harmful

....for possible FDNA paper...continuing a long tradition of
spurious research destinated for the journal of irreproduceable
results (formally known as ACM CCR), 

the past has always been the best predictor of the future (at least
until last tuesday fortnight when i was knocked off a bike for the
first time in 47 years of riding around in various taxi-ridden
cities). 

SO for FDNA, it is clear that we need to trawl the Past Directions
in Network Architectures - as was discussed recently, we need to
think about complexity - 

how has network architecture complexity varied over the years - lets
draw a graph, as that will help us get the paper up to 8 pages

complexity
^
|
|   2     4       6       8 ?
|  ..     . .     . .     ?
| .  . 3 .   . 5 .   . 7 .
|.    . .     . .     . .
 _______________________>time

Notice that the y axis complexity measured as the exponent, on a log
scale, so the peaks of the curve are np complete, the mid points are 
quadratic, and the troughs are linear complexity - the x axis for
time is measured roughly in decades....and is of course a
cabbalistic plot

so the points i've annotated are an excercise for the reader
but let me give some Three Letter Acronyms for you to play with
ATM
GSM
HBN(hilltop beacon network)
MCT (morse code telegraph)
ASN (Ad-hoc sensor networks)
IP4
X25
POTs

and we can see that it is a pretty clear match...

so what is 8? well this is somethign ew should avoid - this argues
that there is No Future in FDNA for 10 years - we should skip
immediately ahead 20 years where at least we can get back to
networks that are quadratic (or adriatic) in complexity, or even 30
years where they will be O(C), for some constant C

A colleague pointed out we need to worry about C - some examples are ATM
where C is basically the entire output of the world economy for a
decade in terms of trees and paper for the ATM B-ISDN specifications
- this is linear complexity in architecture, but of course of no
practical value at all....

next?

jon