Boost logo

Boost :

Subject: Re: [boost] Guidelines to implement Boost library evolution policy (was Boost 2.0)
From: Stephen Kelly (hello_at_[hidden])
Date: 2014-06-09 00:58:16

Niall Douglas wrote:

> On 8 Jun 2014 at 11:16, Stephen Kelly wrote:
>> >>The Jeff Garland case study tells us that the past problem is already
>> >>solved
>> >>using Boost from the present or the past. You don't need to solve that
>> >>problem again.
>> [snip]
>> And it then illustrates that 'older' means 1996 era compilers.
>> If you think it should mean something different, I recommend you edit the
>> document, or qualify what 'older' means.
> No, I think Jeff's use case does refer to that age of compiler. To my
> best understanding, he had a large code base based on ancient
> compilers which he successfully got compiling with C++ 11 mode
> switched on thanks to Boost. I understand he believes that a more
> rapid switch of Boost to requiring all C++ 11 would be a great loss
> to Boost and to those in his situation. I understand he therefore
> believes such ideas should be opposed.

It looks like the document about the future of Boost is making it a policy
that no compiler support should be dropped ever and that ancients are
supported by policy.

That definitely affects the code, and is an idea disconnected from reality -
some ancients were dropped globally already months ago and that has enabled
some of the modularization work ('What is this abstraction? Oh, it's
obsolete. We can remove an edge.'). Modularization is taking advantage of
that, and when the code is more modular, I'm sure people will want to bump
compiler requirements.

As I've said many times though, you don't even have to say 'c++11' in this
discussion - bumping compiler requirements would allow the use of partial
specialization in Boost (a feature of ancient compiler versions). What would
your paragraph look like if you removed all reference to 'c++11' and had
that in mind? Is the position of the document a reasonable position to hold?

It is not clear to me how updating the requirements of Boost master branch
could possibly be a great loss to his situation. That does not make older
Boost releases disappear. However, the document forwards the idea that it
would indeed be a great loss and must not be done.

> I do wish Robert would email this list instead of privately

He emailed the list and CC'd me as it was a reply to my mail. His mail did
not reach the mailing list. That's what I'm suggesting he resolve.



Boost list run by bdawes at, gregod at, cpdaniel at, john at