Re: [epochs] Proposal for an epoch-based organization of Boost libraries

From: Kostas Savvidis (kotika98_at_[hidden])
Date: 2020-05-24 08:53:38

> On May 23, 2020, at 12:56, Joaquin M López Muñoz via Boost <boost_at_[hidden]> wrote:
> 
> Hi,
> 
> Prompted by general feelings about Boost perceived lack of modernization and internal "bloat",

User perceptions about bloat may be related to the download size.
I was curious what in the world makes it weight 600MB and dig around a bit to get an idea.

1) 33MB of lookup tables as part of math special functions - and
numbers are trailing zero padded, such as SC_(7.1529668608000000000000000000000000000000e+10) - (why?)
These are only used for tests, not part of the library per se.

2) There appears to be some huge source files which must have been generated by another computer program of some sort, files such as
    a) epsg_traits.hpp      in geometry - 2MB
    b) make_map50.hpp  in fusion/container - 5MB 
    c) switch_50.hpp        in phoenix - 5MB

3) lots of binary files png, svg, bmp, pdf etc - e.g.
math documentation weights 50MB due to many svg and png

4) 80MB of docs which is not a problem, but some library has its own doc folder with many MB (duplicated?), often the largest subfolder by far is doc and tests also often far larger than actual code

5) ok so the source tree is huge, but maybe the installation is much smaller? Not really, include tree is 150MB and libs are 500MB

I have my suspicions about the wisdom of 2a,2b,2c, but to be clear I am not criticising the authors of these wonderful libraries --- only pointing out where the perceptions of bloat might be coming from. As a constructive proposal, maybe separating boost into three downloads - source,doc,tests - would help?

On another note, it is pretty clear that "eventually" boost will remove everything which is in STL now, but the time is not now. 
But the time is right for decreasing mutual dependencies within boost by replacing stuff with STL equivalents and potentially make more libraries available on standalone basis. Potentially, some people might prefer to keep releasing updates under 1.XX under traditional structure and jump to boost 2.0 for a new boost using STL whenever possible.

Best Regards,
Kostas

=========================================
Institute of Nuclear and Particle Physics
NCSR Demokritos
http://inspirehep.net/author/profile/K.G.Savvidy.1
https://github.com/kotika/random
https://mixmax.hepforge.org