Re: [Starlink] SIGCOMM MIT paper: Starvation in e2e congestion control

[Hesham ElBakoury <helbakoury@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2148 bytes --]

Hi David,

I think someone such as Professor Hari who got many awards including the 
sigcomm 2021 life-achievement award or his student Venkat need to be 
educated on Fair Queuing. There are many publications and text books 
which describe FQ. The results of this paper is for network paths that 
do not use FQ or ECN. Venkat/Hari can provide more details.

For the CAP theorem, do you think I can get C,A,P, if this is what I 
need ? if this is the case, then this theorem is wrong or has limited 
applicability, correct ?

Thanks

Hesham

On 8/8/2022 2:39 PM, David P. Reed via Starlink wrote:
>
> Two things:
>
> 1) Do they understand Fair Queuing among flows? I actually doubt it or 
> they would have commented on it, to be intellectually honest.
>
> But maybe they will publish another paper that includes that? I get 
> really frustrated when reviewers don't force papers to address 
> well-known contradictory evidence. What are reviewers and referees who 
> are expert in the field for if not that? (my crankiness is justified, 
> I believe, by my commitment to research quality. But hey, maybe MIT 
> CSAIL faculty and SIGCOMM don't care about quality? Or maybe there's 
> been nothing published about FQ?
>
> 2) I absolutely hate folks who invent "theorems" that say you can have 
> "any two of three" properties. It's become popular in computer systems 
> research, but it actually creates a huge intellectual mess.
>
> The CAP theorem, for example has some very peculiar definitions in 
> order to make C, A, and P "independent" axes. Of course they are NOT 
> independent in engineering practice. In fact, they aren't even 
> "binary" - there's no "yes" or "no" to C, A or P - they are not even 
> spectra that map to some increasing sequence.
>
> Yes, you can't always get what you want. But you can almost always get 
> what you need, and that is never a specific two out of three.
>
> Especially not in queue management algorithms.
>
> Goddamn cutesy anti-intellectuals.
>
>
> _______________________________________________
> Starlink mailing list
> Starlink@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink

[-- Attachment #2: Type: text/html, Size: 4385 bytes --]