cvs commit: src/sys/compat/linux linux_socket.c

[juli mallett]

* Paul Richards <paul at originative.co.uk> [ Date: 2005-03-10 ]
	[ w.r.t. Re: cvs commit: src/sys/compat/linux linux_socket.c ]
> On Thu, Mar 10, 2005 at 11:16:44AM -0700, Scott Long wrote:
> > Paul Richards wrote:
> > >On Thu, Mar 10, 2005 at 10:44:09AM -0700, Scott Long wrote:
> > >
> > >>Paul Richards wrote:
> > >>
> > >>>On Thu, Mar 10, 2005 at 06:06:16PM +0100, Dag-Erling Sm?rgrav wrote:
> > >>>
> > >>>
> > >>>>Paul Richards <paul at originative.co.uk> writes:
> > >>>>
> > >>>>
> > >>>>>Imagine something like Photoshop being written on the most recent
> > >>>>>version of Mac OS X and finding that compatibility only worked
> > >>>>>forward.  That would mean that most users out there would have to
> > >>>>>upgrade their OS in order to use the most recent version of Photoshop!
> > >>>>
> > >>>>Yes, that is usually how it goes.
> > >>>
> > >>>
> > >>>I don't believe it does. Can anyone provide real world examples of
> > >>>this happening that we can consider?
> > >>>
> > >>
> > >>You know, I'm completely outraged that I can't use MSWord 2005 on my 
> > >>Windows 3.1 system!  I even installed the win32s library!  Don't those 
> > >>bozos at Microsoft care at all about forwards compatibility?
> > >
> > >
> > >We're talking minor releases across stable branches here. This
> > >issue is not whether 6.1 should run on 5.3 it's whether 5.4 should
> > >run on 5.3
> > >
> > >I would expect software developed on XP service pack 2 to be able
> > >to run on a fresh CD install of XP.
> > >
> > 
> > No that's not the case.  Go look at software retail boxes sometime.
> > They often say "Requires Windows FOO Service Pack BAR".  This was
> > especially true with NT 4.0 and its incredibly long lifespan.  I would 
> > fully expect that software that comes out in the next year to start 
> > requiring SP2 for XP.
> 
> What we're discussing here though is changes in the ABI.  If SP2
> fixes a bug in the OS that is essential to your software working
> then obviously it will require SP2.
> 
> Likewise, if a new feature gets added to XP in a later service pack
> and your product is written to make use of it then clearly you'll
> require the latest SP.
> 
> However, what we're talking about here is targetting the original
> version of XP but using a later version as the development platform.
> Obviously anyone sane will thouroughly test their product against
> both versions but they'd expect the same binary to work on both if
> it was written against what they expected to be a stable ABI.

You CANNOT say "if a new feature gets added then clearly" and then try
to say that you might just MAGICALLY make use of a new feature and be
shocked and AMAZED when it doesn't work on an older version.  This is
an OBSCURE FLAG.  That will not EAT YOUR CHILDREN if you do not use it.
IF SOMEONE CHOOSES TO USE IT they should check when support was added.
WE DO NOT REQUIRE THAT SOMEONE USE IT.  If they were to port from Linux
and use MSG_NOSIGNAL, ignorant that it ever wasn't there, then it would
be THEIR PROBLEM if they claimed their code worked fine on versions
without it, without testing that this was the case.