PM's Netscape rant

["Vladimir Z. Nuri" <vznuri@netcom.com>]

I thoroughly enjoyed PM's vituperative, venemous, and vitriolic
Netscape rant. it just wouldn't be the cypherpunks list without 
the pit bull of the internet nipping at everyone's heels here
and slaying any unwarranted peace!! however I haven't seen any 
well-deserved rebuttals however, so..

I'm again going to be a netscape apologist and say, GEEZ, PM, 
will you take it easy, and untangle your underwear knots?

as has been pointed out numerous times, the encryption in the Netscape
code is designed to handle credit card number transport, *not*
actual cash transport. its really silly to have more security
present than is available than the weakest link. it would be like
worrying about a fence around the white house when one is giving
open tours to the public every day!!

PM is rather secretive about the systems he is working on, but
I suspect they are stock systems that must be highly secure because
they actually involve *transfer* of cash, and *large*amounts* of it,
in a *time-critical* environment, with *large corporate clients*.  these
are all inappropriate criteria to judge Netscape by.  in the Netscape 
scenario, the software is *not* transfering cash itself, *not* transferring 
large amounts of it, and *not* in a time critical application, and *not*
geared toward large corporations, but instead individual users. 
it is relying on another infrastructure (credit cards) for the actual 
transaction mechanisms. 

as has been pointed out numerous times, the whole
credit card apparatus is somewhat based on "security through obscurity", 
i.e. the obscurity of a credit card number, and it doesn't make a whole
lot of sense to try to make this more "secure". this is a problem for
credit card companies to fix (I agree it is a horrible problem, that costs
us billions, and should have been fixed a long time ago) .. but holding
credit card *using* companies responsible for this deficiency 
doesn't make sense. they are not the enemy!! they would surely seize the
most secure mechanism available, if there were alternatives.

the distinction is subtle, but I think a relevant one: is
the software itself transfering cash, or building on another system
that does so? hopefully in the latter case, the requirements for a
successful implementation are not so difficult to achieve (so that
even fresh-out-of-college CS students can call the functions to
do so, and perhaps code packages are written such that the user
is protected from their own naivete, or what PM would call stupidity
or incompetence).

--

PM gives some excellent techniques for improving code security (some
of these may not be exactly what he proposed):

1. hiring experts
2. code reviews
3. restrictions of who can work on what code (security clearance)
4. heavy testing
5. antagonistic attacks (i.e. hiring someone trying to crack the code that
others have written)
6. open review of key code

however, put all these things together and you get a company apparatus
a bit more like the NSA than a commercial company. I agree that all
these precaution are relevant for banking and stock transaction software
transferring millions of dollars.  but holding a joe-schmoe GUI and 
Web company responsible for this kind of paranoid oversight is really 
impossible and unrealistic and *unnecessary*.

there will be some companies that specialize in creating the 
*secure*infrastructure* for communications transactions. other companies
will just *latch on* to this existing infrastructure. hopefully the
requirements for *latching on* will not be too difficult, otherwise we
are all in trouble!! 
 
now, admittedly, it would be ideal if the netscape code was highly
secure, but again, I just don't think it is in the best interests of
this company to become security paranoid to the degrees that I have
listed above, and the extreme degrees that people here are ranting about. 
rather they should try to blend in with other companies
who specialize in cryptographic security. the latter companies should
as much as possible provide foolproof modules. they should take care
of all functions that have a potential for problem, such as random
number generation, key exchange, etc.  they should try to provide
a minimum of training where the code is not foolproof.

many have been making the point that one cannot judge the security of
a package based simply on analyzing key modules. I actually don't think this 
has been proven in general and completely resolved yet. I can imagine 
modules that communication with software such that the module 
itself is a "secure environment" in general, and
it is almost impossible to misuse the software itself. (for example,
the software might never store the actual keys of transactions itself, 
this being handled by a secure module, making it impossible to 
accidentally reveal them).

some day we might actually see "secure module support" built into a 
microprocessor. in many ways the microprocessors that guard against
illegal memory accesses and illegal function calls are in a sense
providing a kind of cryptographic security. and people who study
secure OSes generally eventually conclude that for ultimate security,
you almost have to work from the ground up, starting with memory,
microprocessors, and network hardware.

--

so my general point is that PM's rant, while lots of fun to read..

>you @#$%^&* whippersnappers!! you don't have a @#$%^&* clue about
>REAL code!! us old timers were writing code as secure and impenetrable
>as granite bricks, impregnable as a frigid victorian gandmother,
>before you were a twinkle in your mama's eye!! learn some
>sufficent grovelling skills for your superiors or you will
>not only be fired from your JOB but be excommunicated from the
>entire INDUSTRY, perhaps even tarred, feathered, drawn, quartered, and
>hung from your neck in the nearest tree!!! your employer will throw
>you to the wolves, your customers will spit on your flayed carcass,
>your family will look upon your shrivelled remains with shame,  the
>vultures will vomit your undigestable eviscerated entrails,
>and the world immediately explode, if you have a 
>SINGLE BUFFER OVERFLOW *anywhere* in your code!!

(ahem) this is not appropriate in the context of Netscape's aims, unless they
want to become financial transaction experts more in line with banking
expertise. netscape is more a "bring cyberspace to the masses" company,
not "bring secure transactions to cyberspace". it's just because so
few companies are successfully doing the latter, that netscape is forced
to implement some "key" aspects of it to support the former. but I suspect
they may ease out of the cryptographic security business in the long run,
delegating it to other companies' plug-in-packages.

furthermore, cyberspace is growing gradually. the way we get to really
incredible secure transactions is through a growing process, an evolution
in which mistakes are made at different levels, and which in the beginning
the software is not much more than a toy that looks pretty and has the
fewest moving parts and most simplistic design imaginable. 

I fully believe that some day a company in cyberspace will exist
that satisfies PM's and all other cypherpunk's most erotic dreams 
about secure transactions. however that day is years away and it
will take a long time to reach it. and I doubt that it will be the
same company that is playing around with GUI's for the end user and
hiring college programming hot-shots and Java geeks.  IMHO netscape 
is probably not going to be the company that will try to bring the
*lowlevel infrastructure* for cash, judging by the current winds,
although that could change. they will definitely help guide its
progress and be interacting with the companies that do, however.

when the big Secure Transactions Inc.  company is invented for
cyberspace, *then* the kinds of absolutely uncompromising  standards that PM
embodies will be in place. but again, we cannot expect the companies
of today to embody that ideology and atmosphere for a few years yet.

the cypherpunks play a very valuable role in finding these "growing
pain" mistakes of beginning companies such as Netscape,
but we are not really serving our own best interests or the 
harmonious growth of cyberspace by vilifying/ embarrassing/ 
browbeating/ humiliating companies or their employees over their security
problems, at least if they are clearly responsive to far less ammunition.
keep in mind that NSA unbreakable security is *just*not*appropriate* in
every situation, and in fact "weak" encryption does have legitimate 
uses (i.e. in a world where people routinely lock their keys in their
cars). (although I agree, in general one should always try to design a 
system to be as secure as possible.)

(oops, I used the term "we" in that paragraph, a grave cypherpunk sin.. 
my humble apologies; @#$%^&* cryptoanarchist vocabulary)
 
that all said, nevertheless, I do enjoy PM's periodic displays of 
undigestable bile eruptions at least as much as one of the other 
infamous amusing crackpots circulating in this corner of cyberspace..
(but geez, PM, were you raised by a pack of wild wolves or what?)

p.s. to TCM: why do you continually find my login name abbreviation so 
fascinating??? my apologies to anyone if I am missing some kind of 
inside joke here, I'm a little dense at times <g>