Re: Why Jam?

From: bill_kempf (williamkempf_at_[hidden])
Date: 2002-02-21 14:53:41

--- In boost_at_y..., "Jeff Garland" <jeff_at_c...> wrote:
> I agree 100% with Dan's approach.  We should encourage support for 
multiple
> build/install paradigms contributed by the community.  As much as 
Bill can't
> install Cygwin or whatever, working in a different corporate 
environment where
> Make is the norm, it can be difficult to switch people to JAM.  
Usually, they
> won't need to know about boost builds.  However,  if a 
configuration management
> team is involved in doing builds they might very well want control 
and they will
> likely be unhappy if they have to do something special for the 
boost libraries.

I definately agree with this (BTW, I'm not the only Boost member 
who's got problems with installation issues).  The only problem with 
the idea of having two efforts is in keeping the two in synch.  I 
can't (easily) deal with maintaining an autotools solution, but if I 
modify the library it's likely that a "third party" supported set of 
autotools scripts will now be broken.  Other then this headache, 
though, it would be wonderful for such contributions to occur.  In 
fact, for the Windows platform it would be great if some MSI packages 
were available, since most Windows developers are used to that 
paradigm for installation.

> And, there are still issues getting everyone using JAM.  I wrote a 
Makefile for
> a person trying to use the GDTL stuff the other day b/c he was 
having problems
> getting JAM to work under Cygwin.  It was faster for me to write 
the Makefile
> then try to debug the JAM issues.  I'm sure we could have asked for 
support on
> the boost list and got it right away, but that's not the point.  He 
wanted to
> experiment with the code, the build problem was a distraction.  The 
job was
> simple, it took me 10 minutes to write the Makefile and we were 
done b/c I know
> Make, he knows Make, and his environment supported Make.

I have no problems with this either, and is what should occur in a 
volunteer community like this.
 
> Also, Boost packages already support multiple build paradigms.  
John Maddox has
> provided support for various Make-based options from the 
beginning.  I don't
> know if he is going to keep supporting this, but since regex is a 
stable library
> I would think this option would continue to be available.

And this, too, is true.  But Boost should never make such a 
requirement necessary for contributors.  It will scare away too many 
people.

Bill Kempf