draft-ietf-dnsext-ecc-key-07.txt

bind 源码 最新实现 linux/unix/windows平台

   the order defined by the picture.

   Of the remaining parameters, PFHKQABCGY are variable length.  When
   present, each is preceded by a one-octet length field as shown in the
   diagram above.  The length field does not include itself.  The length
   field may have values from 0 through 110.  The parameter length in
   octets is determined by a conditional formula: If LL<=64, the
   parameter length is LL.  If LL>64, the parameter length is 16 times
   (LL-60).  In some cases, a parameter value of 0 is sensible, and MAY
   be represented by an LL value of 0, with the data field omitted.  A
   length value of 0 represents a parameter value of 0, not an absent
   parameter.  (The data portion occupies 0 space.)  There is no
   requirement that a parameter be represented in the minimum number of
   octets; high-order 0 octets are allowed at the front end.  Parameters
   are always right adjusted, in a field of length defined by LL.  The
   octet-order is always most-significant first, least-significant last.
   The parameters H and K may have an optional sign bit stored in the
   unused high-order bit of their length fields.

   LP defines the length of the prime P.  P must be an odd prime.  The
   parameters LP,P are present if and only if the flag M=1.  If M=0, the
   prime is 2.

   LF,F define an explicit field polynomial.  This parameter pair is
   present only when FMT = 1.  The length of a polynomial coefficient is
   ceiling(log2 P) bits.  Coefficients are in the numerical range
   [0,P-1].  The coefficients are packed into fixed-width fields, from
   higher order to lower order.  All coefficients must be present,
   including any 0s and also the leading coefficient (which is required
   to be 1).  The coefficients are right justified into the octet string
   of length specified by LF, with the low-order "constant" coefficient
   at the right end.  As a concession to storage efficiency, the higher
   order bits of the leading coefficient may be elided, discarding high-
   order 0 octets and reducing LF.  The degree is calculated by


R. Schroeppel, et al                                            [Page 6]


INTERNET-DRAFT                                       ECC Keys in the DNS


   determining the bit position of the left most 1-bit in the F data
   (counting the right most bit as position 0), and dividing by
   ceiling(log2 P).  The division must be exact, with no remainder.  In
   this format, all of the other degree and field parameters are
   omitted.  The next parameters will be LQ,Q.

   If FMT>=2, the degree of the field extension is specified explicitly,
   usually along with other parameters to define the field polynomial.

   DEG is a two octet field that defines the degree of the field
   extension.  The finite field will have P^DEG elements.  DEG is
   present when FMT>=2.

   When FMT=2, the field polynomial is specified implicitly.  No other
   parameters are required to define the field; the next parameters
   present will be the LQ,Q pair.  The implicit field poynomial is the
   lexicographically smallest irreducible (mod P) polynomial of the
   correct degree.  The ordering of polynomials is by highest-degree
   coefficients first -- the leading coefficient 1 is most important,
   and the constant term is least important.  Coefficients are ordered
   by sign-magnitude: 0 < 1 < -1 < 2 < -2 < ...  The first polynomial of
   degree D is X^D (which is not irreducible).  The next is X^D+1, which
   is sometimes irreducible, followed by X^D-1, which isn't.  Assuming
   odd P, this series continues to X^D - (P-1)/2, and then goes to X^D +
   X, X^D + X + 1, X^D + X - 1, etc.

   When FMT=3, the field polynomial is a binomial, X^DEG + K.  P must be
   odd.  The polynomial is determined by the degree and the low order
   term K.  Of all the field parameters, only the LK,K parameters are
   present.  The high-order bit of the LK octet stores on optional sign
   for K; if the sign bit is present, the field polynomial is X^DEG - K.

   When FMT=4, the field polynomial is a trinomial, X^DEG + H*X^DEGH +
   K.  When P=2, the H and K parameters are implicitly 1, and are
   omitted from the representation.  Only DEG and DEGH are present; the
   next parameters are LQ,Q.  When P>2, then LH,H and LK,K are
   specified.  Either or both of LH, LK may contain a sign bit for its
   parameter.

   When FMT=5, then P=2 (only).  The field polynomial is the exact
   quotient of a trinomial divided by a small polynomial, the trinomial
   divisor.  The small polynomial is right-adjusted in the two octet
   field TRDV.  DEG specifies the degree of the field.  The degree of
   TRDV is calculated from the position of the high-order 1 bit.  The
   trinomial to be divided is X^(DEG+degree(TRDV)) + X^DEGH + 1.  If
   DEGH is 0, the middle term is omitted from the trinomial.  The
   quotient must be exact, with no remainder.

   When FMT=6, then P=2 (only).  The field polynomial is a pentanomial,
   with the degrees of the middle terms given by the three 2-octet


R. Schroeppel, et al                                            [Page 7]


INTERNET-DRAFT                                       ECC Keys in the DNS


   values DEGH, DEGI, DEGJ.  The polynomial is X^DEG + X^DEGH + X^DEGI +
   X^DEGJ + 1.  The values must satisfy the inequality DEG > DEGH > DEGI
   > DEGJ > 0.

        DEGH, DEGI, DEGJ  are two-octet fields that define the degree of
           a term in a field polynomial.   DEGH is present when FMT = 4,
           5, or 6.  DEGI and DEGJ are present only when FMT = 6.

        TRDV is a two-octet right-adjusted binary polynomial of degree <
           16.  It is present only for FMT=5.

        LH and H define the H parameter, present only when FMT=4 and P
           is odd.  The high bit of LH is an optional sign bit for H.

        LK and K define the K parameter, present when FMT = 3 or 4, and
           P is odd.  The high bit of LK is an optional sign bit for K.

   The remaining parameters are concerned with the elliptic curve and
   the signature algorithm.

        LQ defines the length of the prime Q.  Q is a prime > 2^159.

   In all 5 of the parameter pairs LA+A,LB+B,LC+C,LG+G,LY+Y, the data
   member of the pair is an element from the finite field defined
   earlier.  The length field defines a long octet string.  Field
   elements are represented as (mod P) polynomials of degree < DEG, with
   DEG or fewer coefficients.  The coefficients are stored from left to
   right, higher degree to lower, with the constant term last.  The
   coefficients are represented as integers in the range [0,P-1].  Each
   coefficient is allocated an area of ceiling(log2 P) bits.  The field
   representation is right-justified; the "constant term" of the field
   element ends at the right most bit.  The coefficients are fitted
   adjacently without regard for octet boundaries.  (Example: if P=5,
   three bits are used for each coefficient.  If the field is GF[5^75],
   then 225 bits are required for the coefficients, and as many as 29
   octets may be needed in the data area.  Fewer octets may be used if
   some high-order coefficients are 0.)  If a flag requires a field
   element to be negated, each non-zero coefficient K is replaced with
   P-K.  To save space, 0 bits may be removed from the left end of the
   element representation, and the length field reduced appropriately.
   This would normally only happen with A,B,C, because the designer
   chose curve parameters with some high-order 0 coefficients or bits.

   If the finite field is simply (mod P), then the field elements are
   simply numbers (mod P), in the usual right-justified notation.  If
   the finite field is GF[2^D], the field elements are the usual right-
   justified polynomial basis representation.





R. Schroeppel, et al                                            [Page 8]


INTERNET-DRAFT                                       ECC Keys in the DNS


        LA,A is the first parameter of the elliptic curve equation.
           When P>=5, the flag A = 1 indicates A should be negated (mod
           P).  When P=2 (indicated by the flag M=0), the flag A = 1
           indicates that the parameter pair LA,A is replaced by the two
           octet parameter ALTA.  In this case, the parameter A in the
           curve equation is x^ALTA, where x is the field generator.
           Parameter A often has the value 0, which may be indicated by
           LA=0 (with no A data field), and sometimes A is 1, which may
           be represented with LA=1 and a data field of 1, or by setting
           the A flag and using an ALTA value of 0.

        LB,B is the second parameter of the elliptic curve equation.
           When P>=5, the flag B = 1 indicates B should be negated (mod
           P).  When P=2 or 3, the flag B selects an alternate curve
           equation.

        LC,C is the third parameter of the elliptic curve equation,
           present only when P=2 (indicated by flag M=0) and flag B=1.

        LG,G defines a point on the curve, of order Q.  The W-coordinate
           of the curve point is given explicitly; the Z-coordinate is
           implicit.

        LY,Y is the user's public signing key, another curve point of
           order Q.  The W-coordinate is given explicitly; the Z-
           coordinate is implicit.  The LY,Y parameter pair is always
           present.



3. The Elliptic Curve Equation

   (The coordinates of an elliptic curve point are named W,Z instead of
   the more usual X,Y to avoid confusion with the Y parameter of the
   signing key.)

   The elliptic curve equation is determined by the flag octet, together
   with information about the prime P.  The primes 2 and 3 are special;
   all other primes are treated identically.

   If M=1, the (mod P) or GF[P^D] case, the curve equation is Z^2 = W^3
   + A*W + B.  Z,W,A,B are all numbers (mod P) or elements of GF[P^D].
   If A and/or B is negative (i.e., in the range from P/2 to P), and
   P>=5, space may be saved by putting the sign bit(s) in the A and B
   bits of the flags octet, and the magnitude(s) in the parameter
   fields.

   If M=1 and P=3, the B flag has a different meaning: it specifies an
   alternate curve equation, Z^2 = W^3 + A*W^2 + B.  The middle term of
   the right-hand-side is different.  When P=3, this equation is more


R. Schroeppel, et al                                            [Page 9]


INTERNET-DRAFT                                       ECC Keys in the DNS


   commonly used.

   If M=0, the GF[2^N] case, the curve equation is Z^2 + W*Z = W^3 +
   A*W^2 + B.  Z,W,A,B are all elements of the field GF[2^N].  The A
   parameter can often be 0 or 1, or be chosen as a single-1-bit value.
   The flag B is used to select an alternate curve equation, Z^2 + C*Z =
   W^3 + A*W + B.  This is the only time that the C parameter is used.



4. How do I Compute Q, G, and Y?

   The number of points on the curve is the number of solutions to the
   curve equation, + 1 (for the "point at infinity").  The prime Q must
   divide the number of points.  Usually the curve is chosen first, then
   the number of points is determined with Schoof's algorithm.  This
   number is factored, and if it has a large prime divisor, that number
   is taken as Q.

   G must be a point of order Q on the curve, satisfying the equation

        Q * G  =  the point at infinity (on the elliptic curve)

   G may be chosen by selecting a random [RFC 1750] curve point, and
   multiplying it by (number-of-points-on-curve/Q).  G must not itself
   be the "point at infinity"; in this astronomically unlikely event, a
   new random curve point is recalculated.

   G is specified by giving its W-coordinate.  The Z-coordinate is
   calculated from the curve equation.  In general, there will be two
   possible Z values.  The rule is to choose the "positive" value.

   In the (mod P) case, the two possible Z values sum to P.  The smaller
   value is less than P/2; it is used in subsequent calculations.  In
   GF[P^D] fields, the highest-degree non-zero coefficient of the field
   element Z is used; it is chosen to be less than P/2.

   In the GF[2^N] case, the two possible Z values xor to W (or to the
   parameter C with the alternate curve equation).  The numerically
   smaller Z value (the one which does not contain the highest-order 1
   bit of W (or C)) is used in subsequent calculations.

   Y is specified by giving the W-coordinate of the user's public
   signature key.  The Z-coordinate value is determined from the curve
   equation.  As with G, there are two possible Z values; the same rule
   is followed for choosing which Z to use.






R. Schroeppel, et al                                           [Page 10]


INTERNET-DRAFT                                       ECC Keys in the DNS


   During the key generation process, a random [RFC 1750] number X must
   be generated such that 1 <= X <= Q-1.  X is the private key and is
   used in the final step of public key generation where Y is computed
   as

        Y = X * G (as points on the elliptic curve)

   If the Z-coordinate of the computed point Y is wrong (i.e., Z > P/2
   in the (mod P) case, or the high-order non-zero coefficient of Z >
   P/2 in the GF[P^D] case, or Z sharing a high bit with W(C) in the
   GF[2^N] case), then X must be replaced with Q-X.  This will
   correspond to the correct Z-coordinate.



5. Elliptic Curve SIG Resource Records

   The signature portion of an RR RDATA area when using the EC
   algorithm, for example in the RRSIG and SIG [RFC records] RRs is
   shown below.

                       1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3
   0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                   R, (length determined from LQ)           .../
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                   S, (length determined from LQ)           .../
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

   R and S are integers (mod Q).  Their length is specified by the LQ
   field of the corresponding KEY RR and can also be calculated from the
   SIG RR's RDLENGTH. They are right justified, high-order-octet first.
   The same conditional formula for calculating the length from LQ is
   used as for all the other length fields above.

   The data signed is determined as specified in [RFC 2535].  Then the