From: JoAnne Abbott <[EMAIL PROTECTED]>
> ----- Original Message -----
> From: "Caitlyn Máire Martin" <[EMAIL PROTECTED]>
> > You mean to say that TCP/IP (a networking protocol) is just a consensus?
> > Not hardly. I dare say protocols have *and must have* rigid definitions
> of
> > how things are done in order to work. They are not even necessarily
> > achieved by consensus. Plus, there is no consensus for TCP/IP. Just
ask
> > Novell, if you doubt me.
>
> Consensus does not mean total agreement.
=May= not mean total agreement. Many people do use the word to
say that everyone is in agreement. They may not all like exactly what
they got, but they agree it's the best they are going to get.
> The standards for the TCP/IP protocol are indeed rigid, as they must be,
to
> be
> usefull.
... to those people who wish to use TCP/IP. Protocols become kind of
a self-reinforcing thingy. You like what I provide, you use my protocol.
Someone else sees us using the protocol, they sign up as well. Sooner
or later it's a defacto standard. UUCP wasn't subject to a standards
review process, but there was a time when a huge amount of traffic
flowed over UUCP links and it ran on more than just UNIX boxen.
> TCP/IP is still a protocol and has become the de-facto standard of the
> internet
> but is not yet the "Official" standard.
Consensus != protocol or standard.
> > No, they aren't. Protocols are defined with no agreement whatsoever.
> > Standards are based on marketplace factors and occasionally consensus.
> > Sorry, but you've got your definitions absolutely backwards.
>
> The basis of market place standards is usually who got there firstest with
> the mostest. ie; Unix TCP/IP
Tragically this is often not how they work. Think of the 80's CPU wars.
Y'all make the point in a second ...
> > For example, there was no agreement on what a LAN protocol should be.
> > Novell pushed IPX/SPX, the UNIX world had TCP/IP, IBM had multiple
> > proprietary protocols, Banyan had it's own, many peer-ro-peer networks
> used
> > to use ArcNet, Microsoft used to push NetBEUI, and so on. There still
> isn't
> > consensus. The only reason Novell even bothered with IP support is
> because
> > their customers demanded it. Each company or community unilaterally
> defined
> > their own protocol. The marketplace then reduced the number in actual
> use,
> > but there are still several.
>
> Yes
> In the LAN marketplace the winner has yet to be decided but is almost
> certain.
> But if I don't use your LAN why should I use your protocol?
Now you're cheating. If it isin't "firstest with the mostest", it should
have
been SNA. But Caitlyn barely missed what I think is the correct way that
standards are set -- innovation and openness. Proprietary standards
aren't going to go anywhere because I can't join the club unless I'm you.
Dead standards aren't going to go anywhere because needs change
and I'll pick a standard (or force you to embrace, as with Novell and
their IP support) that meets those needs.
The Z80 was a much better processor than the 8080, which it mostly
replaced. Zilog didn't innovate (okay, the Z8000 doesn't count ...),
so Intel stole the market back with the 8088/86, then innovated to
keep Motorola and friends at bay. ISA was "worse" than MCA, but
MCA was closed and EISA was forced into existence.
Innovate or die. Develop accessible standards or be ignored.
> > I think it came from your original argument that standards prevent
> > innovation. They do not.
>
> The standards are preventing me from making a 128 bit direct connection
> to a video chip as a innovation. I don't have enough room on the worktable
> for Four motherboards.
Now you're confusing extensible standards with inextensible ones.
Consider #pragma or the options field in a TCP header.
> > > > > JoAnne Abbott C.E.T. MSEE, DSM
> >
> > Since we're throwing out credentials:
> >
> > Caitlyn M. Martin
> > Senior Network Engineer
> > IBM Global Services
Oh, no! Another IBMer! I'm starting to feel less and less embarassed
to work for Big Blue ;-)
> > > Even the best standards (inhibit, straitjacket, prevent) (choose one
or
> > add
> > > yours) innovation.
> >
> > No, they don't. As new innovation comes along, the standards change to
> > adapt to the new technologies. Take a look at chip sockets or PC
memory,
> if
> > you want good, recent examples.
>
> These are incremental changes in response to market conditions.
> I wouldn't want to put a Athlon in a standard PIII chip socket.
Uh, why not?
This is what I mean by "extensible". It also touches on proprietary
versus not. What if chip manufacturers standardized on the pinout
of x86-compatible CPUs?
What would the implications be for AMD and Intel if x86 compatible
CPUs were plug compatible?
> > > JoAnne Abbott C.E.T.
> >
> > I can put letters after my name too. I choose not too. The only reason
I
> > could see to do so on a list like this is an attempt to browbeat people
> into
> > accepting your views.
>
> Your choice. This one is of a personal pride in a accomplishment
> and is only intended for that.
A lot of us could put plenty of letter behind our name. I choose not
to because my comments rise and fall on _their_ merit, not on my
credentials.
I'm just a programmer. If you want to be impressed, I'll send you
my resume. But I'd rather think that my analysis of how standards
and protocols and innovation and so on work is right or wrong without
me having to describe myself as a Software Goddess.
-- Julie.
Software Goddess.
_______________________________________________
issues mailing list
[EMAIL PROTECTED]
http://www.linux.org.uk/mailman/listinfo/issues
