What we need next

[Michael Welzl < >]

Hi,

First of all, can we agree to be all on Skype today?  At least Francesco and 
Angelo. I’m online now in Skype, with username mwelzl.
My phone number:  Norwegian +47 47346201  (this is where I use WhatsApp), 
Italian (ancient phone but cheaper to call :-) ): 3 756227384


I’ll mainly work until 4PM and MIGHT have to leave then, depending on the 
state of things.


Looking at the current version of the paper, I think we need to do a few key 
things next:



ALL STUDENTS:
- - - - - - - - - - - - -

From you, I think I need two or maybe three things:


1) the tables, in a compilable form:

To make this happen, I created a new main file called 
"main-for-tables-only.tex” - with this you can try to get it to compile.  I 
am wondering: can we merge them into one table?  Not only would this save 
space, it would also be more convincing, I think, that this is indeed a 
“complete” TCP implementation, and I could then make the text say “congestion 
control is in lines X to Y”.

A question to Angelo: yesterday, you said something about not implementing 
SACK and not implementing Spurious Loss Recovery - but that was just the SW 
implementation, right?  Are Spurious Loss Recovery and SACK parsing parts of 
the tables? If not, why not - could we include them, or would that make the 
tables much longer?   (if we do skip them I need to explain why)


2) the time-sequence plot from Angelo:

Here, we’re almost there. Things overlap nicely between FPGA and SW, and 
getting these to be 100% overlapping is almost impossible, a time-sequence 
plot is REALLY a “zoom” into the TCP behavior, the tiniest timing derivations 
may create the differences that we see now. So, in my opinion, seeing a small 
difference is even better than a 100% overlap because it’s more credible!  A 
100% overlap might seem too magical!  BUT, it is VERY important that we can 
explain why lines are not 100% overlapping, so I need you to figure out why - 
and I think the ns-3 version should really overlap one of the others as well, 
else that’s just strange.


3) The SYN-Proxy use case:

It’s nice that we have text, but I think we need more. How did you agree to 
show what you did, with an XFSM table? If so, please add this to the 
tables.tex file - as a SEPARATE table. So I’m thinking of TWO tables in this 
case: one for the whole XFSM, one for the SYN-Proxy.



FRANCESCO / ANGELO:
- - - - - - - - - - - - - - - - - - -

I think the SW part should be finished ASAP, both with text and plots (at 
most 2 !). This, to me, is perhaps the biggest “hole” in the paper right now!


I’ll stay away from editing the SW file, "52-impl-sw.tex” - so Francesco, 
this is completely in your hands now!
All other files, I’ll have a pass through and adjust a little  (except that 
we get updates in tables.tex, I hope!)


Cheers,
Michael