Re: [boost] [http] Formal review of Boost.Http

Subject: Re: [boost] [http] Formal review of Boost.Http
From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2015-08-11 14:11:00

On 8 Aug 2015 at 14:06, Agustín K-ballo Bergé wrote:

> > The key thing Boost needed was a library development resource guide
> > as we had little which was up to date and especially relevant to C++
> > 11/14. I wrote my personal vision of that, as I was the one who
> > invested the 120 hours or so of my family time to write it and I
> > think that investment of my time gave me that right for the first
> > edition. If you, or anyone else being critical of my substantial
> > effort invested here, would like to write something better then go
> > ahead and do it instead of +1 attacking it when you haven't done
> > anything better yourself.
> 
> Why don't you submit it for the traditional review instead? After all, 
> having spent months (or years!) of your personal time on a library don't 
> automatically entitle you to make it a Boost library.

A traditional review doesn't suit shared documentation: everybody 
will have their own opinion, nobody wants to do the work to merge 
opinions, it all degrades into a bitching about other people's 
opinions match which ultimately is resolved by a single person 
ignoring everybody else and implementing something which nobody much 
likes, but it is what it is: a piece of shared infrastructure.

We've been though all this before. Dave went through this more than 
enough times. The people who maintain Boost.Build and Boost.Test and 
the web services see lots of vitriol spouted against their efforts 
and design choices here and elsewhere, and most of us here have felt 
frustration with the Boost infrastructure at one time or another. The 
tl;dr; of the whole thing is that Boost is not strong at shared 
infrastructure. We don't have the culture, processes nor resourcing 
to do shared infrastructure or shared decision making well.

I have proposed many ways of fixing that over the past three years, 
but to put it charitably there is not community consensus on what to 
do. The Steering Committee can not therefore allocate funding. We 
continue therefore to do nothing, and therefore nothing changes.

> I can't speak for others, but I don't think anyone claimed the Handbook 
> isn't useful. For me personally, the implications that this is a Boost 
> blessed document rubs me the wrong way, just as libraries that call 
> themselves Boost.Whatever without ever having been reviewed do.

The very first sentence makes it very clear it is not a Boost blessed 
document. Moreover, the target audience for that document is the tiny 
number of people wanting to author a Boost library. They tend to be a 
highly sophisticated audience who are subscribed to this list. It's 
not like any Boost library author is going to get misled about 
recommendations in that document being requirements to enter Boost.

And again, I repeat that that document is on the wiki, and is 
editable by anyone with a Trac login. That alone makes it not like a 
Boost library. It's the same as any other content on the Boost wiki - 
probably outdated but maybe useful information for Boost library 
authors and users and the Googlebot.

For the record I personally would far prefer a situation where Boost 
employed a full time permanent infrastructure and maintenance 
engineer whose full time day job is to improve and keep up to date 
all the shared Boost infrastructure. Individual people like me would 
still lead out the substantial effort in writing the first draft of 
some piece of shared documentation like that Handbook, but the full 
time maintenance engineer would de-Niall the shared document into 
something more generic and less personally written.

But without someone independent and acting not according to their own 
opinion and with enough free time to do that rewrite, we're stuck. 
I'm going to defend - hard - my opinions expressed in that document. 
You're going to attack the opinions you disagree with hard - and 
probably not budge me an inch. That's why you need an independent 
editor in between to do the rewrite according to their best guess as 
to consensus.

And we struggle to find review managers for maybe two dozen hours of 
managing a review, let alone the 80+ hours it would require to 
rewrite that handbook after multiple rounds of review and feedback to 
create a consensus document. And then what about someone to maintain 
the thing and keep it up to date after so much shared effort has been 
invested?

There is no point in proceeding with a shared document review until 
we solve the shared infrastructure maintenance problem first. It 
would be wasted effort for all. So we've got what we've got, and 
neither you nor I nor anybody is entirely happy with the outcome, 
same as for all shared Boost infrastructure.

Niall

-- 
ned Productions Limited Consulting
http://www.nedproductions.biz/ 
http://ie.linkedin.com/in/nialldouglas/