[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: hurd does NOT need /hurd



On Sun, May 19, 2002 at 07:21:34PM -0700, Thomas Bushnell, BSG wrote:
> Adam Heath <doogie@debian.org> writes:
> 
> > On 19 May 2002, Thomas Bushnell, BSG wrote:
> > 
> > > The other is the RPC interfaces that the servers use to talk to each
> > > other.  These are basically static, and don't change except by
> > > extension.
> > 
> > You can't honestly expect this to always be the case.  No one can implement
> > everything perfectly the first time around.
> 
> Um, actually we can!  Because we have experience with how it works.
> 
> We can extend the interface at liberty without breaking compatibility
> at all.  That's just a property of the way the RPC interfaces work.
> Also, the names of interface functions are not part of the RPC
> interface itself, but only of the library ABI interface.  
> 
> As a result, a new RPC can be added without breaking the old RPCs at
> all.  Indeed, it can even take over the name of the old RPC.  The
> result is then that the library ABI changes--and it is versioned--but
> the RPC itself doesn't need a versioning.

And if I understand all this correctly, if a subsystem was designed
for some specific functionality and you can't easily extent that
functionality with adding RPCs, you can just make a new subsystem. A
server could then implement both subsystems for backwards compatiblity
and drop the old one when it isn't needed anymore.

Incompatibility isn't forced by nature.

Jeroen Dekkers
-- 
Jabber ID: jdekkers@jabber.org  IRC ID: jeroen@openprojects
GNU supporter - http://www.gnu.org

Attachment: pgp89QLkD1rdF.pgp
Description: PGP signature


Reply to: