Re: [tcpm] draft-minshall-nagle

[Joe Touch <[e-mail]>]

Hi, Wes,

Although I don't recall this and only scanned it briefly, I don't quite 
understand the problem.

I thought it was widely known that Nagle was useful for single-character 
interactive traffic, but also widely known as something that should be 
disabled for any multi-character interactive traffic - including 
multi-character encodings, HTTP, etc.

If Nagle is off - as it should be for interactive web traffic - then the 
optimizations in this draft won't have any impact.

So what kind of traffic does this actually help? Or is this an 
optimization of a path that either isn't or shouldn't be used?

Joe

On 6/2/2014 6:58 PM, Wesley Eddy wrote:
> I was working on an issue with Nagle in a space system, and
> noticed that the Linux kernel actually implements the "Minshall"
> version of the Nagle algorithm, described nicely in:
> http://tools.ietf.org/html/draft-minshall-nagle-01
>
> There are a couple later papers I was able to find also that
> describe this.
>
> To me, this looks like clearly a good thing that should be
> recommended in the RFCs where we currently just have discussion
> of "standard Nagle".  Apparently it was adopted in Linux, but I
> wonder if it was picked up or not in other OSes as well.
>
> Does anyone have history on why this draft never progressed?
> Maybe someone has current contact information for Greg Minshall?
>