Re: [boost] [threads] making parts of Boost.Threads header-only

Subject: Re: [boost] [threads] making parts of Boost.Threads header-only
From: Andrey Semashev (andrey.semashev_at_[hidden])
Date: 2009-04-07 17:11:44

Emil Dotchevski wrote:

>>> I am against such a move. Boost Threads requires linking for other
>>> features which makes it one of the few libraries in Boost that can be
>>> properly designed to avoid unnecessary physical coupling. Unless
>>> something is proven to cause performance problems it should not be
>>> inlined, regardless of how simple it is.
>> I don't see your point. AFAIK, mutexes do not require any features that have
>> to reside in a separately compiled library.
> 
> In principle, nothing ever has to reside in a separately compiled library.

That's an exaggeration, just as having everything separately compiled.

> You don't see my point because you think of the header-only approach
> as a good thing,

Not exactly. There are pros and cons, and in this case I find pros more 
valuable. And there are other header-only Boost libraries that I would 
rather have separately compiled.

> whereas in my mind headers should be limited to
> things that must be in a header, such as functions for which inlining
> is critical, as well as template definitions. This reduces physical
> coupling, which is only a problem in large scale projects;
> unfortunately by the time it becomes a problem it is already too late.

I think, this particular case doesn't increase coupling in any 
considerable way. It won't introduce any header includes, classes or new 
functions, AFAICT. Besides, the need to link to the library is a way of 
coupling, too, so we'll remove it with the change.