Re: [boost] The future and present of Boost

Subject: Re: [boost] The future and present of Boost
From: Robert Ramey (ramey_at_[hidden])
Date: 2018-10-24 19:58:19

On 10/24/18 11:12 AM, Deniz Bahadir via Boost wrote:

Answering as a library maintainer.

The library I maintain is C++03 compatible.  This is because that was 
the standard in existence when it written.

As new standards come into existence, sometimes a problem comes up. This 
is usually because of something like: the

a) library depends on undefined behavior. This library could not be 
written in such a way to avoid this.  This includes working around 
optimizer dropping functions in release mode.  These functions are 
necessary but are not explicitly named.

b) some undefined behavior either gets defined or becomes undefined in a 
different way and one or more of the tests fails. Or some new feature is 
implemented.  The current example is that visibility is now being supported.

c) I go back and make an improvement.  Currently I keep the improvement 
consistent with C++03.  it's not hard to do.

d) The library is still compatible with C++03, C++11, C++14 and likely 
C++17.

So what should I, as a maintainer be doing differently?

Of course if the library were be written today, it would save a lot of 
time and grief to just use the latest version. Number of lines of code 
would probably drop by 50% or more.  But the library is not being 
written today, so this question is not relevant.

Should the library have been designed differently so it would depend 
less on undefined behavior?  Maybe. but the library is in wide use and 
it was required to support wide functionality in order to pass the boost 
review process.  With the benefit of hind sight, a couple of features 
which complicated the design weren't necessary and not including those 
would make the library simpler and easier to maintain.  But there aren't 
very many of such features.  Pretty amazing given that the library was 
released 15 years ago.  Since all this is in the past and not that big a 
deal - this question is not relevant

Should the library be re-written to exploit the knowledge gained over 
the past 15 years and make it smaller and easier to maintain?  Maybe, 
but no one has volunteered lately.  I get a lot of suggestions though.

So - I just don't see any of this discussion as being at all relevant to 
the maintainer of an old library. Now if it were somehow mandated that 
libraries are prohibited from being useful with versions of C++ earlier 
than C++11?, that would effect me as a library maintainer.  I would 
likely have to insert one statement which fails to compile under earlier 
versions to fulfill this requirement.

Sooooooo - I honestly can't see any this discussion is relevant to the 
present and future of boost - the subject line of the thread.

Robert Ramey