(fwd) e$: PBS NewsHour, Path Dependency, IPSEC, Cyberdog, and the Melting of Mr.Bill.

-----BEGIN PGP SIGNED MESSAGE-----


- --- begin forwarded text

Sender: e$@thumper.vmeng.com
Reply-To: e$@thumper.vmeng.com
Mime-Version: 1.0
From: rah@shipwright.com (Robert Hettinga)
Date: Mon, 22 Jan 1996 10:37:28 -0500
Precedence: Bulk
To: Multiple recipients of <e$@thumper.vmeng.com>
Subject: e$: PBS NewsHour, Path Dependency, IPSEC, Cyberdog, and the
Melting of Mr. Bill.

I thought I'd crank this out in light of Friday's NewsHour segment about
Apple and the path-dependency of the microcomputer market.


Contrary to what PBS would have us believe, ;-), the concept of path
dependency in technology, and in economics, for that matter, is a proven
fallacy.

At the risk of sounding credentialist (*I'm* not an economist, either), the
NewsHour's Mr. Solman seems to be proof that journalism isn't economics, no
matter the journalist's academic credentials. Journalists have to get a
story out, and sometimes there's no story in the actual economics of a
situation. Certainly there's no story in the non-existance of path
dependency.


The most famous example of path dependency, the idea that our previous
economic choices doom us to repeat those choices forever, is the QWERTY
keyboard, which has been proven *not* to be significantly slower than the
Dvorak keyboard, its supposedly more efficient alternative. Dvorak, the
designer of the alternative keyboard, was also the same person who
conducted the "ergonomic" studies (spending our WWII tax dollars, I might
add) "proving" it's efficacy, and elevated his keyboard to the status of an
urban legend. Dvorak was at worst a fraud and at best deluded with his own
grandeur. Reviews of Dvorak's own data show some significant flaws in both
research methdology and data handling. In addition, several independent
studies since then have shown that randomly selected beginning typists,
starting out on one keyboard or the other, have *never* shown any
significant difference in typing speed.

The Betamax/VHS videocassete war, another example of path dependence, was
more one of Sony not having an open standard than anything else. Sony
played dog-in-the-manger with it's own technology, and consequently ended
up owning the most lucrative market on a profit-per-machine basis, the one
in television broadcasting. It also means they were leaving big money on
the table where the largest market was, in consumer electronics. Doesn't
this sound familiar to Mac fans?

Path dependence had nothing to do with it. Consumer Reports did comparisons
at the time showing only a marginal difference between Betamax and JVC, and
nowadays there is absolutely no percievable difference between the two.

I now challenge anyone (including, unfortunately, Apple's own
*psychologist*, quoted in the NewsHour piece) to *prove* path-dependence in
the current market "hegemony" of Microsoft on the desktop. It ain't so. The
reason that Microsoft has business computer market dominance today is *not*
because of it's original *perceived* incompatibility with legacy mainframe
equipment ("nobody ever got fired for buying IBM -- or Microsoft").
Technically, mainframe compatibility was a non-issue at the time. It
certainly wasn't for the first 5 years of the Mac's life, anyway. Apple
could have done something about the mainframe compatibility issue with
simple marketing communications if they had paid any attention to it at
all.

Apple's heart was never in the business market. First of all, for all their
lip service to business, they really weren't ever attracted to the idea of
building better word-processing and spreadsheet boxes, even if they did
have the best one, before Excel and Word moved to Windows, at the time of
the big ramp up of the business microcomputer market.  It showed in their
attitude to most business people. Outside consultants and mavericks were
always the heroes in Apple's commercials, and so outside consultants and
mavericks were attracted to the Mac as a computing platform, but large
businesses and conformists weren't.  When compatibility with mainframes
actually did become an issue, for the short time when people were
offloading their mainframe data onto LANs, Apple didn't want to be there
anyway. With the advent of LANs, Apple didn't build the technology to deal
with LANs head on on their own turf, large corporations. Apple built
peer-to-peer networks of collegial desktop machines. Unfortunately, they
never paid attention to the bandwidth or the multitasking premia necessary
for those networks to function properly from the high-volume user's point
of view, and, so, when someone downloaded a file from your machine, and you
printed something in the background at the same time, you suffered a
performance hit if you tried to do anything else. Your mouse jerked around
the screen, or your words wouldn't show up in a window as fast as you typed
them. With PC file and print servers, this was less of a problem, because
those two jobs were offloaded to a seperate machine, whose job it was to do
nothing but run a printer, or to serve files. Since everyone had to be
connected to these servers the local area network, or LAN, was born. On
Apple networks, every machine is potentially a server for everyone else,
and everyone is their own print server. Only after PC LANs became
ubiquitous did Apple ever build servers of their own. Again, their hearts
weren't in it, because they were more interested in the possiblities of
more distributed, collegial, peer-to-peer networks. Fortunately, the first
problem, network bandwidth, has been solved, because almost all Macs now
come with ethernet, while the second problem, preemptive multitasking on
faster processors, is being solved slowly. This is all very good for Apple,
because peer-to-peer architecture is where the world's going to go anyway.
The whole internet is a peer-to-peer, "geodesic" network, where each
machine is optimized for it's own particular function, be it serving, or
switching information. There is no central repository of anything. That has
been Apple's view of networks since day one.


If it's any consolation, we won't even need LANs to do business with,
anyway.  A couple months ago, I saw Netscape running in the bond trading
room of this country's largest institutional trustee bank, of all places.
In their case, Netscape beat Powerbuilder hands-down in a prototype
development shootout. The prototype *was* the production version.  Netscape
can do anything from secure outside-the-firewall SQL calls to actually
conducting cash commerce. Game over. By the way, Netscape is not special in
this regard at all. So can any other sufficiently secure browser server
combination. Either one, client or server, can be developed for a dime a
dozen even now. This is especially true when compound document
architectures come on-line, like Apple's Cyberdog, an internet
implementation of their OpenDoc software object technology.

The reason we won't need LANs is because the only real difference between a
LAN and the internet is a firewall for security, and the need for clients
to speak Novell's TCP/IP-incompatible proprietary network protocol.  With
internet-level encryption protocols like the IETF IPSEC standard, you won't
even need a firewall anymore.  The only people who can establish a server
session with *any* machine connected to the net will be those issuing the
digital signatures authorized to access that machine, no matter where those
people are physically. When that happens, networks will need to be as
public as possible, which means, of course, TCP/IP, and not Netware. It's
like Heinlein's old joke about space, "once you're in Earth orbit, you're
halfway to anywhere". So, once you've gotten *rid* of the firewall, you're
everywhere.  So much for the path dependence of the LAN market.

What happens to the information concentrated behind those firewalls  -- or
proprietary software markets, for that matter --  when, because of strong
cryptography, firewalls disappear? Remember what happened to those floating
globs of grease in the detergent commercial? Surfacted away into little
tiny bits. I can hear Bill Gates now: "I'm melting!, I'm melting!". Ding,
dong, Mr. Bill is dead... Game over. Now you see why he's fighting so hard
to be net-compatible all of the sudden.

In this "decade of the internet", the [user interface, platform, desktop,
LAN, whatever] is meat, and real life is on the net, to paraphrase William
Gibson.


For the time being, I have come to the conclusion that the Mac, at least as
long as Apple makes most of them, is the computer for the "best of us",
and, unfortunately, not "the rest of us". I've learned to live with that. I
no more worry about Apple's prospects than I do about Porsche's.  I expect
that Apple management, like Herr Doktor Porsche, is just waking up to the
fact that even though they designed the Volkswagen, they can't possibly
mass produce them efficiently at a decent enough profit to advance the
state of the art, which is really where their hearts have been all along.
Sooner or later, Apple will go back to cranking out 917s, to demonstrate
the power of the technology, 911s, for a more affordable version of that
power, and 928s, for those of us who only want to look the part. ;-).
Fortunately, there are lots of companies, like Power Computing, to produce
those Volkswagens for those of us who can't afford Porches, and "Macintosh"
won't mean just "Apple" anymore.


So, for developers, and for me, a fully-credentialled Mac Bigot and
camp-follower, the future for Apple means Cyberdog, because Cyberdog means
breaking down large "glops" of information and software "grease" and
surfacting them, fractally, into little bitty bits out into the net, where
*all machines*, not just dumb Java-terminals, can use them better. It also
means developing cryptographically strong internet-level security, so that
anyone can talk to any machine from anywhere if they have permission to do
so, and *nobody* with out permission can either get in or see what those
authorized people are doing, with a packet sniffer, or worse, with a
key-cracker. It means building into all network applications the ability to
do digital commerce. That is, the ability to handle digital bearer
certificates, like Digicash's ecash, and the ability to handle
micropayments, like the MicroMint protocol or it's successor technologies.
Imagine if your code could send you money in the mail, or if a router did
real-time load balancing by changing it's micropayment
price-per-thru-packet when traffic got too high or too low. The future of
the net's going to be a strange place, indeed.


Until that happens, I suppose Porsche parts is still a lucrative business,
as long as developers keep in mind what business they're really in.


Cheers,
Bob Hettinga

- --------------------------------------------------
The e$ lists are brought to you by:

Making Commerce Convenient (tm) - Oki Advanced Products - Marlboro, MA
Value-Checker(tm) smart card reader= http://www.oki.com/products/vc.html

Where people, networks and money come together: Consult Hyperion
http://www.hyperion.co.uk                    info@hyperion.co.uk

See your name here. Be a charter sponsor for e$pam, e$, and Ne$ws!
See http://thumper.vmeng.com/pub/rah/ or e-mail rah@shipwright.com
for details...
- -------------------------------------------------

- --- end forwarded text


-----BEGIN PGP SIGNATURE-----
Version: 2.6.2

iQCVAwUBMQO/sPgyLN8bw6ZVAQGuSgP/fkKrI6aTSmPIGOu+LOxRzO5Ptt7QZNxh
48+b7975jIfUMgovphKBWdWtO+jGMCyUWxUVqjVbN8nmwfLT1RZFckOdLK0iM4nD
Fgl5+s9yoI0OllHS+oOMcAIyuLIkzazUgtQojm8qBFGSGulW0Keq2dIRNsThGLrk
Kk7K3oGMrQs=
=71fv
-----END PGP SIGNATURE-----

-----------------
Robert Hettinga (rah@shipwright.com)
e$, 44 Farquhar Street, Boston, MA 02131 USA
"Reality is not optional." --Thomas Sowell
The NEW(!) e$ Home Page: http://thumper.vmeng.com/pub/rah/

rah@shipwright.com (Robert Hettinga) writes:
[interesting article about the future, which includes..]

>The reason we won't need LANs is because the only real difference between a
>LAN and the internet is a firewall for security, and the need for clients
>to speak Novell's TCP/IP-incompatible proprietary network protocol.  With
>internet-level encryption protocols like the IETF IPSEC standard, you won't
>even need a firewall anymore.  The only people who can establish a server
>session with *any* machine connected to the net will be those issuing the
>digital signatures authorized to access that machine, no matter where those
>people are physically. When that happens, networks will need to be as
>public as possible, which means, of course, TCP/IP, and not Netware.

I'm all for the end of ridiculous non-TCP/IP protocols, but does
anyone believe this point about encrypted IP traffic eliminating the
need for firewalls?

I guess I don't trust the ability for people to keep secrets secret.
Nothing like refusing to pass packets at all..

Nelson Minar writes:
> I'm all for the end of ridiculous non-TCP/IP protocols, but does
> anyone believe this point about encrypted IP traffic eliminating the
> need for firewalls?

There is division in the IETF community on this point.

Phil Karn (who I have the greatest respect for) thinks IPSEC means we
can get rid of the firewalls. I, for one, don't -- they are there
largely because people don't trust that their networking software is
free of security holes, and cryptography doesn't fix security holes
for the most part.

Perry

Perry writes...
> 
> can get rid of the firewalls. I, for one, don't -- they are there
> largely because people don't trust that their networking software is
> free of security holes, and cryptography doesn't fix security holes
> for the most part.

	Perhaps I'm nieve, but I've always understood that one of the
primary functions firewalls accomplish is insulating from most easy
attacks large numbers of random machines in an organization that may not
be all perfectly administered, 100% under control of competant security
wise users, and configured correctly for maximum security with all the
latest rev's of stuff.

	Seems unclear that IP level security and authentication will
totally eliminate the problems caused by buggy software and 
clueless or careless users, or overloaded security staffs who
don't have time to update everybody and check everything immediately
on networks with thousands of machines.

	Having one or two machines to keep secure instead of thousands
seems like a big win.

						Dave Emery

I tend to oscillate between the two positions; at the moment I think that 
firewalls are still needed with IPSEC.

Firewalls cannot be removed if

	1) You need to control outbound as well as inbound traffic
	2) There are still non IPSEC machines on the network.
	3) There are network services on IPSEC machines that do not 
	   understand IPSEC security, and which cannot be easily secured 
	   through IPSEC aware wrappers.

I can't see anyway to cope with the first problem- however the latter two 
are legacy headaches, which tend to clear up given time.

What I do see happening is more and more IPSEC machines moving out into
a quasi-DMZ as it becomes much easier to make ordinary machines secure 
enough to go over-the-top; however, it'll take more than just IPSEC to 
make this fool-proof enough to move everybody out there.

One worry I do have is that if such a machine is misconfigured it could 
cause more damage as that machine is trusted more because it's using 
IPSEC. 

Simon


(defun modexpt (x y n)  "computes (x^y) mod n"
  (cond ((= y 0) 1) 	((= y 1) (mod x n))
	((evenp y) (mod (expt (modexpt x (/ y 2) n) 2) n))
	(t (mod (* x (modexpt x (1- y) n)) n))))

IPsec will not change the role of firewalls.  It will change some
technical details about them.

Firewalls do a couple of things:

	Enforce a policy boundary between us & them.  Reduce the
number of systems to be 'well secured' (This is because really
securing a machine is tough, and often involves sacrifices of
useability.) Provide job security/ass covering (see also, satisfy
auditors.)

	The fact that some traffic passing through is encrypted will
not change any of this.  Only allowing traffic to people who provide a
signature is only useful for some things.  Besides, there will always
be shitty protocols, like NFS, yp, SMTP, etc that need a firewall to
protect them.  Legacy systems are with us forever.  (I was in a
meeting last Thursday where we discussed how to handle a Sun3 that
needs to be a router in a CIDR environment.  No option to upgrade this
box for complex reasons.  I bring it up to illustrate the persistance
of legacy systems.)

Nelson Minar wrote:
| rah@shipwright.com (Robert Hettinga) writes:
| [interesting article about the future, which includes..]
| 
| >The reason we won't need LANs is because the only real difference between a
| >LAN and the internet is a firewall for security, and the need for clients
| >to speak Novell's TCP/IP-incompatible proprietary network protocol.  With
| >internet-level encryption protocols like the IETF IPSEC standard, you won't
| >even need a firewall anymore.  The only people who can establish a server
| >session with *any* machine connected to the net will be those issuing the
| >digital signatures authorized to access that machine, no matter where those
| >people are physically. When that happens, networks will need to be as
| >public as possible, which means, of course, TCP/IP, and not Netware.
| 
| I'm all for the end of ridiculous non-TCP/IP protocols, but does
| anyone believe this point about encrypted IP traffic eliminating the
| need for firewalls?
-- 
"It is seldom that liberty of any kind is lost all at once."
					               -Hume

reply from attila:

	I agree there will be "universal" secrecy --there will always 
    someone who manages to decode one or two "signatures" including 
    handshakes, and spoofs them, after burying the sucker machine in
    response commands so it has a chance to grab the handshaking.

	a little group effort, a couple of fast machines to
    coordinate the attack, and rest just might be history.  seems
    to me both Netscape and the abominable creature from the 
    Pacific Northwest said they could not be broken....

	Personally, I think NSA has figured out how to break PGP --
    enough specialized DSPs and prime factoring tables on magneto-
    optical disks can go along way.  If you have traffic both ways,
    you have the hash as well.

	dropping Phil accomplished two basic things: a cheap give-
    away to look good in public; and, they avoided defending ITAR
    in court  --and the ninth circuit can be pretty cranky on the
    Bill of Rights  --they don't follow Washington's line too 
    well.




On Mon, 22 Jan 1996, Nelson Minar wrote:

> rah@shipwright.com (Robert Hettinga) writes:
> [interesting article about the future, which includes..]
> 
> >The reason we won't need LANs is because the only real difference between a
> >LAN and the internet is a firewall for security, and the need for clients
> >to speak Novell's TCP/IP-incompatible proprietary network protocol.  With
> >internet-level encryption protocols like the IETF IPSEC standard, you won't
> >even need a firewall anymore.  The only people who can establish a server
> >session with *any* machine connected to the net will be those issuing the
> >digital signatures authorized to access that machine, no matter where those
> >people are physically. When that happens, networks will need to be as
> >public as possible, which means, of course, TCP/IP, and not Netware.
> 
> I'm all for the end of ridiculous non-TCP/IP protocols, but does
> anyone believe this point about encrypted IP traffic eliminating the
> need for firewalls?
> 
> I guess I don't trust the ability for people to keep secrets secret.
> Nothing like refusing to pass packets at all..
> 

__________________________________________________________________________
    go not unto usenet for advice, for the inhabitants thereof will say:
      yes, and no, and maybe, and I don't know, and fuck-off.
_________________________________________________________________ attila__

    To be a ruler of men, you need at least 12 inches....
    There is no safety this side of the grave.  Never was; never will be.