On Mon, May 23, 2011 at 9:09 PM, Lars Viklund <zao@acc.umu.se> wrote:

On Mon, May 23, 2011 at 07:02:45PM +0200, Ovanes Markarian wrote:
> On Mon, May 23, 2011 at 1:52 PM, Lars Viklund <zao@acc.umu.se> wrote:
>

[...]

You read me wrong. I said to copy the boost/ directory as a whole.

Then of course, if you feel brave, you can always remove libraries that
you know it will not depend on.

Personally, I do not see the need to extract single libraries, as the
cost (hours/days of work, which may blow up) way outweighs the benefits
(what, saving a couple of megs of disk space?).

Lars,

It is not the matter of saving some megs. It can easily happen, that you need to implement some tooling where boost would be extremely helpful. But you are unable to reference some existing boost lib distro in the enterprise repository, because it is not there. It is also not possible to put boost libs within some short term into the common libs repository, since there might be some evaluation processes when a lib is allowed to be a part of common repository. 

On the other hand the tool to be developed is so small (a couple of files) that you still don't want to put into the tools repo the whole boost headers. What do you think how many header files are in the boost header directory? In the 1.45 there are over 7500 files. In that case it means, that your user will need to check-out over 7500 files to build some tiny tool. Do you really want the users of that tiny project to be required to check out with your 20 source files about 7500 files, and that only because you just wanted shared pointer or some other small lib form boost? 

It might be hard to understand but there are some enterprise policies in big companies and there is no way to just to overcome them.