1                       INFORMAL HEARING
 2    - - - - - - - - - - - - - - - - x
 3    In the matter of:               :
 4    APL9800007/Z066051/G006298      :
 5                                    :  PAGES 1 - 78
 6    - - - - - - - - - - - - - - - - x
 7
 8    Hearing held at the offices of Department of Commerce,
 9    14th and Constitution Avenue, N.W., Washington, D.C.,
10    20230, commencing at 10:30 a.m., Monday, January 25,
11    1999, before Moderator Anstruther Davidson.
12
13
14
15
16
17
18
19
20
21    Reported by:  ROBERT M. JAKUPCIAK, RPR
22
Page  2
 1    A P P E A R A N C E S   O F   C O U N S E L:
 2    FOR THE APPELLANT:
 3                BY:   LEE TIEN, ESQUIRE
 4                1452 Curtis Street
 5                Berkeley, Calfornia  94702
 6                (510) 525-0817
 7
 8    FOR THE EXPORT ADMINISTRATION:
 9                U.S. DEPARTMENT OF COMMERCE
10                BY:   HOYT H. ZIA, ESQUIRE
11                      THOMAS C. BARBOUR, ESQUIRE
12                14th & Constitution Avenue, N.W.
13                Washington, D.C.  20230
14                (202) 482-5301
15
16                     P  R  E  S  E  N  T
17                         Hugh Daniel
18                         John Gilmore
19                   Joanne M. Kelly, Esquire
20                        James A. Lewis
21
22
Page  3
 1                    P R O C E E D I N G S
 2               MR. DAVIDSON:  All right.  I would like to
 3    welcome everybody to an informal hearing of an appeal
 4    submitted by Mr. Lee Tien on behalf of Hugh Daniel.
 5    This is Appeal Number 980007 or Z, as in Zulu, 066051
 6    or G, as in golf, 006298.  It's an appeal under
 7    Section 756 of Export Administration Regulations and
 8    appeals the propriety of a June '98 letter from
 9    Mr. Jim Lewis revoking a prior commodity
10    classification of EAR 99 and issuing one of ECCN 5D, as
11    in delta, 002.
12               You have requested an informal hearing and
13    Mr. Reinsch, the Under Secretary, has granted the
14    request.  The primary purpose for the hearing is for
15    you and Mr. Daniel to present any additional matters
16    for consideration that are relevant to the matter.
17    Neither side is required to present its case here, so
18    that applies on both sides.  Anything that, for
19    example, was in your pleading or in your letter that
20    you don't talk about is still certainly relevant and
21    something that Mr. Reinsch would address.
22               My role is that of moderator.  I am not
Page  4
 1    making the decision.  Mr. Reinsch is making the
 2    decision.  I would suggest we go around the room --
 3    oh, and one final thing.  Most of us know this but we
 4    probably need to remember since especially in this
 5    building we don't do that this much.  We can only
 6    speak one at a time because the court reporter can
 7    only take down one person at a time.  So if you want
 8    to say something, sort of raise your hand, take a
 9    breath and we will try to get you recognized.
10               To my left is Jim Lewis, who is the
11    Director of the Office of Strategic Trade and Foreign
12    Policy.
13               MR. LEWIS:  How are you?
14               MR. DAVIDSON:  To his left is Mr. Hoyt Zia,
15    who is the Chief Counsel for Export Administration and
16    is the primary legal advisor to the Under Secretary.
17               MR. ZIA:  Good morning.
18               MR. DAVIDSON:  And to his left is Mr. Tom
19    Barbour, who is an attorney in the Office of Chief
20    Counsel.  Good morning, sir.
21               MR. BARBOUR:  Good morning.
22               MR. DAVIDSON:  To my right is Mr. Lee Tien,
Page  5
 1    the attorney who filed the case.  And why don't I let
 2    you introduce your client and anybody else.
 3               MR. TIEN:  I represent the exporter in this
 4    case, Mr. Hugh Daniel, who is to my right, and to his
 5    right at the table is Mr. John Gilmore, who is the
 6    technical consultant.  And then sitting outside the
 7    table is Ms. Joanne Kelly of the McKenna & Cuneo law
 8    firm, and she will primarily be observing.  I will be
 9    expecting Mr. Gilmore and Mr. Daniel to offer
10    technical details that I am unable to offer.
11               MR. DAVIDSON:  Unless anyone has a question
12    or other matter, I would propose to allow you, let you
13    start and --
14               MR. TIEN:  Why don't you put your 12(c)
15    issue on the record.
16               MR. DAVIDSON:  The other issue is that we
17    would normally not allow in the room someone who is
18    not the exporter or a legal representative of the
19    exporter, since we would presumably be discussing the
20    commercial and other proprietary affairs of the
21    exporter and of his equipment.  Section 12(c) of the
22    Export Administration Act prohibits us from making
Page  6
 1    that public and we often speak of it as a right of the
 2    exporter, although it's probably broader than that.
 3               And as I said on the telephone, we have no
 4    objection to the other people being in the room, but
 5    we would never allow that if it were not acceptable to
 6    you and your client.  So I would like to indicate for
 7    the record whether it is or is not.
 8               MR. DANIEL:  They are here at our request.
 9               MR. DAVIDSON:  Thank you.  Go ahead.
10               MR. TIEN:  I would like to thank the
11    Under Secretary and Mr. Davidson for giving us an
12    opportunity to respond to Mr. Lewis's letter on the
13    BXA reclassification.  Let me begin by summarizing our
14    main point.  And they are two, with some subpoints.
15    The first main point is about the technology and
16    policy surrounding authentication.  As we all know,
17    the software in this matter is software for
18    authenticating the Internet, specifically domain name
19    security.
20               I will be talking about something called
21    DNS, and that refers to the domain name system, which
22    is kind of a traffic cop for information flowing over
Page  7
 1    the Net.  Now, the two subpoints under there are first
 2    that this kind of authentication requires some element
 3    of cryptography, and the regulations as they are
 4    currently written exempt authentication.
 5               The second, as a general policy matter this
 6    particular type of authentication, DNS security, is
 7    something that the government has a public policy
 8    interest in, as evidenced by the role of DARPA and
 9    DISA in funding and attempting to productize DNS
10    security research products.  And that as a technical
11    matter, and we can have Mr. Gilmore and Mr. Daniel go
12    into this in more depth, encryption is needed for the
13    particular authentication application here of DNS
14    security.  And so let me summarize that point by
15    saying that there are reasons for the authentication
16    exemption that currently exists.
17               My second point is more of a legal point
18    and I'm going to start at the same place, which is
19    that authentication is exempt from encryption item or
20    EI controls.  And the regulations are clear on this.
21    The regulations recognize that authentication contains
22    encryption and this is a matter of long standing.
Page  8
 1               Openly, at least since April 1992, when the
 2    ITAR recognized the authentication exemptions,
 3    these -- there has been this line between, in quotes,
 4    encryption and authentication.  And we believe that
 5    the key here in understanding authentication exemption
 6    is that it focuses on what the item is designed to do.
 7    And for short I sort of think of this as the "as is"
 8    principle.
 9               And for this reason, and I will go into
10    this in more detail later if necessary, our position
11    is that the initial classification of integrated
12    DNSSEC, the software in this case, was correct, it was
13    authentication software.
14               Now, the subpoint under this is that on the
15    second review of the software, which found it to be an
16    encryption item subject to EI controls, BXA applied a
17    different standard, did not apply the standard stated
18    in the regulations.  Rather, it focused on what the
19    software could be if it were modified by someone else.
20               This is, we argue, ultra-vires action.  The
21    standard needs to be in the regulations.  Second, it
22    undermines the authority of the written regulations
Page  9
 1    and the policy behind them, which recognizes that
 2    authentication requires encryption.  And this is again
 3    especially true for the DNS security situation.
 4               Third, the standard that's actually used in
 5    the -- on the second classification, which is that, I
 6    call it modifiability or easy modifiability, is
 7    unworkable and if it were left in the current form it
 8    would be vague.
 9               There is a general principle that we should
10    construe statutes and regulations so as not to render
11    any language in them superfluous.  And the problem
12    with this construction is that it would essentially
13    swallow up the authentication exemption.
14               And finally, we believe that there is a due
15    process problem.  Certainly, we do not say that BXA
16    lacks the authority to promulgate the standard, but it
17    should be done in a procedurally fair fashion.  You
18    need to have clear and consistent rules and
19    application of rules.
20               Agencies are bound by their written
21    regulations.  It would be good for exporters if the
22    agency followed due process, because under the EAR
Page 10
 1    self-classification is encouraged and exporters need
 2    clear rules in order to do that.  But as we know,
 3    there is strict liability for civil violation of EAR,
 4    and that will make it very difficult for
 5    self-classification to occur.  And on the other hand
 6    we think it would be good for the agency because what
 7    we have here is a situation where BXA is undermining
 8    its own written regulations and the legitimacy of the
 9    process is something that encourages compliance.  When
10    significant changes are going to be made in the
11    regulations, as we believe this would be, then doing
12    so ultimately will enhance the legitimacy.
13               And finally, whereas traditionally most of
14    these kinds of decisions are not subject to judicial
15    review, we believe that under the IEEPA there
16    is a potential for review of this kind of decision.
17               So let me summarize just by saying that
18    there is no such thing as a perfect rule, but today's
19    written regulations are pretty good.  They have been
20    essentially unchanged since 1992.  There is no reason
21    to depart from them.  I think that the standard
22    enunciated as a reclassification of the software works
Page 11
 1    a significant change and that should not occur without
 2    an open process of actually changing them.
 3               That concludes my summary.  I would like to
 4    ask either Mr. Gilmore or Mr. Daniel to talk a little
 5    bit about the technical aspects of why authentication
 6    is integral to domain name security.
 7               MR. DANIEL:  The Internet is growing and
 8    growing at quite a great pace and we are all coming to
 9    depend on it more and more.  There is a fundamental
10    mapping that happens out there between human named
11    space and computer number space, and this mapping is
12    done via the domain name system that converts names
13    that all of us use into the numbers, the addresses,
14    like 1909 Main Street, that the computers have to use
15    to communicate with each other.
16               Right now that entire system is open to
17    being messed with by people with appropriate skill.
18    Anyone who wants to can go grab a piece of the main
19    name space, and they don't have to do this obviously.
20    They can do it very subtly.  They can say simply get
21    the numbers from me and hand out perfectly valid
22    numbers for maybe all but one address and for that one
Page 12
 1    address it's not even clear to most users or even
 2    sometimes to experts like me that I'm getting a bogus
 3    number, because they can take all the packet data for
 4    that valid other end of the communication, look at it,
 5    maybe modify it, maybe not, and then pass it on to the
 6    other end, and the other end is none the wiser because
 7    it can fake my address to the other end.  This is
 8    called the man in the middle attack.
 9               Currently the entire Internet is open to
10    this and it's not something that can be fixed in any
11    one place.  The entire Net has to be fixed because
12    somebody sitting in Egypt could literally do this for
13    me in looking up regulations on the BXA site from
14    California to Washington.
15               The way this is fixed is with something
16    that the IETF is actually done designing and now we
17    are trying to deploy called domain name security,
18    where we use the art and techniques of authentication
19    to make certain that the name to number mappings are
20    what they claim to be.  So that when I'm working with
21    say deposits.citibank.com, I'm really getting
22    deposits.citibank.com directly and not say
Page 13
 1    deposits.medellin.cartel.
 2               This is the technology we are trying to get
 3    out and trying to get deployed.  I want my Net to be
 4    solid and secure and I want to use it just as easily
 5    to send love letters or to look up regulations or to
 6    do business with my government.  That is not going to
 7    be doable in the future if the domain name system
 8    stays as vulnerable as it is now.
 9               MR. DAVIDSON:  Excuse me.  Can I ask a
10    question at this point?
11               MR. DANIEL:  I was done.  I was trying to
12    make it quick.
13               MR. DAVIDSON:  Thank you.  You did.  Two
14    questions.  The vulnerability is in up-loading the
15    domain name information, the translation between the
16    domain name and, I guess, the IP address?
17               MR. DANIEL:  That's correct.
18               MR. DAVIDSON:  Or is the vulnerability once
19    they are up-loaded somewhere in the stream someone
20    misdirects, if I'm using the right word, and I may not
21    be, the request for data between the user and the
22    server?  That's the question, right, if I --
Page 14
 1               MR. DANIEL:  In actuality it's both.
 2               MR. DAVIDSON:  And is your software
 3    designed to address both?
 4               MR. DANIEL:  The second answer is yes and
 5    no.  The first answer is that you have to -- whenever
 6    a user types in an address, like
 7    deposits.citibank.com, that address has to be looked
 8    up and several computers along the chain are involved.
 9    The DNS itself is a distributed database system.  It's
10    actually rather unique in networking technology.
11    Pieces of it are kept all over the place, because if
12    you had to go to a central repository for that bit of
13    information you would have to wait long amounts of
14    time.  And whenever that central repository was down
15    or overloaded you would not get the data you were
16    looking for.
17               So the data in the DNS is distributed
18    around the many nodes of the hierarchical system so
19    there is both a storage in all the remote places where
20    pieces of the database are kept and there is the
21    fetching of the data.  Each one of those nodes uses
22    the DNS system to get the data from further up or down
Page 15
 1    the hierarchy.
 2               MR. TIEN:  Can we think of it as a big
 3    distributed phone directory?  Is that an unreasonable
 4    analogy?
 5               MR. DANIEL:  No.  No.  It's more like the
 6    neuro-mapping system in a human body.  Because no, a
 7    phone directory, if you get it wrong you call up and
 8    say, hi, is this Joe's Garage?  And they say, no, this
 9    is Joe's Cleaners.  This is more integral to the
10    operation of the Net than that.
11               MR. DAVIDSON:  Let me --
12               MR. TIEN:  I don't think we -- did we --
13               MR. DANIEL:  There was a second part of
14    your question I think and I'm spacing what it was at
15    this point.
16               MR. DAVIDSON:  I probably didn't ask the
17    question exactly right.  I guess I'm a little bit
18    interested -- I think I sort of understand the domain
19    name system.  I probably have some mistakes.  Where
20    does the authentication occur?  What is the packet of
21    data that you are authenticating?
22               MR. DANIEL:  Give John a shot.
Page 16
 1               MR. GILMORE:  Why don't I try this.  The
 2    way DNS security is designed, the person who
 3    originally puts the data into the system, who owns it,
 4    basically signs it with a private key known to them.
 5    And so, for example, BXA would hold its own data and
 6    sign that.  And then as that data propagates around
 7    the network that signature is checked at each point.
 8               So the software that we were actually
 9    trying to publish only checks it at intermediate
10    servers.  So, for example, at my machine, say my
11    laptop wouldn't check it but the server in my house
12    would check it.
13               Eventually the software will be set up so
14    that it gets checked all the way end to end but we
15    don't have all of that written yet.
16               MR. DAVIDSON:  So the major thing it
17    authenticates is that the IP address that correlates
18    to the bxa.doc.gov always stays -- and when that gets
19    replicated through the domain name system that they
20    stay the same?
21               MR. GILMORE:  Right.  Unmodified.
22               MR. DANIEL:  So the address and the name
Page 17
 1    stay the same and are checkable from whatever the
 2    owner entered it as.
 3               MR. TIEN:  If I try to pretend to be
 4    bxa.doc.gov, they will look and say, whoa, that's not
 5    signed with your key so we reject it.
 6               MR. DAVIDSON:  Or you try to up-load to the
 7    domain name server someplace that -- all right.  I
 8    understand.  Thank you.  If you were finished or if
 9    you have anything else?
10               MR. GILMORE:  I guess I had one other thing
11    to say.  People have known about this vulnerability on
12    the Internet for maybe ten years.  It's been known
13    theoretically, but it actually happened more than a
14    year ago.  There was someone who was upset with the
15    monopoly of Network Solutions over issuing domain
16    names, a guy names Eugene Kashpureff, who wanted to
17    start up a competing business and he was in Canada and
18    one day he sent the wrong information to about 5,000
19    domain name servers, sort of he picked a good 5,000
20    and sent bad information to each one of them that
21    redirected www.internic.net so it would come to his
22    web page instead.
Page 18
 1               This was all covered in the New York Times.
 2    He ended up being the subject of a criminal
 3    prosecution for it and, you know, it was a serious
 4    problem.  It woke a lot of people up to the
 5    vulnerability and increased the urgency of actually
 6    getting the domain name security deployed.
 7               MR. DAVIDSON:  Let me ask another question
 8    then along that line.  The software in issue here was
 9    in source code?
10               MR. TIEN:  That's correct.
11               MR. DAVIDSON:  That's one of the big
12    issues.  For the purpose of protecting the Internet
13    and addressing the concerns you raised, what would be
14    the utility of having the software available in object
15    code, in uncompiled -- or compiled form?
16               MR. DANIEL:  Compiled form.  The utility,
17    it would be useful for people who happened to have
18    that rev of software operating system would run that
19    binary, but it would not be useful in the general
20    case.  There are too many different machine types.
21               Remember, we are talking about servers here
22    that handle the intermediary data and it's not like
Page 19
 1    Windows where Bill Gates has a monopoly and 99 percent
 2    of the machines are this.  The machine types are
 3    spread between different CPUs and different operating
 4    systems, et cetera.
 5               Also, the software that we were doing was
 6    specifically designed as a first pass, as research,
 7    you know, how do you make this work.  And we need to
 8    get people in there figuring out what are the problems
 9    with making the system work widely deployed.  And
10    doing that with binaries is -- well, a binary would
11    not encourage this work that needs to happen.
12               MR. DAVIDSON:  So you were not just asking
13    people to download it, you wanted to collaborate with
14    them and seek their --
15               MR. DANIEL:  Right.  Manufacturers of these
16    class of systems worldwide, in England, in Japan and
17    places like this that we work with need to include
18    this software in their systems and they need the
19    source for DNSSEC to do this.
20               MR. GILMORE:  I guess I would also add, the
21    network will only really be secured when the secure
22    version of DNS comes in each product.  When it comes
Page 20
 1    as part of Microsoft Windows, when it comes as part of
 2    NT, when it comes as part of the HP operating systems
 3    and DEC's operating systems, et cetera.  None of those
 4    vendors would take a binary that they can't maintain,
 5    they can't fix other bugs in.
 6               The security is only a small part of this
 7    applications.  It's a distributed database and large
 8    parts of it, you know, are unrelated to the security.
 9    Those parts are still, still need evolution.  They
10    still have plenty of bugs.  And vendors need to have
11    access to the source code to work on fixing those and
12    evolving it to meet the needs of the future.
13               MR. DANIEL:  And if I might add, you
14    don't -- in serious computer security you don't trust
15    things that you cannot examine.  It's not security if
16    someone says it is.  It's security if you can pore
17    over the actual source code and prove to yourself it's
18    secure.
19               MR. DAVIDSON:  Okay.  Thank you.  Do you
20    have anything else?
21               MR. TIEN:  Yeah.  I went on for a long time
22    and that was actually just a summary, so --
Page 21
 1               MR. DAVIDSON:  Okay.  Go ahead.
 2               MR. TIEN:  But I will probably cut down
 3    what I was going to say because some of it is a little
 4    long and John and Hugh have talked a lot about the
 5    technical stuff.  I want to review the background of
 6    why we are here today and we are here today because
 7    BXA changed its mind about what the software should
 8    be.  And on this I'm going to -- our position, of
 9    course, is that the original classification was
10    correct and the reason why it was correct is because
11    it fit within clear language of the regulations and
12    here I'm referring to 5A002, notes F and G, which
13    relate to authentication for access control and
14    authentication, as well as data authentication,
15    calculating message authentication codes.  And both of
16    those exemptions recognize that cryptography is going
17    to be present within those kinds of applications and
18    would permit that software or hardware to be
19    considered authentication even with the presence of
20    cryptography so long as the application doesn't use it
21    to encrypt for confidentiality and only for
22    authentication.
Page 22
 1               And there is a third exemption, and I'm not
 2    actually sure of its status under the current EAR, but
 3    under the old ITAR from 1992, software that protects
 4    against malicious damage to computers, such as
 5    viruses, would also fall within the authentication
 6    category.  And in the various revs to the EAR under EI
 7    controls this virus thing was moved around.
 8               Currently they took it out, but I think
 9    that the philosophy behind it is consistent with what
10    we have been talking about, which is that this is an
11    application that serves the function of not just
12    protecting a particular computer but the Internet
13    itself from malicious damage, such as the case
14    mentioned.
15               MR. DANIEL:  Actually, I would comment that
16    I don't believe he did malicious damage in the sense
17    of damaging anyone.
18               MR. TIEN:  Okay.  He certainly disrupted
19    things greatly.
20               MR. DANIEL:  Actually, he only disrupted
21    the people who thought they were in control of that
22    space.  All the users downstream from him got the data
Page 23
 1    they expected, they just didn't have any way to prove
 2    it.
 3               MR. GILMORE:  No.  Actually, when he
 4    redirected it they got a different web page and the web
 5    page included a diatribe about how they shouldn't
 6    register names with Network Solutions.
 7               MR. DANIEL:  For all but the one site.
 8               MR. TIEN:  So in any case, I see that the
 9    software in this case, that it can clearly be fit into
10    those exemptions and, therefore, there is no
11    difficulty with BXA's original classification.
12               Now let me turn to the second part of the
13    story, which was the reclassification.  Now, as we all
14    know, that reclassification was apparently triggered
15    by the raising of an apparent inconsistency in BXA's
16    treatment of software of Mr. Karn's diskette in the
17    Karn case and the software here.  And what I -- or
18    what I would like to say to that is I think that, that
19    the government could have responded in that case by
20    resting on the language of the regulations.  That is,
21    the software on Mr. Karn's diskette was not the
22    software at issue in this case.
Page 24
 1               Integrated DNSSEC is not RSAREF.  RSAREF is
 2    a part of it, but they are different applications.
 3    And the application at issue here is authentication
 4    application.  And it would I think have been very
 5    reasonable and consistent with everything in the
 6    written regulation to simply say you are comparing
 7    apples and oranges here.  One is authentication and
 8    one is not.
 9               Unfortunately, that's not how it turned
10    out.  And what we have instead is an attempt to
11    produce a consistency by I think departing from the
12    regulatory language, which does look at items as they
13    are, and focusing too much on the items as they could
14    be.
15               Now, there is two possible tracks here.  I
16    think Mr. Lewis's declaration to some extent implies
17    that the reclassification of the software was based on
18    new information, and while it might have been new in
19    one sense, as the person who wrote the classification
20    request, I want to emphasize that we did not in any
21    way hide what was in the software.
22               We were very clear about what it contained.
Page 25
 1    We were very clear that we wanted to get a
 2    classification for it in source code form.  And that
 3    was why we -- and we, in fact, provided the source
 4    code on diskette to BXA for its technical review.  We
 5    were not interested in trying to -- to be anything but
 6    up-front.
 7               So on this record, I think given that BXA
 8    did not have, really have new facts for it to
 9    reclassify the software, our position is that the only
10    reason for reclassification is a change in the law as
11    applied to those facts.  And we believe that that is
12    the problem here.
13               Now, I will not belabor the applicability
14    of the exemptions or the language of the regulatory
15    exemptions for authentication any further.  What I
16    want to move on to is the standard that was used on
17    reclassification.  And that standard was what I have
18    called modifiability or easy modifiability.  I
19    believe Mr. Lewis's specific words were something like
20    with minimal programming effort or with programming
21    effort.
22               And this poses a number of problems.  And
Page 26
 1    the first and most obvious is the ultra-vires problem,
 2    that this is not the written standard.  And I have
 3    talked about that a lot already so I will go on to the
 4    next one.
 5               And the next one is simply that this
 6    standard would probably swallow up the authentication
 7    exemption at least for any kind of authentication
 8    software.  First, there is no authentication function
 9    without some encryption.  We have talked about that
10    already.  BXA recognizes this in the regulations,
11    which specifically mention that these applications can
12    have encryption.
13               Well, in software it's always possible to
14    modify the program.  So if the rule is going to be
15    modifiability, then all authentication software is
16    going to be crypto software for the purposes of the
17    regs, and that will destroy the exemption.  And as a
18    policy matter, as I have already said, that would be
19    unwise.
20               As a legal matter, it is not what the
21    regulations say.  As a matter of construction, it's an
22    unreasonable way to read the regulations so as to
Page 27
 1    render the authentication exemption superfluous.
 2               Next, we would argue that the standard as
 3    articulated is essentially unadministrable or
 4    unworkable and to the point that it would, if applied
 5    in its present form, be legally vague,
 6    unconstitutionally vague.  It is -- it sweeps very,
 7    very broadly.  Once you say that the issue is can you
 8    change the thing and not how is the thing designed
 9    today, you have opened up an enormous can of worms.
10               I can't even really begin to say how one
11    might attempt to define the scope of that standard
12    once you actually tried to do it.  I think Mr. Lewis
13    tried to do that in the notion of easily, but I think
14    that that doesn't work because you still have to
15    define easy.  Easy for whom?  All these sorts of
16    issues.  And we have seen these issues before in the
17    encryption context with the problems of crypto-shaped
18    holes and they raised the same kinds of issues.
19               Now, having said that it is that we believe
20    that it is unworkable and would be vague, then I think
21    the final point is to go to the due process issue.
22    And you know, we -- our position is that right now we
Page 28
 1    have a relatively clear rule.  It's tied to the
 2    software as it's designed.  What does it do?
 3               To go away from that is to work a very
 4    significant change in the regulations, and I am
 5    concerned that it would not only be a significant
 6    change for this particular area of regulatory
 7    interpretation, but that it would affect the use of
 8    the notion of designed and modified throughout the
 9    EAR.
10               The idea that something is -- and that is
11    the regulatory language that appears throughout the
12    regulations, that something is designed and modified
13    to do X, and in general we are talking about the
14    function of the item as it comes to us.
15               Once you start breaking away from that you
16    are going to have considerable ripple effects I think
17    in the higher administrative machinery.  I will bring
18    up, I mentioned this before, and we have seen a notice
19    for inquiry involving the definition of the term
20    specially designed, which is not directly relevant in
21    this case, but I have reviewed the comments on that
22    and it is -- and I believe the same sort of issue
Page 29
 1    comes up.
 2               You have a much longer history with
 3    specially designed running back to the '50s, which is
 4    a really long time.  But it's stable, it is basically
 5    workable, and there is no need -- there is no need to
 6    introduce uncertainty there.
 7               Here we have a standard that, that at least
 8    dates from 1992, possibly earlier in an unwritten
 9    sense as to what authentication is, how this is going
10    to be handled.  Six or seven years is I think a pretty
11    long time in the encryption area for some kind of line
12    to be drawn.  There is an understanding of it.  I
13    don't -- I see serious problems with changing it and
14    there are ripple effects.
15               From a legal standpoint, agencies are bound
16    by their own regulations, and the BXA clearly can make
17    changes to the rules but they did constrain them in
18    discretion when they created the rules covering the
19    present authentication exemptions.  If you want to
20    change that, you should go through an open process.
21               I think one of the problems is, has been
22    the traditional unreviewability of the BXA
Page 30
 1    item-specific types of decisions.  But I think that in
 2    the end this, although this may give an agency a
 3    certain sense of freedom at one level in dealing with
 4    individual exporters, I think in the end it is
 5    unproductive.
 6               We believe that BXA would be sacrificing
 7    the long-term legitimacy of the export rules for a
 8    small short-term gain.  And I think primarily with
 9    respect to the litigation in the Karn case, I don't
10    believe that this is even a short-term gain with
11    respect to this particular software because it was
12    lawfully published after we received the initial
13    classification.  It was available for a significant
14    period of time without restriction.  It is overseas
15    and we don't think that the short-term benefits of
16    deeming this to be encryption software with the
17    collateral -- outweighs the collateral damage to the
18    scheme as a whole, and certainly to the policy
19    importance of -- of that formation for DNS security.
20               The arbitrariness of making a small change
21    for one person that is actually a significant change
22    for the meaning of the regulations is something that
Page 31
 1    will undermine the legitimacy of the BXA control
 2    regime.  People are encouraged to follow a system that
 3    they perceive to be fair and reasonable, even if it is
 4    not exactly what they want it to be.  Because we all
 5    recognize that systems are not perfect.  But when you
 6    tinker at the margins in order -- in such a way as to
 7    undermine the overall legitimacy, you create a serious
 8    problem.
 9               So I'm going to conclude by saying that
10    there are no perfect rules, but this is pretty good.
11    And the standard that was used for this software is
12    one that is significant and at the very least if it's
13    going to be adopted must be done so in an open fashion
14    with public input on the record because otherwise you
15    are going to have serious problems in the general
16    scheme of regulations.
17               MR. DAVIDSON:  Anyone over here want to say
18    anything?
19               MR. LEWIS:  My thought was that the intent
20    here was to get us to take a second look at this
21    reclassification, so I wondered if you could describe
22    for us what it is that's in the product?  If you could
Page 32
 1    describe that in some detail, that would be helpful.
 2               MR. DAVIDSON:  This is the --
 3               MR. LEWIS:  In the product you are asking
 4    for declassification.  This is better than written.
 5               MR. TIEN:  I certain am not going to try to
 6    answer that.
 7               MR. LEWIS:  Tell us what's in the product.
 8    What is it comprised of?
 9               MR. DANIEL:  Maybe from your point of view
10    it's comprised of three basic blocks.  The first block
11    is a traditional DNS server system that is run on
12    server computers that know how to participate in the
13    hierarchy of DNS-distributed database systems.  This
14    technology has been deployed since sometime in the
15    '80s, mid to late '80s.
16               It was then modified via contract from the
17    U.S. government by TIS to include -- this is for us
18    the critical part -- the software modifications to DNS
19    to implement DNSSEC, DNS secure systems.
20               MR. LEWIS:  Is this still on the first
21    block?
22               MR. DANIEL:  Well, that's the second block.
Page 33
 1    There is the old stuff, there is the new stuff and
 2    there is a library.  Give it that overview.  So the
 3    new stuff is the ability to take a segment of data, of
 4    pairs of information like a DNS name and a DNS -- or
 5    an IP address and then sign those.  And the signature
 6    is simply a crypto hash that if the two things above
 7    when processed by the right math don't add up to the
 8    crypto hash, then you have bogus data.
 9               So TIS wrote that software and that's what
10    makes it DNSSEC.  And then you add in the RSAREF
11    library that actually does that mathematical test for
12    you, which is all we use it for, and that's the last
13    of the three parts.
14               MR. LEWIS:  Okay.
15               MR. DANIEL:  I will point out, I guess,
16    that if you can't actually do the math that says that
17    these two numbers or, you know, this DNS name and this
18    IP number equal this crypto hash, if you can't do that
19    math, there is no security.  If you don't have that
20    last step there there is no reason to bother doing any
21    of this.
22               MR. LEWIS:  And that's the application you
Page 34
 1    use the RSAREF library for?
 2               MR. DANIEL:  Yes.  Well, there is also the
 3    generating of the key pairs originally, and the
 4    signing of the data and stuff like that.  There is a
 5    lot of ancillary things that go with this.
 6               MR. LEWIS:  Okay.
 7               MR. DAVIDSON:  Let me ask a question here.
 8    I'm trying to phrase it intelligently.  Is it correct
 9    or incorrect that the library you spoke of could be
10    used to encrypt text data for confidentiality, as the
11    licensing authorities have said in this case,
12    versus -- do you disagree on a technical matter as to
13    the ability to take this, the source code for this
14    package, out of the program and use it for more than
15    authentication?
16               MR. DANIEL:  The mathematics for doing
17    authentication and encryption is the same.  So
18    mathematically anything that can do encryption --
19    excuse me -- can do authentication can also do
20    encryption.  So any piece of code, be it binary or
21    source, you can pull out and make do the other.
22               MR. GILMORE:  But as the product was
Page 35
 1    created and as we intended to export it, there is no
 2    way for a user to access that functionality.  If they
 3    just build the product, install it, there is nothing
 4    that they can do that will encrypt text data for
 5    confidentiality.
 6               MR. DANIEL:  Even more so, the entire DNS
 7    protocol that we are working with on the security DNS
 8    protocol, the modern version of it, nothing in there
 9    can be transmitted encrypted.  The protocol that 90
10    percent of the software implements or more does not
11    have a place for encrypted communications.  So it
12    would take an expert level programmer to go in and rip
13    out the parts, the mathematics in the bottom that does
14    authentication and make it do encryption.  But that
15    would be no different than given to a binary more or
16    less.
17               MR. TIEN:  There is no secret button in
18    there that allows you to turn it around to use it for
19    something else.  It is to the maximum extent possible
20    under our constraints tied to the authentication.
21    That's what it does.
22               MR. LEWIS:  Could you perhaps describe that
Page 36
 1    in a little more detail, including the way there is no
 2    way for the user to access, just in some more detail?
 3    That would help us.
 4               MR. DAVIDSON:  And excuse me.  Before you
 5    answer, at least for me, do whatever you want, but if
 6    you feel it relevant, differentiate between the
 7    ultimate compiled program and the source code on that.
 8    The question I would guess, maybe the answer is
 9    different, maybe it's not.  Go ahead.
10               MR. TIEN:  John, why don't you.
11               MR. DANIEL:  Yeah, I'm not certain how I
12    would answer your question.
13               MR. DAVIDSON:  If my question is wrong tell
14    me.  But I thought Mr. Gilmore's comment a minute or
15    two ago was something which I don't believe there is
16    any disagreement over, that you can compile the
17    program, you put it on the server and you use it, but
18    it has no capability of encrypting data for
19    confidentiality.  I don't understand that to be the
20    issue here.
21               I think the concern is with the source code
22    and taking the module or whatever out of the source
Page 37
 1    code and then putting it in something else and
 2    compiling that and then having a piece of software
 3    which will encrypt data for confidentiality.  I'm
 4    sorry.  Did you want to say something?
 5               MR. GILMORE:  Sure.  It's certainly
 6    possible for an expert programmer to sort of take a
 7    piece of this and build some scaffolding around it and
 8    create something that would do encryption, but I think
 9    it takes a similar level of effort -- there are many
10    other ways that they could do the same thing,
11    including ways that involve taking nothing out of DNS
12    Security, just using standard textbooks in the field
13    to do the same thing, for example.
14               So I guess having this library in the
15    source code in a sense makes it, you know, makes
16    another path that people could use to produce software
17    that encrypts texts for confidentiality.  But I don't
18    think it significantly increases the likelihood of
19    that happening.  I think if people want to encrypt
20    text for confidentiality outside the States there are
21    ample ways for them to do that.
22               MR. DANIEL:  This is not a simple nor a
Page 38
 1    straightforward way of going about getting
 2    confidentiality.
 3               MR. LEWIS:  Are there things built into the
 4    product that make it more difficult or that affect the
 5    ability to do that though?  Is there any modification
 6    or feature of the product that --
 7               MR. DANIEL:  By contract we cannot modify
 8    the RSAREF library in any way, shape or form.
 9               MR. LEWIS:  That's a contract with RSA?
10               MR. DANIEL:  That's RSA's licensing with
11    everyone.
12               MR. TIEN:  Intellectual property.
13               MR. DANIEL:  So there is nothing we can do
14    at that level to change things.  It's certainly not
15    easy to make it work even right.  That's why this is
16    cutting edge and not easy for everyone to do, because
17    this is a hard problem.
18               MR. ZIA:  The next question is is there a
19    difference, just so I understand, of how programming
20    works?  Which is probably more basic than probably
21    anyone else on this side of the table.  But I
22    understand why you want to allow user access to source
Page 39
 1    codes so they can modify it to fit the application
 2    that they are using, because one size doesn't fit all.
 3               And you talk about there being a
 4    requirement or rather that an expert programmer could
 5    probably rip the RSAREF piece away.  But at what
 6    level, I guess, of programming skill is required just
 7    to modify it to fit a regular application that someone
 8    has?  Don't you have to have -- doesn't that require
 9    more of this expert level in order to use the source
10    code to modify it to, you know, whatever application
11    you want to use it for?  Do you understand my --
12               MR. DANIEL:  You have to write the
13    application and if you are going to write the
14    application there are better, newer, faster, cleaner
15    public domain versions of the same mathematics
16    available worldwide written by non-U.S. citizens that
17    if you use you won't have a pile of RSA lawyers
18    running on your doorstep charging you with
19    intellectual property theft.
20               MR. ZIA:  My point is the person who would
21    be able to use the source code would have to be on an
22    expert level so wouldn't the person be able to figure
Page 40
 1    out a way to pull out the RSAREF?  I mean, are you
 2    saying at one level a person -- sort of a day-to-day
 3    person who has a lot more knowledge than I have but
 4    not qualified or knowledgeable enough to take the next
 5    step?
 6               MR. DANIEL:  It takes a fair amount of
 7    paranoia and experience to do any crypto-related
 8    programming for authentication or privacy and, yes,
 9    you could take a lower level person out there and have
10    them, you know, try to rip a library out of something
11    like integrated DNSSEC and put it into an application
12    to provide privacy and there is maybe one chance in 30
13    that it will be properly done and provide privacy
14    unless you have got someone who is really aware of all
15    the various problems there are using crypto
16    mathematics.
17               MR. TIEN:  If I could restate that.  There
18    is a difference between being good at programming and
19    being good at information security.  There is a big
20    difference.  You could be a really, really good
21    programmer and be able to code your way all around
22    that thing, but that doesn't mean that you can build
Page 41
 1    an application that is actually going to encrypt well
 2    the way that people want to use it.  They are really
 3    two different things.
 4               MR. DANIEL:  I guess maybe I can expand on
 5    this by pointing out that you will notice that the
 6    RSAREF is a small library that we use actually for a
 7    minuscule amount of time that the application is
 8    running, just to run the mathematics that we have to
 9    have to get the authorization.  Most of the rest of
10    the time the program is doing other things.
11               This is -- the same is also true for other
12    applications.  So if you were to do a privacy
13    application, you know, rip out this library from a
14    certain PGP or someone like this, most of the work is
15    doing this other stuff all of the time and actually
16    most of the security is also coming from other places.
17    Where do you get the random numbers?  Where do you get
18    the protocol word?  Much the same as for secure DNS.
19    It's the protocol design, it's the where are these
20    numbers coming from, are they generated properly, et
21    cetera, et cetera.
22               MR. TIEN:  And one of the things -- I mean,
Page 42
 1    Hugh and John were the programmers here, will I think
 2    with some precision continue to refer to RSAREF as a
 3    library.  What it is -- RSAREF is not, as I understand
 4    it, an application.  It --
 5               MR. GILMORE:  Correct.
 6               MR. TIEN:  It is a bunch of sub-routines
 7    that do the mathematical function, the exponentiation,
 8    and various other, that somebody else doesn't really
 9    understand, but it's like having little smart agents
10    that can crunch certain kinds of odd numbers and do
11    certain kinds of mathematical tasks for you.  But you
12    still have to line them up the right way.  You have
13    to -- I mean, all that is involved in an encryption
14    for confidentiality is how you use those sub-routines,
15    not the sub-routines themselves.
16               MR. LEWIS:  What would happen to the
17    product if you took the RSAREF out of it?
18               MR. DANIEL:  It would continue to function
19    as a DNS server in all respects but security.  At that
20    point it would not be able to test the mappings
21    against the crypto hashes within it and you would have
22    no idea if you were talking, as I said in my earlier
Page 43
 1    example, to deposits at Citibank or deposits at
 2    Medellin cartel.  There would be no way to tell.
 3               MR. LEWIS:  Okay.
 4               MR. DANIEL:  So the authentication function
 5    disappears entirely when you take out that math.
 6               MR. LEWIS:  Got it.
 7               MR. ZIA:  One of the earlier questions
 8    about -- you had said that there is no way to modify
 9    the software to protect against the RSAREF without
10    breaching, violating the agreement to use the library.
11    I'm wondering if there is an alternative way to
12    building the -- without modifying the library itself,
13    building a stronger cage, something with bars around
14    it to protect from people that have the capability or
15    making it more difficult for them to try to break
16    through and steal that part of it.
17               MR. GILMORE:  I guess I can speak to that.
18               MR. TIEN:  I'm not going to.
19               MR. GILMORE:  The DNS part is already a
20    cage around it effectively.  We could build another
21    internal cage inside the cage, but I think that the
22    position the department is taking is that it doesn't
Page 44
 1    matter how many layers of cages there are if you can
 2    just let the animal out and reuse it somewhere else.
 3    In source code it's hard to put a cage around a
 4    particular function.
 5               MR. DANIEL:  In the source of binary I can
 6    still go through the cages because at some point my
 7    CPU has to execute those instructions.
 8               MR. GILMORE:  As it turns out, one of the
 9    problems, the intellectual property licensing on
10    RSAREF was one of the problems with this software.  It
11    only permits non-commercial use.  The RSA library only
12    permits non-commercial use.  And this would have
13    prevented widespread deployment of DNS Security.
14               TIS used that library because it was
15    convenient as they were in the research phase.  But I
16    ended up negotiating a license with RSA that permits a
17    different library to be used called DNSsafe that is
18    tailored specifically to the requirements of DNS
19    Security and removes all -- anything other than that,
20    and negotiated with them a worldwide license to permit
21    both commercial and non-commercial use.
22               Now, we negotiated that license more than a
Page 45
 1    year ago and we have that software available but we
 2    got hung up at the RSAREF level with actually trying
 3    to integrate this into the regular DNS maintenance
 4    process, because it's -- because of the department
 5    contending that it's not exportable.
 6               DNS maintenance happens worldwide.  There
 7    is an collection of people, a mailing list.  The
 8    server itself is called BIND, the Berkeley Internet
 9    Name Daemon, and there is a list called BIND-workers of
10    people who work on it intensively and it has about 400
11    people on the list and these people are all scattered
12    around the world.  They go through new releases of the
13    software every month or a couple of weeks, try it on
14    their machines, make sure that it's all working, send
15    back in any local changes that are required and
16    gradually work as a team to produce stable, solid
17    releases that then go out to the vendors, to the main
18    user community.
19               Those people basically were not able to
20    touch the software because of the export uncertainty
21    around it.  They couldn't add the security features.
22    And so part of -- well, we had two issues there.  We
Page 46
 1    had the intellectual property problem that it only
 2    permitted non-commercial licensing and we had the
 3    export.  So I negotiated the license with RSA to deal
 4    with the intellectual property problem and Lee and
 5    Hugh had filed the request to clear up the export
 6    uncertainty and when we got here in '99 we thought
 7    that was all cleared up but then further complications
 8    have made it uncertain.
 9               MR. LEWIS:  Go ahead.
10               MR. TIEN:  I just wanted to add one quick
11    point, because I think a part of what both Jim and
12    Hoyt are asking about is is -- was there -- is there
13    more that can be done to sort of limit the crypto,
14    potential crypto functionality and I think that the
15    basic problem goes back to what Hugh said, which is
16    that this is the math for authentication and the math
17    for encryption are the same.  It's raising and
18    multiplying big numbers and raising numbers to powers
19    and then, and factoring and this sort of stuff and
20    that's common.  You could probably -- you might be
21    able to do things, to take more stuff -- you might
22    have been able to take more stuff out and have the RSA
Page 47
 1    license not been so constraining, but it would still
 2    at the end of the day I think be possible for your
 3    department to say this still contains the math.
 4               MR. DANIEL:  And is still modifiable.
 5               MR. TIEN:  Right.  And therefore is still
 6    modifiable.  And that's why I keep going back to say
 7    that the regulations as they are written already
 8    understand that.  Because they say there can be
 9    cryptography in authentication, they recognize, as you
10    know, by text and policy that the math is the same.
11               MR. ZIA:  Is there authentication software
12    that is different to the extent that you cannot take
13    the, I guess the crypting source code out and use it
14    for encrypting -- for use other than authentication?
15               MR. DANIEL:  At my level, no.  At your
16    level --
17               MR. TIEN:  At our level everything is
18    impossible.
19               MR. DANIEL:  At your level, you are caring
20    about the application, which is what we are doing, and
21    our application does in no way anything privately.  It
22    is just authentication.  Is that distinction clear?
Page 48
 1               MR. ZIA:  I guess what I'm trying to
 2    understand, and you probably may have to go into a lot
 3    more detail is essentially your point, I think, Lee,
 4    is that we have eviscerated this exemption for
 5    authentication so it's an empty box as it were,
 6    because you are saying, I think, that any
 7    authenticating software is going to have the same
 8    problem, that is, someone can go into it if it's a
 9    source code and take out the encryption part of it and
10    use that encryption part to do the same thing that we
11    said the RSAREF will do, but be able to use that to
12    encrypt data as opposed to just perform the
13    authentication function.  And that's what I'm trying
14    to understand, if that's what you are saying, that's
15    true across the board.
16               MR. TIEN:  It is.  Based on my technical
17    level of understanding, which is not that much greater
18    than yours, my understanding is it is not possible.
19    That you will have a problem.  The hardware might be a
20    little bit different, but it's still -- but if there
21    is a chip that does exponentiation, you are still
22    going to be able to modify it or hook it up in a
Page 49
 1    different way.  At the bottom you have -- my
 2    proposition is if you want to have an exemption for
 3    authentication, it's going to have to allow it to have
 4    math.
 5               MR. LEWIS:  Could you describe the
 6    differences between the DNSsafe and RSAREF?
 7               MR. GILMORE:  Well, they come from
 8    different origins within RSA Data Security, who wrote
 9    both of them.  The DNSsafe is based on an old version
10    of their commercial library called BSAFE and they
11    rewrote that library some years ago and decided that
12    it wouldn't threaten them commercially to release
13    parts of the older version for free public use.
14               I think RSA has as much interest -- well,
15    RSA has publicity problems in general.  They have been
16    administering their patents with kind of an iron hand
17    for a long time and they need people on the Internet
18    to feel good about them because the patent is about to
19    expire.  So this was sort of I think a good will
20    gesture on their part saying, look, we are letting
21    this be used for good purposes, we are not charging
22    you for it, we are out there helping to secure the
Page 50
 1    Internet, et cetera.
 2               So they took what was a fairly complete
 3    library called BSAFE and took out everything that DNS
 4    Security didn't need, which takes lots of algorithms
 5    out and takes lots of protocols away to encode data.
 6    They still had to leave in encryption for
 7    confidentiality because the private key that secures
 8    your DNS data has to be encrypted on your disk,
 9    otherwise if someone breaks into your machine they can
10    pull out your private key and impersonate you.
11               But that's the only use of authentication
12    for confidentiality is to protect the keys.  And we do
13    have the ability to modify that library, you know, if
14    we need to, which we didn't have with RSAREF.  So
15    there are certainly things that could -- if you can
16    think of things that could be done to it that would
17    continue to allow it to be used for authentication and
18    would somehow make it more acceptable for you, then
19    that would be an interesting thing to explore.
20               MR. LEWIS:  But that wasn't -- you didn't
21    submit a DNSsafe for us to review?
22               MR. GILMORE:  No.  It took -- well,
Page 51
 1    integrating that into the rest of the TIS stuff took
 2    some time.
 3               MR. TIEN:  Because the TIS release that --
 4    that Hugh started with was based around the RSAREF
 5    library.
 6               MR. GILMORE:  Because DNSsafe didn't exist
 7    at that time.
 8               MR. TIEN:  DNSsafe didn't exist until when?
 9               MR. GILMORE:  I would have to look.  It was
10    last year, I think fairly early.
11               MR. TIEN:  But that was in early 1998.
12               MR. DANIEL:  No.  It was mid spring by the
13    time we were done with that.
14               MR. TIEN:  And we submitted the original
15    classification in '97 and actually I started the
16    process in December of '96.
17               MR. DANIEL:  One thing that might be
18    helpful to point out here is that this is critical
19    infrastructure software that people who are going to
20    use this have long lead times on.  This is going into
21    operating systems like UNIX and Linux and other
22    operating systems that we don't even understand and
Page 52
 1    know about.  It's going into financial networks.  So
 2    there is a fair amount of time that anyone needs to
 3    start trusting the software, to start working with it
 4    and fitting it into their environments, which is why
 5    we have been working on this for several years already
 6    even though the problem is just beginning to creep up
 7    into annoyance level.
 8               MR. LEWIS:  Do you have a version of the
 9    product that incorporates the product, integrated
10    DNSSEC that incorporates DNSsafe rather than RSAREF?
11               MR. GILMORE:  I don't have a version but Trusted
12    Information Systems does and we could try talking to
13    them about it.
14               MR. LEWIS:  Okay.
15               MR. DANIEL:  But certainly it's not in the
16    main line of development of the software at this point
17    in time.
18               MR. LEWIS:  Why is that?
19               MR. DANIEL:  Partially because we had
20    export permission for integrated DNSSEC and that's
21    what people started using.
22               MR. TIEN:  Resources started going into
Page 53
 1    that.
 2               MR. DANIEL:  That's why there are copies
 3    available worldwide on various web servers is because
 4    people grabbed it and started working with it.
 5               MR. DAVIDSON:  So is development on the
 6    problem that you identified at the hearing of the
 7    vulnerability of the DNS system, is development
 8    worldwide still progressing based on this software
 9    package?
10               MR. DANIEL:  Based on integrated DNSSEC
11    because that's all that is available.
12               MR. GILMORE:  It's happening on multiple
13    tracks.  The DARPA has paid TIS to continue evolving
14    the software, which in our version is frankly a
15    prototype.  It wasn't ready for prime time.  When we
16    first submitted it it was a while ago and TIS has done
17    more work since then.  But we didn't see the point in
18    continually trying to update the software in your
19    process because the result should be the same.  You
20    know, the software itself was really not at issue
21    here.  What's at issue is the principle of is it
22    possible to export authentication software in source
Page 54
 1    code.  If it's not possible, then it doesn't matter
 2    which software you are going to export.
 3               MR. LEWIS:  Putting aside some of the IP
 4    questions and your licensing questions with RSA, is it
 5    possible to make RSAREF more like DNSsafe?  And you
 6    were describing how DNSsafe was limited by RSA.  Is it
 7    possible to do that to RSAREF?
 8               MR. GILMORE:  It might be possible.
 9               MR. LEWIS:  Technically possible and then
10    you have the IP problems.
11               MR. DANIEL:  You could certainly go in and
12    delete unused portions of the RSAREF library in the
13    integrated DNSSEC.  When you were done deleting those
14    portions you will have deleted parallel encryption
15    abilities and parallel authentication abilities but
16    there would still be encryption or there would still
17    be the math that is the same for both encryption and
18    authentication in there.
19               MR. LEWIS:  Go over that a little more
20    slowly.  So you could delete unused portions?
21               MR. DANIEL:  You could delete the unused
22    portions, but you would be like cutting off a man's
Page 55
 1    left arm and being annoyed that he still has a right
 2    to hold the saber in.  And we are going to have to
 3    have an arm no matter what because the mathematics is
 4    the same.  So you can get rid of some of the
 5    duplication and make the library smaller and more
 6    focused, and in some sense from your point of view you
 7    are simply making it easier for other people to use
 8    the library at that point because it is now a smaller,
 9    simpler target that makes more sense and has more
10    examples of how it's used.
11               MR. LEWIS:  But what would you cut out?
12    What would you delete?
13               MR. DANIEL:  Mostly there is -- actually, I
14    don't know what all else is in there because I don't
15    use it.  The only calls I have looked at are ones on
16    the main line of what we are doing.  There are some
17    other encryption algorithms in there that aren't used
18    by integrated DNSSEC that -- mostly they are the
19    symmetric key stuff.  Isn't there a triple --
20               MR. GILMORE:  I don't actually know.  I
21    haven't gone through that exercise either.
22               MR. TIEN:  The government claims that there
Page 56
 1    is some sort of --
 2               MR. DANIEL:  Yeah.  Can you delete that?
 3    There are some other modes that it has for calling the
 4    RSA functions in.  There is a lot of weird
 5    cryptographic, is it in CBC mode or blah, blah, blah.
 6    There are some other calls into the system that we
 7    don't use that can be deleted.  But once again, that's
 8    kind of trimming the fat off and the steak is still in
 9    the middle there.
10               MR. TIEN:  We certainly, I don't believe in
11    all our discussions, we don't have -- those are not --
12    there is no resistance to doing that.  It just
13    couldn't --
14               MR. GILMORE:  Well --
15               MR. DANIEL:  It's a waste of time.
16               MR. TIEN:  Well, okay.
17               MR. GILMORE:  There is a resistance on one
18    level because there is an issue that under one
19    interpretation of the regulations the EAR doesn't even
20    apply, because it's public domain authentication,
21    which is exempted from the strict -- or exempted from
22    the entire controls of the EAR.
Page 57
 1               So on that interpretation it's basically
 2    sort of a First Amendment cut-out that says you have
 3    the right to publish this and the government can't
 4    tell you what you are allowed to say or not say.  On
 5    this other interpretation of the regulations then, the
 6    government has full discretion to say what can go
 7    across the border and what can't.  And there is a very
 8    sharp cliff between those that sort of rests on what
 9    exactly do those regulations mean?
10               And that sort of goes back to Lee's earlier
11    point that says, I think it behooves the department to
12    make -- to constrain itself to actually following its
13    own regulations, even though it can't necessarily be
14    forced to do that by courts because it encourages
15    people to actually follow them themselves.
16               If as an exporter I'm likely to get
17    arbitrary treatment when I send in a license
18    application I won't send one in.  I will do my best to
19    follow the regulations as written and then I will just
20    go with the luck, you know, rather than poke my head
21    up and attract the attention of an arbitrary
22    department.
Page 58
 1               MR. DAVIDSON:  I had a couple follow-ups
 2    and let me make a preface.  As we were talking before,
 3    this is appeal number seven of '98 so we don't have a
 4    lot of these, but we often hear that the adverse
 5    decision or the impact and harm of the adverse
 6    decision of the department on the exporter, and you
 7    have talked about that to a great extent on the
 8    vulnerability of the system and all of that.
 9               I guess my question, without knowing
10    whether the answer is even relevant, is what other
11    impacts, if any, are there?  Is there any pecuniary
12    impacts on you, any proprietary impacts on you?  And I
13    guess a part of that is while I noted your
14    conscientiousness and apparent good faith at solving a
15    very difficult problem that I'm not sure anyone
16    denies, I'm not sure I understand your role in this
17    software and its development.  And I don't know
18    whether that information would be relevant to the
19    appeal or not.
20               It would probably not be supremely relevant
21    but it might be interesting to know what role you play
22    in this in either a pecuniary measure or an
Page 59
 1    intellectual measure or whatever.
 2               MR. TIEN:  I'm going to let them answer
 3    that question but I want to preface their answer with
 4    a general comment.  One of the problems I think that
 5    your export control regime in the existing EAR has
 6    with people like my client is that you have been
 7    oriented toward a commercial, very commercial world of
 8    large businesses that have a lot of contracts with
 9    arrangements, with business contracts and arrangements
10    with the government or other governments or other
11    foreign entities and it's all a very commercial sort
12    of world.  But the -- the Internet, the people that
13    built the Internet or have been maintaining the
14    Internet since the U.S. Government sort of really
15    jump-started it, are -- at most have one foot in that
16    world.  There is certainly a major commercial
17    component to Internet maintenance and development, but
18    there is equally a major, you know, hacker,
19    non-commercial, you know, wizards who stay up late
20    component of organizations like the IETF, which I
21    believe stands for the Internet Engineering Task
22    Force, the Internet Society, the way that domain names
Page 60
 1    used to be handled.
 2               There is a huge, one might call it
 3    non-commercial infrastructure of programmers, of
 4    computer scientists, of systems administrators and
 5    engineers who have been doing this, and the work they
 6    do, the way they put out software, is not the same way
 7    that Microsoft puts out software or, you know, these
 8    other companies.  And it is a much more collaborative,
 9    much more non-profit and much more sort of sharing of
10    the information and that's one reason why I am
11    detecting, and don't want to put words in anybody's
12    mouth, but a big part of the problem here is that we
13    asked for a classification for the source code and
14    your reaction to that seems to be, oh, my God, why are
15    they doing this?  If you are doing this in source code
16    you must be doing it in order to make it easier for
17    people to mess with.
18               But the tradition in what I am calling the
19    sort of other side of the Internet maintenance is
20    source code, and not for nefarious reasons, but
21    because that's the way to get the work done.
22               MR. GILMORE:  I do have some small
Page 61
 1    pecuniary interest in this in the sense that I paid
 2    Hugh as a contractor to work on this stuff, I paid Lee
 3    to do all the legal work on this stuff.  To the extent
 4    that the project doesn't accomplish its objective of
 5    securing the Internet, I have spent money and gotten
 6    nothing out of it.  So I have an interest there.
 7               But the amount of money is relatively small
 8    compared to lots of other things that I do.  The
 9    overall interest, the reason that I'm doing this at
10    all, is because I care about the stability of the
11    Internet.  I'm on the Board of Directors of the
12    Electronic Frontier Foundation which looks at civil
13    rights and civil responsibility on the Net.  I'm on
14    the Board of the Internet Society, which exists to
15    foster the culture of the Internet and continue the
16    evolution, technical evolution of the Internet.
17               As Lee pointed out, the Internet
18    Engineering Task Force is one of the things that
19    happens under the Internet Society umbrella, and there
20    has been a much stronger focus in the last five years
21    in Internet engineering on realizing that the network
22    that we all put together was a research prototype,
Page 62
 1    funded by an advanced research agency to collect, to
 2    connect up centers of excellence in universities.
 3               It wasn't designed to withstand massive
 4    attacks.  It wasn't designed to underpin the entire
 5    infrastructure of commerce or of how citizens interact
 6    with their government or even to carry the volume of
 7    love letters between husband and wives that it now
 8    carries.
 9               So there has been a conscious and
10    significant effort to re-engineer protocols of the
11    Internet so that they do actually provide utility
12    quality protections against disruptions, whether
13    malicious or accidental and such.
14               And I have spent a long time in the
15    Internet culture.  I worked at Sun Microsystems in
16    the early days, which benefited greatly from being an
17    Internet company.  Frankly, it made me rich.  It
18    raised me out of the middle class and has given me the
19    opportunity to work on what I choose to work on and
20    what I have been choosing to work on is to try to pay
21    back society for the wealth it's given me by using
22    some of that wealth to try to raise the
Page 63
 1    infrastructure, to try to make it something that as we
 2    all come to depend on it it won't fail us in times
 3    when we need it.  So that's what brought us into this
 4    whole process.
 5               MR. LEWIS:  Can I ask who the target
 6    audience was for the product?  Who are you thinking
 7    would get this?  If you want to export it, it's not a
 8    mass market product, right?
 9               MR. DANIEL:  It's not a mass market
10    product.  There are two categories of people.  Early
11    adopters, large sites, universities, maybe a small
12    nation state domain like dot IE for Ireland or
13    something like this that wanted to start getting this
14    into their processes, into their operations, learning
15    the technology, how it works, how they can use this to
16    authenticate their end-users.  Stuff like this.
17               And the other major category is people
18    doing these systems, doing UNIX distributions, OSs,
19    maybe even large applications like say a bank to start
20    figuring out how do I test, what trust can I put in it
21    knowing that this IP address equals this name, et
22    cetera.  It certainly wasn't going to be used by your
Page 64
 1    average Joe small business, although they need to use
 2    it eventually.
 3               MR. LEWIS:  But you would see them getting
 4    it from a service provider, being built into something
 5    they purchased not something they would --
 6               MR. DANIEL:  Yeah.  In the long run this is
 7    too much fiddling around by a long shot for people to
 8    add in.  Although, right now that's what I'm working
 9    on getting people to do.  For instance, until you run
10    the software you can't even mirror my data or John's
11    data in the DNS space.  Because it breaks the old
12    servers.  So getting it out there is important to us
13    because right now our -- since we run this stuff
14    experimentally our data isn't well distributed across
15    the space because servers that get it often crash.
16               I wanted to answer the previous question
17    and that is there are reasons I'm involved in doing
18    this too, not just because it's work for hire or
19    something like that.  I'm doing this work because I
20    have spent a long time building community in
21    cyberspace.  I was the person who suggested and helped
22    create the first UNIX box that provided access to the
Page 65
 1    Internet for anyone in the community anywhere.  This
 2    was back in Michigan in the early '80s.
 3               So I have been trying to provide access to
 4    the Net and tried to explain how it's useful to people
 5    for many years.  I have also been doing consulting in
 6    many industries, including in commercial and financial
 7    industries, and when somebody says I want to do my
 8    banking over the Net, how do I make certain I'm
 9    talking to the right people, I kind of go, well, you
10    don't.  You don't have a choice of securing your
11    infrastructure until this problem is generically
12    solved.
13               And I foresee that someday in the future
14    when I'm bored and I need to go off and find some
15    other work and I need these tools in my civil society,
16    and that this really is for me building a civil
17    society; extending the tools that we have been
18    developing in our day-to-day world into the Internet
19    world where it's appropriate and useful.  And,
20    finally, I'm going to warn you as John was saying, the
21    IETF is developing a security -- security --
22               MR. GILMORE:  Architecture.
Page 66
 1               MR. DANIEL:  Not architecture.  There is a
 2    consensus that this has to go in everything everywhere
 3    all the time and the IETF is specifying protocols that
 4    implement this stuff.  And either this particular one
 5    is unusual in that this is a base technology that we
 6    all have to share, but many of the future things that
 7    the IETF is defining are more application level things
 8    than this, and they are going to have crypto in them
 9    and those products can come from within the U.S. or
10    not.
11               MR. LEWIS:  Do you find it, your product
12    now being used in the United States and Canada to
13    which the export -- you have a community of people you
14    want to distribute this to and some are in the U.S.
15    and some are not.  Did you make the distribution to
16    the ones in the U.S. and are they using it or is there
17    some other --
18               MR. DANIEL:  It is in use in a small number
19    of places inside Canada and the U.S., and maybe even a
20    smaller number of places outside.  It really does need
21    to be reved, it does need to be pushed forward.  We
22    need to go through taking the work that TIS and others
Page 67
 1    have been doing and folding it back into this
 2    mainstream stuff all at once and get it out again.
 3               MR. LEWIS:  And what are they using it for?
 4               MR. DANIEL:  It doesn't do anything but
 5    DNSSEC.
 6               MR. GILMORE:  All the people who are using
 7    it are basically experimenters.  They are people who
 8    are trying to push the technology forward.
 9               MR. DANIEL:  Universities, corporations.
10               MR. GILMORE:  One of the big issues around
11    deploying this is nobody has ever built at this big a
12    database with authentication in it, this widely
13    distributed a database.  You tend to have sort of
14    centralized Oracle servers and they tend to have much
15    different characteristics.
16               We really don't know how easy or hard it
17    would be to operate a widely distributed database that
18    has keys that expire every six months and the data
19    needs to be resigned.  There is a huge operational
20    component that we have never gotten to because we
21    couldn't get the software out to people.
22               So we have tried to deploy that inside the
Page 68
 1    United States to individuals to say please start using
 2    this, tell us what operational issues come up and as
 3    we evolve the software we can make it easier for
 4    addressing those issues.
 5               MR. TIEN:  But it's a scale problem.  We
 6    just don't know whether it scales up.
 7               MR. GILMORE:  Yeah, there is a scale
 8    problem.  There is an ease of use problem, that
 9    something that a wizard administrator can do is much
10    different than an ordinary administrator or end-user
11    can do.  But there is a very sharp line between being
12    able to publish something and having to give out
13    individual copies to people.  And the way the fluid
14    and kind of evolutionary process of Internet
15    development has occurred has been through publication.
16               MR. LEWIS:  On the Net you mean?
17               MR. GILMORE:  Well, even before the Net,
18    though.
19               MR. LEWIS:  You don't mean a book, right?
20               MR. GILMORE:  No, I don't mean a book.  But
21    public access.  For example, I think the reason the
22    Internet exists today is because Berkeley made their
Page 69
 1    version of UNIX freely available.  DARPA paid U.C.
 2    Berkeley to write the Internet protocols and put them
 3    into a version of UNIX and Berkeley would basically
 4    sell that on tape to anybody who wanted it.
 5               MR. DANIEL:  And that wasn't a sale for
 6    profit.  It was just the cost of duping a tape.
 7               MR. GILMORE:  Right.  All of the computer
 8    companies like Sun, went to Berkeley, got the tape,
 9    put that software in their systems and all of a sudden
10    all those systems could talk on the Net and it sort
11    of mushroomed from there, because it was free.
12               MR. TIEN:  Like HTTP protocol invented by
13    what, Berners-Lee.
14               MR. GILMORE:  At CERN.
15               MR. TIEN:  Right.  And it was made public
16    and which allowed everybody to adopt it and talk to
17    each other.
18               MR. GILMORE:  I call it reducing the
19    transaction costs of cooperation.  Basically, you
20    don't need to ask someone for permission, you don't
21    need to ask somebody to get a copy, you can get a copy
22    if you have the interest, mess with it and send back
Page 70
 1    improvements to it.  And that has caused much more
 2    fluid, much more evolutionary development on the
 3    Internet than the traditional style of we are going to
 4    cut a major release every two years and the customers
 5    can't modify it and can't see how it works.
 6               MR. DAVIDSON:  Are we close to finishing?
 7    If so, let me --
 8               MR. TIEN:  I have nothing more to say.
 9               MR. DAVIDSON:  All right.  Let me just make
10    one comment, which you may want to respond to now,
11    respond to in writing or not respond to at all.  And,
12    in fact, I think Mr. Reinsch has indicated that he
13    would consider any further information you want to
14    submit.
15               MR. TIEN:  It is my intent to do so.
16               MR. DAVIDSON:  Speaking only for myself, I
17    understand your arguments and I think for the most
18    part they are well set out and I think the issues are
19    pretty clear.  The one argument that I do question is
20    this.
21               Let's assume for the sake of argument that
22    Mr. Reinsch were to decide that it is correct to
Page 71
 1    classify the software as 5D002 whatever.  Then he goes
 2    and -- well, he may want to end at that point.  But
 3    you, I believe, would want him to consider the reason
 4    for the change from '97 to '98, whatever the years
 5    were.  And you address that in your letter, you
 6    addressed it here.  I guess I don't sort of as a legal
 7    matter see what is the test he develops.
 8               If he assumed, if he finds that it is a
 9    correct classification -- if he finds it's an
10    incorrect classification then it doesn't matter
11    because presumably if it was incorrect it will make it
12    into the EAR 99 and you are happy except for the year
13    it wasn't.  What if, however, he determines that it's
14    correctly classified in June of '98?  What is he
15    looking for legally?  What's the test?
16               I guess you could get off into things like
17    the guilty man goes free because the magistrate messed
18    up and those kinds of things.  I don't know if you
19    want to address this now.  Maybe you don't.  It just
20    seems to me that it's not clear how he would say while
21    the June '98 decision was correct on the regulations,
22    something which you don't concede of course, but let's
Page 72
 1    assume he gets to that point, but that he should then
 2    say well because of some procedural error I'm going to
 3    make it EAR 99?
 4               MR. TIEN:  Well, this is a tentative
 5    response but, you know, of course our position is that
 6    the -- first of all, I have to start with our
 7    position, which is that the original decision was
 8    correct and the original decision was correct because
 9    of the way the Regs are written.  I actually have a
10    very hard time sort of grappling with the hypothetical
11    that you set forth because I see no way that it can be
12    done consistently with the language of the regulation,
13    given that there is new information, given the
14    language of the regulations that permit authentication
15    which contain crypto, which would seem clearly to
16    understand the technical point we have been attempting
17    to develop.  What I think that Mr. Reinsch would be
18    better off explaining, and deciding that this was I
19    guess 5D002, that this is authentication software and
20    explaining why it is authentication software and not
21    encryption software along the lines that I discussed
22    previously, when I was, you know, criticizing the way
Page 73
 1    that the government responded in the Karn case, that
 2    it was not simply not necessary for the government
 3    to -- the government was in a sense, when it
 4    reclassified integrated DNSSEC, was answering the
 5    wrong question.  It was -- it was comparing two things
 6    that were not the same.  It was attempting to make --
 7    essentially, we believe that Mr. Karn raised a false
 8    inconsistency that --
 9               MR. DAVIDSON:  Excuse me.  If I could
10    interrupt.  I think you have made that argument before
11    and I understand that.
12               MR. TIEN:  Okay.
13               MR. DAVIDSON:  And I can't speak for
14    Mr. Reinsch, but I believe that if I were deciding it
15    I would first address the issue you just talked about,
16    whether or not it was correct to change.  And you
17    don't have to answer this.  I'm just saying, if you
18    lose on that, what -- how does he -- you are writing
19    his opinion for him.  He has unfortunately for you
20    decided that it was correct to do, at least as a
21    matter of the software and the Regs, but you want him
22    to come out with an answer that is EAR 99.  How do you
Page 74
 1    write that opinion?  That's my question.
 2               MR. TIEN:  I think I will have to answer it
 3    because I believe the only real answer I can give will
 4    be after some reflection in my submission, but I
 5    believe that the half the answer that I'm prepared to
 6    give now is that this was a -- the factual situation
 7    here is quite unusual.  I believe it's quite unusual
 8    to have something receive a classification and then as
 9    a result of that classification be made publicly,
10    lawfully publicly available for a significant period
11    of time, not a day, not a week, no accidental clerical
12    error, but a full reasonable reliance on a
13    classification from the department, which led to its
14    worldwide distribution and that there are some errors
15    that, you know, which I think that the classification
16    was, which when you look at all the facts it doesn't
17    make sense to put a classification on it.  It just is
18    out of variance with reality.  But that's only half
19    my -- I will expand.
20               MR. DAVIDSON:  I understand.  That is my
21    view on it.  Not necessarily anybody else's.  Did you
22    want to say something?
Page 75
 1               MR. GILMORE:  Sure.  To me it looks like an
 2    end-use issue.  The new regulations that came out in
 3    December make significant differences in exportability
 4    of various kinds of crypto based on the end-use.
 5    Whether you can do this, you can send the same
 6    software out of the country for banking but you can't
 7    send it out for metallurgy or whatever.  In this case
 8    I think you could draw a clear distinction that says
 9    if this software was used for confidentiality it would
10    be 5D002.  Because it's end-use is authentication then
11    it falls into this exemption.
12               MR. LEWIS:  Usually the way we do that is,
13    I was thinking along similar lines, but we would do it
14    as a practical matter a little differently, which is
15    that the end-use doesn't determine the classification,
16    the end-use will determine whether or not an item can
17    be approved for export and that's where the one other
18    scenario is even if this is 5D002, it doesn't mean it
19    can't be exported to some group of end-users and that
20    would be depending on the outcome of the appeal.
21               MR. DANIEL:  It's not clear to us who the
22    end-users, who the early adopters are.  Sometimes it's
Page 76
 1    universities, sometimes it's a small company.
 2    Remember, we are dealing largely here also with the
 3    Linux community, so often you will end up with some
 4    small company that has got maybe a bank and they might
 5    go off and work on it, make a bunch of changes, make
 6    it operational and then feed those changes back.
 7    Maybe they will feed them back to South Carolina or
 8    Germany for the distribution.  We don't know.  It's
 9    whoever gets there first and is most interested in
10    doing the work.
11               MR. LEWIS:  There are vehicles that would
12    accommodate that.  I'm not sure they will work in this
13    case.  But you can get away from identifying.  You
14    want people just to have access to it.  You want to
15    reduce transaction costs.  There is ways to do that.
16    Whether it works, again whether --
17               MR. DANIEL:  It's more than reducing our
18    transaction costs.  It's making it easy enough for
19    people who are suckers enough to actually go off and
20    be the first in line to try to make this work, to
21    actually get at the tools to try.
22               MR. LEWIS:  Sure.
Page 77
 1               MR. DAVIDSON:  I would be happy if any
 2    person wants to keep going, but this is hard on the
 3    court reporter right now, so if there is nothing more
 4    to say on the record I would suggest we close the
 5    hearing.  Is that okay?
 6               MR. TIEN:  Fine with me.  Unless you -- you
 7    are the one.  If you have more questions I am happy to
 8    answer.  But if you are done, then we are done.
 9               MR. DAVIDSON:  Thank you very much.
10                          -   -   -
11          (The hearing was concluded at 12:15 p.m.)
12                          -   -   -
13
14
15
16
17
18
19
20
21
22
Page 78
 1               I, ROBERT M. JAKUPCIAK, DO HEREBY CERTIFY
 2    THAT THE TESTIMONY OF THE PARTIES WAS TAKEN BY ME IN
 3    STENOTYPY AND THEREAFTER REDUCED TO PRINT UNDER MY
 4    DIRECTION; THAT SAID HEARING IS A TRUE RECORD OF THE
 5    TESTIMONY GIVEN BY THE PARTIES; THAT I AM NEITHER
 6    COUNSEL FOR, RELATED TO, NOR EMPLOYED BY ANY OF THE
 7    PARTIES TO THE ACTION IN WHICH THIS HEARING WAS TAKEN;
 8    AND, FURTHER, THAT I AM NOT A RELATIVE OR EMPLOYEE OF
 9    ANY ATTORNEY OR COUNSEL EMPLOYED BY THE PARTIES
10    HERETO, NOR FINANCIALLY OR OTHERWISE INTERESTED IN THE
11    OUTCOME OF THE ACTION.
12
13                        NOTARY PUBLIC IN AND FOR THE
                          DISTRICT OF COLUMBIA
14
15    My Commission Expires:
      November 2004
16
17
18
19
20
21
22
                   BLOCK COURT REPORTING, INC.
         The High-Tech Leader in Court Reporting Services
             (202) 638-1313     (800) 735-3376 (DEPO)