third_party (ex-lib_3p) now ready for review.

[Jeremy Allison]

On Fri, Jul 25, 2014 at 10:23:05AM +1200, Andrew Bartlett wrote:
> 
> What I propose (having the tarball creation process strip these out)
> also achieves the same goal. 

No it doesn't, as it leaves the source code
still in our tree, only removed in the
release process (if that's what I think
you're suggesting).

I want this source code out of our tree.
I can't be clearer than that !

> The way I see it, if autobuild is still using it, if the build farm is
> still using it, then it isn't out of our tree - it might be in a
> different git repository, but we are still involved in work to ensure
> they build, specifically including Samba wscript files. 

This is a neccessary step in moving to
getting rid of maintenance of this
stuff (including the Samba wscript files
from the thirdparty libs) completely.

If we leave them in the current git tree,
we will be no better off than we are now.

> If we are still involved, then putting it in a distinct GIT repo only
> creates work, and means we have to keep two different git repositories
> in line.

Why is this difficult ? This is a straw
man argument IMHO.

> What problem does moving these to a distinct git repo, that we still
> have to maintain, solve?  

Unmaintained code out of tree, eventually to
be gone entirely (into a .tgz file that is
downloaded by the folks who need it).

> The arguments you make seem to be the arguments for just totally
> removing all these external libs, including the build rules and
> everything.  I see this as a distinct step, and one that still needs
> some support for the build farm.

Which I am happy to do. Ira is trying to be nice
by doing things in small steps. I'd personally just
remove all these external libs - build rules and
everything. But I'm not doing the work, so I defer
to the person who is taking this on.

> Finally, it seems we are putting far more effort into this than we ever
> put into maintaining these libraries in the first place.

Sure - that's because this is *important* to get
right. Just doing another crappy import of source
code we won't maintain and won't update again until
someone complains is the easy way to do things.

Doing things right takes more effort than doing it
wrong (again :-), and I appreciate the work Ira is
doing to get this right.