Re: Free system that could be "real Unix".

[Christopher Browne]

After takin a swig o' Arrakan spice grog, mike3 <mike4ty4@yahoo.com> belched 
out:
> On Oct 12, 10:13 am, Christopher Browne <cbbro...@acm.org> wrote:
>> Quoth mike3 <mike4...@yahoo.com>:
>>
>>
>>
>> > On Oct 10, 5:49 pm, John Hasler <j...@dhh.gt.org> wrote:
>> >> mike3 writes:
>> >> > Would it be possible to construct a Free (note the capital "F") 
>> >> > operating
>> >> > system that would be capable of meeting all the UNIX standards (POSIX,
>> >> > SUS, etc.) precisely enough to be able to be certified by The Open Group
>> >> > as UNIX(R)...
>>
>> >> Any major Linux distribution could qualify, but nobody cares any more.
>>
>> > You sure? I thought they would require modification. But I'm not
>> > sure just how extensive it would need to be. If it is not be a huge
>> > amount, perhaps, just perhaps, one might "drift" over the "sweet
>> > spot" but of course nothing would really happen since the creators
>> > would either a) not notice it or b) not have the money to actually
>> > pay The Open Group to certify it. And, UNIX is just a label really,
>> > so even if it is not certified to be legally branded as such that
>> > does not necessarily make it any worse in terms of quality,
>> > capability, etc., especially if it were to otherwise meet all the
>> > relevant standards.
>>
>> The *biggest* part would be the cost of having "whatever remains of
>> The Open Group" evaluate the system to validate that it conformed to
>> their requirements.
>>
>
> Well the thing I was asking more about was making a system
> that would be "real Unix" in the sense that it would be *capable
> of passing* such evaluations -- so it would be "real Unix" in all
> but the legal sense (ie. what you can market it as) (unless of
> course all the money needed to get certification was paid.).

I *think* that STREAMS would need to be added in order to pass the
official tests.  That might no longer be true, but I think it was, at
one time...

>> The biggest *change,* as far as I was last aware, that would be needed
>> would be to introduce the STREAMS abstraction, which both BSD and
>> Linux folk have generally eshewed.
>>
>
> What was so bad about it, anyway, that they didn't like it?

There was an implementation:
  <http://www.gcom.com/linux/lis/>

It was discussed back in 1997:
<http://herbie.ucs.indiana.edu/hypermail/linux/kernel/9705.1/0378.html>

A cogent response was this one from Larry McVoy:

-----------------------------------------------------------------
: Now that there is a GPL'ed STREAMS package available (albeit for
: 2.0.29) which is being actively maintained, isn't it time for STREAMS to
: be integrated into the base system? I seem to remember Linus objecting
: to this but can remember why - heck, if it is just a compile time
: option, or even better a loadable module why should Linux be deprived of
: such a nice framework?

While I don't want to start another flame war, please understand that
it is the widely held opinion of people who have used STREAMS that it
is not a nice framework. It is complicated and slow and there are no
really high performance networking stacks that use STREAMS. Sun tried
and has gone back to sockets. Even Ritchie, the streams (streams !=
STREAMS) inventor hates STREAMS.

Sockets is a fine architecture. Learn it. Use it. Live free and prosper.
-----------------------------------------------------------------
-- 
"cbbrowne","@","gmail.com"
http://linuxdatabases.info/info/nonrdbms.html
The IETF motto: "Rough consensus *and* working code."