RFC: ABI break

[Robert Millan]

On Thu, Dec 29, 2005 at 12:08:48PM +0100, Aurelien Jarno wrote:
> >I think we don't need more problems than what we have.  If ABI changes can 
> >be
> >handled with versioning, we could even diverge from upstream and keep 0.1
> >where they have 1, in case upstream wanted to start with 1.
> 
> Are you sure they want to start with 1?

I'm just guessing, but it's likely.  I think it's also the most reasonable,
since once the glibc port is in upstream, this won't be debian-specific anymore.
It'll potentialy support other distros and even homegrown systems, so I think
this stops us from breaking abi anymore.

> AFAIK, hurd have libc 0.3, and 
> support of hurd is merged (at least partly) in upstream.

Two differences:

  - Glibc+Hurd has a fundamentaly different dessign, with a lot more
    functionality being drawn to glibc library calls instead of Hurd RPCs
    then it would if there were monolithic syscalls.  This makes it more
    prone to ABI change (I remember libc0.2, and presumbaly there was 0.1
    too ;)).

  - GNU/Hurd is still an experimental system (especialy when you consider L4).

-- 
Robert Millan