Re: [boost] [beast] Request for Discussion

Subject: Re: [boost] [beast] Request for Discussion
From: Rob Stewart (rstewart_at_[hidden])
Date: 2016-09-25 08:37:45

On September 23, 2016 5:17:08 PM EDT, Peter Dimov <lists_at_[hidden]> wrote:
> Vinnie Falco wrote:
> ...
> > This is the lowest level of functionality a library can offer with
> respect 
> > to HTTP. All higher level operations can be expressed in terms of
> these 
> > primitives. The corollary is all higher level operations are
> impossible 
> > without any of these operations.

[snip important points about expectations and focus]

> Plus, there's also the matter that to evaluate the lower-level API,
> you have 
> to have something built on top of it with which to evaluate it. 

I don't disagree with the idea of getting a robust, functional lower level API, but you can't know how appropriate that API is until you implement higher level functionality atop it. Consider the OSI model vs. TCP[1]. The layered approach was a good idea from a theoretical perspective, but it failed miserably for real world use cases. That and other experience led to the design decisions in Aeron[2], for example. Thus, your efforts to implement example clients, etc. are more than appropriate. I consider them essential proofs of Beast.

--
Rob
[1] https://tools.ietf.org/html/rfc3439#section-3
[2] https://github.com/real-logic/Aeron
(Sent from my portable computation device.)