🚀 go-pugleaf

Up front I should mention that even if you have read standards.texi
recently, you should probably do so again.  Whatever standards and
compatibility outlines exist for the GNU project are documented there.
There are other "conventions" that are evolving.  Some of them even
compete.  When it becomes clear that one is emerging as a winner, it
tends to be documented in standards.texi.

In article <upui4INNjbv@network.ucsd.edu> mbk@lyapunov.ucsd.edu (Matt Kennel) writes:

   Re: GNU configurations.

   Instead of having many configuration files, why not have a single
   "GNU compatibility library & include files" so that everybody can
   write code to a single standard?  You have to configure the compat
   library once.

I think that there are a number of reasons why this has and hasn't
been done.

In a sense, it has been done.  Initially most GNU software was written
to the "BSD" compatibility library & include files.  :-).  More
recently, things have been moving toward ansi C and posix.

Then there's also gdb-4.5/libiberty, prep:~/pub/gnu/glibc, and a
handful of others.

The trouble is that there are a couple of competing trends going on.

First, GNU developers are spread across the world with only weak
coordination between them.  They swipe code from each other when
convenient, but they also think little of reimplementing pieces to
better fit their needs.  In general, this is very cool because it
means that things develop faster, with lots of room for
experimentation.

Second, GNU software is generally packaged in small, self contained
pieces.  This allows for much shorter periods between releases of any
particular program, easier ftp/uucp'ing, etc.  Imagine what would
happen if:

* you had to ftp/uucp/cp/whatever 6 other libraries before you could
  even build gdb

or

* all 6 libraries AND glibc were included in every release of every
  program on prep.

In contrast, how long was it between the last two X11 releases?  How
much work did someone go through just to get the source onsite?

In summary, the GNU project has a lot of very strong willed people
working on it.  The management and coordination is near non-existent,
which seems to be generally good for the development of GNU.  Code
*is* reused.  Formal interfaces are few and far between.

Someday, GNU software will be packaged as a complete system.  At that
point there will probably be common libraries and headers, but at that
point there will be considerably less need for "compatibility".

--
Disclaimer:  I *might* speak for Cygnus; subject to modification at
any time by nearly anyone.  I DO NOT speak for the FSF, although I try
to represent their views to the best of my knowledge.