Re: [Ecn-sane] draft tsvwg slides #1

Re: [Ecn-sane] draft tsvwg slides #1

[Rich Brown]

[-- Attachment #1: Type: text/plain, Size: 1394 bytes --]


> On Mar 25, 2019, at 3:34 AM, ecn-sane-request@lists.bufferbloat.net wrote:
> 
> Working on getting better plotable data
> 
> What do y'all think? https://lists.bufferbloat.net/pipermail/ecn-sane/attachments/20190325/2cbfe1a1/attachment.odp <https://lists.bufferbloat.net/pipermail/ecn-sane/attachments/20190325/2cbfe1a1/attachment.odp>
Here are a couple comments on the progression of the slides:

1: Intro - good
2: Blank - placeholder - you'll fix this
3: The Problem - good. Maybe say that *loss* only provides binary signal per RTT -> Oscillation -> Bang-bang control theory

Then... swap slides 4 & 5

4: Two Congestion Signals - major bullet points
	- Congestion should be an earlier signal than loss
	- Swap CE & SCE, noting CE gives coarse info
	- We propose SCE, to give high fidelity information

5: Rename heading - s/The Solution/The Implementation/

6: State Diagram - good
7: Experimental Plan - good

8: Plot - You need to have an explanation of the graph on the final slide. It's a bit hard to read the axes, but way more important is a clear description of what the heck we're looking at. It won't be as detailed as [Explaining RRUL Charts](https://www.bufferbloat.net/projects/bloat/wiki/RRUL_Chart_Explanation/) but someone reading (only) the slides should be able to understand what the plot shows.

9: Some kind of summary

Good luck!

Rich

[-- Attachment #2: Type: text/html, Size: 3725 bytes --]

Re: [Ecn-sane] draft tsvwg slides #1

[Jonathan Morton]

> On 25 Mar, 2019, at 1:27 pm, Rich Brown <richb.hanover@gmail.com> wrote:
> 
> Maybe say that *loss* only provides binary signal per RTT

That's also true of CE, because of the feedback mechanism which involves making ECE sticky for an RTT.  I'm sure that was part of making sure CE was equivalent to loss.

 - Jonathan Morton

Re: [Ecn-sane] draft tsvwg slides #1

[Rich Brown]

> On Mar 25, 2019, at 8:34 AM, Jonathan Morton <chromatix99@gmail.com> wrote:
> 
>> On 25 Mar, 2019, at 1:27 pm, Rich Brown <richb.hanover@gmail.com> wrote:
>> 
>> Maybe say that *loss* only provides binary signal per RTT
> 
> That's also true of CE, because of the feedback mechanism which involves making ECE sticky for an RTT.  I'm sure that was part of making sure CE was equivalent to loss.
> 
> - Jonathan Morton

OK. I haven't been following this discussion carefully, so am happy to be corrected. 

I would still swap slides 3 & 4 (introduce the new idea before telling people that you're taking over ECT(1))

The plot is (much) better now but could benefit from a narrative. Perhaps the chart could be smaller (since it's much more readable) with a legend on the right to describe what the colors show. (I know you'll say it in the presentation, but a reader won't know what this is about.)

Thanks again!

Rich

PS - And how many minutes do you have left before presenting? :-) Good luck, and have fun!

[Ecn-sane] Fwd: draft tsvwg slides #1

[Bruno George Moraes]

[-- Attachment #1.1: Type: text/plain, Size: 177 bytes --]

| - We propose SCE, to give high fidelity information

Would be nice to put a Pulse Density Modulation figure to be crystal clear
how high fidelity it is!!!

[image: image.png]

[-- Attachment #1.2: Type: text/html, Size: 601 bytes --]

[-- Attachment #2: image.png --]
[-- Type: image/png, Size: 2882 bytes --]

Re: [Ecn-sane] Fwd: draft tsvwg slides #1

[Dave Taht]

[-- Attachment #1.1: Type: text/plain, Size: 732 bytes --]

That's really nice. A bit late tho. Next time! Please tune into this - talk
should be *soon*.

https://congresshall1.conf.meetecho.com/q-meetecho/login.jsp?ietf=tsvwg

Please Tune into

On Mon, Mar 25, 2019 at 4:26 PM Bruno George Moraes <brunogm0@gmail.com>
wrote:

>
> | - We propose SCE, to give high fidelity information
>
> Would be nice to put a Pulse Density Modulation figure to be crystal clear
> how high fidelity it is!!!
>
> [image: image.png]
>
>
> _______________________________________________
> Ecn-sane mailing list
> Ecn-sane@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/ecn-sane
>


-- 

Dave Täht
CTO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-831-205-9740

[-- Attachment #1.2: Type: text/html, Size: 1945 bytes --]

[-- Attachment #2: image.png --]
[-- Type: image/png, Size: 2882 bytes --]

Re: [Ecn-sane] draft tsvwg slides #1

[Jeremy Harris]

On 25/03/2019 12:27, Rich Brown wrote:
> 3: The Problem - good. Maybe say that *loss* only provides binary signal per RTT -> Oscillation -> Bang-bang control theory

Why can't you divide amount-of-missing by (RTT * goodput), for
a proportional signal?  Only one sample of signal per RTT, but
better than binary.
-- 
Cheers,
  Jeremy