Professional Documents
Culture Documents
PDF Communications in Interference Limited Networks 1St Edition Wolfgang Utschick Eds Ebook Full Chapter
PDF Communications in Interference Limited Networks 1St Edition Wolfgang Utschick Eds Ebook Full Chapter
PDF Communications in Interference Limited Networks 1St Edition Wolfgang Utschick Eds Ebook Full Chapter
https://textbookfull.com/product/networks-and-communications-
netcom2013-proceedings-of-the-fifth-international-conference-on-
networks-communications-1st-edition-natarajan-meghanathan/
https://textbookfull.com/product/full-duplex-communications-and-
networks-1st-edition-lingyang-song/
https://textbookfull.com/product/security-aware-device-to-device-
communications-underlaying-cellular-networks-1st-edition-aiqing-
zhang/
https://textbookfull.com/product/intelligent-vehicular-networks-
and-communications-fundamentals-architectures-and-solutions-1st-
edition-anand-paul/
LTE communications and networks : femtocells and
antenna design challenges First Edition Safdar
https://textbookfull.com/product/lte-communications-and-networks-
femtocells-and-antenna-design-challenges-first-edition-safdar/
https://textbookfull.com/product/machine-learning-and-cognitive-
computing-for-mobile-communications-and-wireless-networks-1st-
edition-krishna-kant-singh/
https://textbookfull.com/product/interference-analysis-modelling-
radio-systems-for-spectrum-management-1st-edition-john-pahl/
https://textbookfull.com/product/big-data-computing-and-
communications-second-international-conference-
bigcom-2016-shenyang-china-july-29-31-2016-proceedings-1st-
edition-yu-wang/
Communications
in Interference
Limited
Networks
Signals and Communication Technology
More information about this series at http://www.springer.com/series/4748
Wolfgang Utschick
Editor
Communications
in Interference Limited
Networks
123
Editor
Wolfgang Utschick
Technische Universität München
Munich
Germany
vii
viii Preface
ix
x Contents
1.1 Introduction
Y = AX,
where the matrix A ∈ Fqñ×n describes the transformation of the network. If nodes
perform a random linear combination of incoming packets during one shot, random
additive errors within the network can be described as
Y = AX + BZ,
where B ∈ Fqñ× represents the linear transformation of the network and the rows
of Z ∈ Fq×(n+m) are the error vectors z0 , . . . , zñ−1 of the corresponding edge of the
network with elements from Fq . E.g, zi = 0 means that on the edge corresponding
to the ith row of B there did not occur any error.
Note that the only property which is not changed by a transmission over an
error-free random linear network coding channel is the row space of the transmit-
ted matrix X. Therefore, information is encoded into subspaces instead of specific
matrices.
In [36], it is shown that subspace codes and rank-metric codes are closely related.
In [36], a construction is given where Gabidulin codes are used for error control with
respect to errors and erasures. We transmit a codeword
X = In M ,
Y = AX + Eout ,
is received, where Eout is the error matrix. Now, let S be a matrix which brings Y to
echelon form. Then, we have
GR
SY = .
0 C
The matrix R has r rows such that C consists of ñ − r rows. The paper [36] extended
R to n rows by inserting specific zero rows to obtain R̂ and showed that R̂ can be
decomposed into
R̂ = M + LM + DC + Er est
= M + Er ow + Ecol + Er est ,
While in [5] network coding with error correction is studied for the operator channel,
in [16] the benefits of network coding over routing in random networks are discussed.
In [35], codes for network coding have been introduced, which are based on rank-
metric codes, in particular Gabidulin codes. Since these codes are not commonly
known, we will briefly describe them in the following.
The notion rank norm outside of coding theory is known (implicitly) in linear
algebra by the well known inequality
Explicitly, the rank norm was introduced by Hua [17]. Later, it was introduced to
coding theory in [6, 11, 29].
For a given basis of Fq m over Fq , there exists a vector space isomorphism which
maps each vector x ∈ Fqn m to a matrix X ∈ Fqm×n . Let rk(x) denote the rank of X over
Fq and let Rq (X), Cq (X) denote the row and column space of X in Fqn . We use the
notation as vector (e.g. from Fqn m ) or matrix (e.g. from Fqm×n ) equivalently, whatever
is more convenient. The rank distance of two vectors x, y ∈ Fq m is defined by
dR (x, y) = rk(x − y)
The codes introduced in [11] have length n, dimension k, minimum rank distance
d = n − k + 1, and are nowadays called Gabidulin codes. Gabidulin codes are
constructed over an extension field Fq m of Fq . A generator matrix is given by:
⎡ ⎤
g1 g2 · · · gn
⎢ g [1] g2[1] · · · gn[1] ⎥
⎢ 1 ⎥
⎢ [2] ⎥
G=⎢
⎢ g1 g2[2] · · · gn[2] ⎥,
⎥ (1.2)
⎢ .. .. . . .. ⎥
⎣. . .. ⎦
g1[k−1] g2[k−1]
· · · gn[k−1]
GH = 0.
Gabidulin codes are the rank-metric analogs of Reed-Solomon codes and found
many applications including network coding. Interleaving or the direct sum of
Gabidulin codes allows both decreasing the redundancy and increasing the error cor-
recting capability for network coding. For Gabidulin codes, we proposed a transform-
domain algorithm correcting both errors and erasures [32]. We showed how to gener-
alize this algorithm for interleaved Gabidulin codes. The transform-domain approach
allows to simplify derivations and proofs and also simplifies finding the error vector
after solving the key equation. Details on these results can be found in [33].
The classical definition of Gabidulin codes requires the concept of linearized
polynomials. A linearized polynomial (or q-polynomial) over Fq m is a polynomial
of the form
t
f (x) = f i x [i]
i=0
The symbolic product is distributive and associative, but noncommutative. The set
of linearized polynomials over Fq m with the operation of polynomial addition and
symbolic multiplication forms a noncommutative ring.
There exist several efficient algorithms for decoding Gabidulin codes. A syndrome-
based bounded minimum distance (BMD) decoding algorithm based on solving
a key equation was introduced by Gabidulin [11]. The first step of the decoding
process is to compute the syndrome. This can be done by multiplying the received
word r with the transposed of the parity check matrix H. We use the syndrome
values s = (s0 , . . . , sn−k+1 ) := r · HT to calculate the syndrome polynomial s(x) =
n−k−1
i=0 si x i . The key equation which has to be solved is given by
Ω(x) ≡ Λ(s(x)) = Λ(x) ◦ s(x) mod x [n−k] , deg Ω(x) < deg Λ(x).
This key equation can be solved for the error span polynomial Λ(x). The direct way
to accomplish this task is to solve a linear system of equations based on the key
equation:
i
[ j]
t
[ j]
Ωi = Λ j si− j = Λ j si− j = 0, ∀i ∈ [t, n − k − 1 − 1].
j=0 j=0
and degq r (x) < degq b(x). The existence of such an algorithm is ensured by the
fact that the ring of linearized polynomials with addition + and multiplication ◦ is a
Euclidean domain.
Algorithm 1:
rout (x); u out (x); vout (x) ← RightLEEA a(x); b(x); dstop
Input: a(x); b(x) ∈ Lq m [x] with degq a(x) ≥ degq b(x);
stopping degree dstop
Initialize: i ← 1,
r (−1) (x) ← a(x), r (0) (x) ← b(x),
u (−1) (x) ← 0, u (0) (x) ← x [0] ,
v (−1) (x) ← x [0] , v (0) (x) ← 0
1 while degq r (i−1) (x) ≥ dstop do
2 q (i) (x); r (i) (x) ← RightDiv r (i−1) (x); r (i−2) (x) u (i) (x) ← u (i−2) (x) − q (i) (u (i−1) (x))
v (i) (x) ← v (i−2) (x) − q (i) (v (i−1) (x)) i ← i + 1
Output: rout (x) ← r (i−1) (x); u out (x) ← u (i−1) (x); vout (x) ← v (i−1) (x)
Therefore, we use the syndrome polynomial s(x) and x [n−k] as input to the LEEA
which in turn gives us Λ(x). Next, we need to find a basis of the root space of
Λ(x). This is relatively easy due to the structure of linearized polynomials. This
basis can be used for determining the error. Therefore it is necessary to solve a
system of equations. The overall complexity is in the order O(n 2 ) over Fq m . We
can further reduce the complexity by applying a Gao-like decoding approach [47].
This algorithm uses an alternative transformed key equation which can be solved by
applying the LEEA and outputs directly the linearized evaluation polynomial of the
estimated codeword.
Another algorithm that can be applied for solving the key equation efficiently is a gen-
eralization of the Berlekamp-Massey algorithm [31]. A problem which is equivalent
to the problem of solving the key equation is the problem of linearized shift-register
synthesis: Let s = s1 , s2 , . . . , s N be a sequence over a field Fq m . Find the smallest
integer l ≥ 0 for which there is a vector σ = (σ1 , . . . , σl ) over Fq m such that
l
qi
sn = − σi sn−i f or n = l + 1, . . . , N
i=1
8 M. Bossert et al.
and find a suitable vector σ which is called the connection vector. Given a connection
vector σ = (σ1 , . . . , σl ) over Fq m , the corresponding connection q-polynomial is
defined by
l
σ (x) = σi x [i] , (1.5)
i=0
Δ
where σ0 = 1 and degq σ (x) ≤ l. A linearized-feedback shift-register is com-
pletely determined by length l and connection vector σ , hence it can be denoted by
(l, σ ). Given a sequence s and a shift-register (l, σ ), let the discrepancy dn (l, σ ) of
a sequence element sn be defined by
⎧
⎨0 for n = 1, . . . , l,
dn (l, σ ) = l
[i] (1.6)
⎩ σi sn−i for n = l + 1, . . . , N .
i=0
Algorithm 2:
l; σ (x); σ (x); n ← Richter–Plass(s)
Input: s = s1 , . . . , s N
Initialize: l ← 0, σ (x) ← x
σ (x) ← x, n ← 0, d ← 1
3 for each n from 1 to N do
[ j]
4 d ← lj=0 σ j sn− j if d = 0 then
5 if 2l ≥ n then
[n−n ]
6 σ (x) ← σ (x) − d dx ⊗ σ (x)
7 else
x [n−n ]
8 σ̃ (x) ← σ (x) σ (x) ← σ (x) − d d ⊗ σ (x) l ← n − l σ (x) ← σ̃ (x),
n ← n, d ← d
elements of s. Within the loop, the discrepancy dn (l, σ ) is computed using the cur-
rent shift-register (l, σ ). The discrepancy is zero if and only if (l, σ ) is able to generate
the element sn . In this case there is no need to change the shift-register. Hence, the
algorithm directly continues with the processing of the next element. If the discrep-
ancy is not equal to zero, the shift-register (l, σ ) is modified in order to obtain the
shortest shift-register (l,ˆ σ̂ ) which is able to generate sn and all previous elements.
The algorithm outputs both, length l and a connection polynomial σ .
The time complexity of this algorithm may be estimated by O(l N ) ≤ O(N 2 )
operations in Fq m since the algorithm has to process N elements of the sequence s
and the processing of each element requires at most O(l) ≤ O(N ) operations in Fq m .
In addition, the uniqueness of the solution can be checked. For decoding Gabidulin
codes up to half the code distance, a decoding failure can be declared when the
solution is not unique. Furthermore, it is possible to calculate all solutions which is
interesting for decoding Gabidulin codes beyond half the code distance.
1.2.3 Remarks
Fast decoding up to half the minimum rank distance of Gabidulin codes by applying a
Gao-like algorithm is explained in [47]. Sidorenko et al. [31] introduces the general-
ization of the Berlekamp-Massey algorithm and proves its correctness. Furthermore
it is explained how to check uniqueness of the solution and how to find all possible
solutions.
Interleaving or the direct sum of Gabidulin codes with rank distance d can be
applied for random network coding. For codes over Fq m , a fast decoding algorithm
exists which corrects errors of rank up to +1 (d − 1) with high probability. Thus,
with interleaving the error correcting radius can be enlarged. The algorithm is based
on fast skew-feedback shift-register synthesis and for fixed it has time complexity
O(m 3 log m) operations in the base field Fq .
In network coding, rows of a matrix C ∈ Fqm×n (with a prefix of length m) are
sent via the network as packets. In many practical cases the packet length is much
more than the number m of packets in the message, so we would like to have n m
in this case. In order to increase the parameter n by a factor of ∈ N, the direct sum
of codewords of the code C can be used. Given a matrix code C ⊆ Fqm×n and a
number , the -interleaved code IC ⊆ Fqm×n consists of all m × n matrices C over
Fq having the following block form
C = C(1) C(2) . . . C() , C() ∈ C, (1.7)
10 M. Bossert et al.
where l = 1, . . . , . The rate and the distance of the interleaved code are the same
as for the original code C.
We assume m = n for the interleaved Gabidulin code. Note, that Loidreau and
Overbeck in [24, 25] used interleaving of transposed Gabidulin codes.
We describe a decoder for any -interleaved Gabidulin code IC ⊆ Fqm×n having
distance d. This decoder guarantees to correct all error words of rank less than d/2
and it corrects with high probability all error words of rank t if
t≤ (d − 1).
+1
The probability Pf (t) of a decoding failure can be upper bounded by Pf (t) < 4/q m
for the cases that words of fixed rank are equiprobable. This probability is very small
in practice and coincides with the bound from [24, 25] and Pf (t) practically vanishes
if t < +1 (d − 1).
Here c = c(1) . . . c() ∈ Fqnm denotes the concatenation of vectors c() . The
codeword c ∈ IG(q m , ; n, k) is transmitted and a vector r = r(1) . . . r() ∈ Fqnm
is received. The error on the channel is e = e(1) . . . e() ∈ Fqnm , where r = c + e.
Let rk(e) = t then we can represent the error vector as
where
n
f i(l) = Bi,(l)j h j , i = 1, . . . , t. (1.12)
j=1
t
σ (x) = σi x [i] (1.14)
i=0
1.3.2 Remarks
Details on these results can be found in [32]. Further details and bounds can be
found in [30]. Further, in [46], an interpolation-based unique decoder as well as a
list decoder for interleaved Gabidulin codes were presented.
Definition 1.4.1 (Convolutional Code [40]) A rate k/n convolutional code C over
F with memory m is defined by its k × n generator matrix in polynomial form:
⎛ ⎞
g11 (D) g12 (D) . . . g1n (D)
⎜ .. ⎟ ,
G(D) = ⎝ ... ..
.
..
. . ⎠
gk1 (D) gk2 (D) . . . gkn (D)
def
νi = max {deg gi j (D)}.
1≤ j≤n
def
m = max {νi }.
1≤i≤k
The overall constraint length ν is defined as the sum of the constraint lengths νi :
def
k
ν = νi .
i=1
(P)UM codes are a special class of convolutional codes with memory m = 1 [21,
22], i.e., the semi–infinite generator matrix G is given by:
⎛ ⎞
G0 G1
⎜ G0 G1 ⎟
G=⎝ ⎠, (1.16)
.. ..
. .
where G0 and G1 are k × n matrices. For an (n, k) UM code, both matrices have full
rank. For an (n, k | k1 ) PUM code, rk(G0 ) = k and rk(G1 ) = k1 < k:
G00 G10
G0 = , G1 = ,
G01 0
where G00 and G10 are k1 × n matrices and G01 is a (k − k1 ) × n–matrix. For both
cases, we have the following encoding rule:
c j = u j · G0 + u j−1 · G1 .
Note that the overall constraint length for UM codes is ν = k and for PUM codes
ν = k1 , since this is the number of symbol which influence the next block.
There are restrictions on the code rate of (P)UM codes when a certain number of
full–rank submatrices Hi should exist. This full–rank condition, rk(Hi ) = n − k for
all i = 0, . . . , m H is used in the following construction.
Lemma 1.4.1 (Rate Restriction for UM Codes [40]) An (n, k) UM code with overall
constraint length ν = k has rate
(n − k) · m H
R=
(n − k) · (m H + 1)
k mH
R= > ,
n mH + 1
1 Coding Techniques for Transmitting Packets … 15
If we use the parity–check matrix to construct (P)UM codes, the following theorem
guarantees that there is always a corresponding generator matrix that defines a (P)UM
code.
Definition 1.4.3 (Sum Rank Weight [40]) Let a vector v ∈ Fn be given and let it be
decomposed into subvectors:
v = v0 v1 . . . v−1 ,
with vi ∈ Fn for all i. We define the sum rank weight wtr k (v) as the sum of the rank
norms of the subvectors:
−1
def
wtrk (v) = rk(vi ), (1.17)
i=0
for 0 ≤ ≤ ∞.
Hence, we define the sum rank distance between two sequences v(1) , v(2) of length
n by
−1
d(v(1) , v(2) ) = wt rk (v(1) − v(2) ) = dR (vi(1) , vi(2) ).
def
(1.18)
i=0
Let C r () denote the set of all codewords c() corresponding to paths in the minimal
code trellis which diverge from the zero state at depth j and return to the zero state
for the first time after + 1 branches at depth j + + 1. W.l.o.g., we assume j = 0
as we only consider time–invariant convolutional codes.
The extended row rank distance of order is defined as the minimum sum rank
weight of all codewords in C r ().
Definition 1.4.5 (Extended Row Rank Distance [40]) The extended row rank dis-
tance of order = 1, 2, . . . is defined as
def
dr = min wt rk (c) . (1.20)
r
c∈C ()
The minimum of the th order extended row rank distances gives the free rank
distance:
dfree = min dr .
As for Hamming metric, the extended row rank distance dr can be lower bounded by
a linear function dr ≥ max{α + β, dfree } where β ≤ dfree and α denotes the slope
(the average linear increase). The slope is an important parameter for determining
the error–correcting capability and is defined as follows (analog to [8]).
Definition 1.4.6 (Slope [40]) The average linear increase of the extended row rank
distance (slope) is defined as r
def d
α = lim . (1.21)
→∞
dfree ≤ dfree
H
, and dr ≤ dH,r , = 1, 2, . . . .
For the rank norm rk(vi ) and the Hamming norm wt H (vi ) of a vector vi it holds that:
rk(vi ) ≤ wt H (vi ) and hence also
−1
rk(vi ) ≤ wt H (v0 . . . v−1 ).
i=0
1 Coding Techniques for Transmitting Packets … 17
Consequently, the upper bounds for the free distance and the slope of (P)UM
codes based on Hamming metric [27, 39] also hold for (P)UM codes based on the
sum rank metric.
Corollary 1.4.1 (Upper Bounds [40]) For an (n, k) UM code, where ν = k, the free
rank distance is upper bounded by:
dfree ≤ 2n − k + 1. (1.22)
For an (n, k | k1 ) PUM code, where ν = k1 < k, the free rank distance is upper
bounded by:
dfree ≤ n − k + ν + 1. (1.23)
For both UM and PUM codes, the average linear increase (slope) is upper boun-
ded by:
α ≤ n − k. (1.24)
1.4.5 Construction
for all i = 1, . . . , m H . G (c) is an (n (c) , k (c) ) Gabidulin code and G (ri ) is an (n (ri ) , k (ri ) )
Gabidulin code with
Hence, not only each submatrix has to define a Gabidulin code, but also the rows
and columns of submatrices of H.
Now, we give an explicit construction that fulfills the requirements of Defini-
tion 1.4.7. To ensure that (1.25) is fulfilledand H(c) defines a Gabidulin code, H1
has to be the continuation of H0 , i.e., h(1) = (h (0)
1 )
[n−k]
, (h (0)
2 )
[n−k]
, . . . , (h (0)
n )
[n−k]
.
(2) (1)
Also, h has to be the continuation of h and so on. Hence,
H(c) = V(m H +1)(n−k) h(0)
= V(m H +1)(n−k) h (0) (0) (0) .
1 h2 . . . hn (1.27)
In order to fulfill (1.26), we have to ensure that all elements from F in the set
This h(0) is used to define H(c) (1.27) and hence also H is defined (1.25).
To make sure that also (1.26), (1.28) are fulfilled, we require a certain minimal
field size. If (n − k) divides n, then h(0) can be divided into subvectors, each of length
(n − k) (1.29) and the field size has to fulfill s ≥ (m H + 1) · n to ensure that all
elements in H are linearly independent (1.28) and that (1.26) is fulfilled. If (n − k)
does not divide n, the last subvector in h(0) is shorter than n − k. Equations (1.28)
and (1.26) can be
⎛ ⎞
a [0] a [1] a [4] a [5] a [8] a [9]
⎜ a [1] a [2] a [5] a [6] a [9] a [10] ⎟
⎛ ⎞ ⎜ ⎟
H0 ⎜ ⎟
⎜ ⎟
⎜ H1 H0 ⎟ ⎜ a [2] a [3] a [6] a [7] a [10] a [11] a [0] a [1] a [4] a [5] a [8] a [9] ⎟
⎜ ⎟ ⎜ ⎟
H=⎜ . ⎟ = ⎜ a [3] a [4] a [7] a [8] a [11] a [12] a [1] a [2] a [5] a [6] a [9] a [10] ⎟.
⎜
⎝ H1 . . ⎟ ⎜
⎠ ⎜
⎟
⎟
.. ⎜ ⎟
. ⎜ a [2] a [3] a [6] a [7] a [10] a [11] ...⎟
⎜ ⎟
⎝ a [3] a [4] a [7] a [8] a [11] a [12] ...⎠
(1.30)
Another random document with
no related content on Scribd:
as was the case years ago, an unhealthy excitement in the minds of
220
the people.”
But, facts being more convincing than official denials, the exodus
grew more alarming, because the forces to which it owed its origin
continued in operation. The “Jewish Colonization Association” now
came to the aid of the indigent exiles, and endeavoured to save
them from additional suffering by preventing those who were not
provided with the necessary passage money, or were not physically
221
fit, from leaving their homes. These wise measures restrained to
a certain extent indiscriminate expatriation, but, as might have been
foreseen, failed to check it entirely. The exodus continued, and the
outcry against Roumania spread, for now the countries into which
the undesirable current flowed were compelled by self-interest to do
what they had hitherto vainly attempted to effect from a sense of
philanthropy.
America, the favourite haven of refuge for the fortune-seeker of
every colour and clime, undertook the task of spokesman. The late
Mr. Hay, Secretary of State, in September, 1902, through the
representatives of the United States in the countries which took part
in the Congress of Berlin, reminded the Governments of those
countries of Art. 44 of the Treaty signed by them in 1878, urging
them to bring home to Roumania her flagrant and persistent failure
to fulfil the conditions on which she had obtained her independence.
After a handsome tribute to the intellectual and moral qualities of the
Jew, based on history and experience, the American Minister
protested, on behalf of his country, against “the treatment to which
the Jews of Roumania are subjected, not alone because it has
unimpeachable ground to remonstrate against resultant injury to
itself, but in the name of humanity.” He concluded with a vigorous
appeal to “the principles of International Law and eternal justice,”
and with an offer to lend the moral support of the United States to
222
any effort made to enforce respect for the Treaty of Berlin.
This powerful impeachment, coming as it did from a distant party
in no way connected with the affairs of Continental Europe, may
have caused heart-searchings in nearer and more immediately
concerned countries; but it failed to awaken those countries to a
proper sense of their interests, not to say duties. The only quarter in
which America’s appeal to humanity found an echo was England. A
number of representative men, such as the late Archbishop of
Canterbury, the present Bishop of London, Lord Kelvin, the
Marquess of Ripon, the late Mr. Lecky, Sir Charles Dilke, the Master
of Balliol, and others, publicly expressed their profound sympathy
with the victims of persecution. Mr. Chamberlain also seized the
opportunity of declaring that, as history proves, the Jews, “while
preserving with extraordinary tenacity their national characteristics
and the tenets of their religion, have been amongst the most loyal
subjects of the states in which they have found a home, and the
impolicy of persecution in such a case is almost greater than its
223
cruelty.” Other Englishmen also joined in the denunciation of
Roumania not so much from pity for the victims of oppression as
from fear lest, unless the Roumanian Government was compelled to
change its policy, England should have to face another inroad of
“undesirable” Jewish immigrants.
In like manner, the only Government which volunteered to
second Mr. Hay’s Note was the British, and on the common basis of
these two representations, the signatory Powers of the Treaty of
Berlin “exchanged views.” The results of this exchange can be
summed up only too easily. The historian of the future will probably
derive therefrom some interesting lessons regarding European
politics and ethics in the beginning of the twentieth century. They are
as follows:
Germany, under whose presidency the stipulation concerning the
Jews of Roumania was framed, did not choose to consider herself
called upon to insist on the execution of that stipulation. The Liberal
section of the German press received the American Note with
sincere, but ineffectual, appreciation; while of the Conservative
majority some pronounced it naïve, and others affected to regard it
as an attempt on America’s part to interfere in European affairs, or
even as an electioneering trick having for its sole object to enhance
President Roosevelt’s political prestige! The German Government,
though more courteous than the German press, proved equally cold.
As we have already seen, that Government was the last to join in the
efforts to improve the lot of the Roumanian Jews and the first to
declare itself satisfied with the deceptive revision of Article 7 of the
Roumanian Constitution. This attitude, when considered in
conjunction with the fact that a Hohenzollern reigns in Roumania,
and with that kingdom’s place in the present political combinations of
the Continent, enables us to understand, if not to applaud,
Germany’s reception of Mr. Hay’s Note.
Austria-Hungary, whose proximity to Roumania pointed her out
as the Power primarily concerned, and entitled to act, declined to
take any steps singly or collectively. The self-restraint of Austria, like
that of Germany, and even in a greater degree, was dictated by
political considerations, Roumania being practically the only State in
the Balkans, where the influence of Austria-Hungary and of the Triple
Alliance still counts for something. Besides, the Vienna Cabinet
could not decently join in advocating Jewish emancipation, for it was
Austria which in May, 1887, concluded with Roumania a treaty
whereby some seventy thousand Jewish residents in the latter
kingdom—who, according to a practice common in Mohammedan
countries, had enjoyed Austrian protection while Roumania was
under Ottoman rule—were deprived of the status of Austrian
subjects, without receiving any other status in exchange.
Italy was deterred from lending her support to the American Note
by Roumania’s relations with the Triple Alliance and also by the
vogue which the “Roman” idea obtains in the land which the
Roumanians are pleased to regard as “the cradle of their race.”
Russia, whose treatment of her own Jewish subjects would have
made an appeal to “humanity and eternal justice” on behalf of the
Jews in another country a sad mockery, decorously refrained from
supporting the American Note. It is true that the Russian press
imitated the Teutonic in scoffing at America’s action as a pretext for
gaining admission to the counsels of the European Areopagus, and
in condemning it as an impertinence! But the Czar’s Government,
with better taste, extricated itself from an awkward position by basing
its refusal on the ground that the grievances set forth in Mr. Hay’s
despatch were so old that it was hardly worth while troubling about
them. In the opinion of the Russian Ministers, the Jews must by now
be thoroughly accustomed to starvation.
France, with all the good intentions in the world, could do nothing
without Russia’s consent and, therefore, contented herself with the
expression of a modest hope that the Roumanian Government might
of their own accord decide to fulfil their obligations, seeing that the
real sufferer is Roumania itself, and with pointing to the lack of
224
means of enforcing such fulfilment.
In brief, the European Powers considered that they did their duty
by expressing their platonic concurrence with that part of the
American Note which referred to the obligations of humanity and
civilisation generally. But to the more definite appeal to the Treaty of
Berlin they refused to pay any attention whatsoever. Nor can we
wonder at their refusal. The appeal was not a very happy one; for
every party to that contract has conscientiously broken it in turn.
Russia, in defiance of its provisions, has fortified Batoum; Turkey has
not even attempted to carry out the reforms in the European
Provinces of the Empire, ordained by the Treaty; Great Britain has
done nothing for the Armenians. Why then should poor Roumania
alone be called upon to carry out her share of an agreement, already
disregarded with impunity by everyone else concerned?
Such a retort would, of course, have been too candid and too
rational for diplomacy. Instead, the Roumanian Government had
again recourse to the more correct, if somewhat hackneyed,
expedient of an official contradiction of the truth. The Roumanian
Minister in London declared that “the idea that any persecution
existed was absolutely erroneous.” The Jews were foreigners, and
“the disabilities imposed upon foreigners were absolutely necessary
for the protection of his countrymen, who had bought their
independence with the sword, and had a right to manage their
225
economic affairs according to their requirements, etc., etc.” What
the Roumanian conception of such a right is has been very
eloquently explained by Roumania’s accomplished Queen. After
having drawn a pitiful and, although exaggerated, in the main faithful
picture of Roumania’s economic misery, Her Majesty declares that,
under such conditions, the civilised world ought not “to require her to
harbour and support others, when she herself stands in dire need of
assistance.” Those “others” are “foreigners,” that is, Roumanian
Jews; their exodus is represented as the voluntary emigration of “a
foreign population” due to the instinct which prompts a rat to quit a
sinking ship, and their departure is welcome, because they, being
traders, drain the country of its wealth. This interesting economic
doctrine is expounded by Her Majesty as follows: “It is a fact that no
money has ever been introduced into Roumania through any one in
trade. Any that such a man may possess goes abroad, first to
purchase his stock and outfit, and later for supplies to carry on his
business, even such articles as buttons and the commonest kinds of
braids not being manufactured here except on the very smallest
226
scale.” Here again the Jewish apologist is more convincing than
his Roumanian accuser. Admitting that, on the whole, the Queen’s
statements are correct, he asks: “But why is it so? For the reason
that the ruling class prohibits ‘foreigners’ to acquire lands in the
country, and by means of this and other laws keeps foreign capital
227
from coming in.”
Protests pass away, grievances remain. The well-meant action of
Mr. Hay and Lord Lansdowne, far from bettering, really aggravated
the condition of the people on whose behalf it was taken. The
Roumanian politicians, with characteristic astuteness, perceived that
the immediate cause of the complaint was the emigration of the
Jews to the United States, England and Canada, and, naturally
enough, arrived at the conclusion that the one thing needful was to
remove the ground of complaint by stopping emigration. A
telegraphic order was sent to all the local authorities, forbidding the
issue of passports to the Jews. Those who had already reached the
frontier were forcibly turned back, and hundreds of others, who had
sold all they possessed in order to raise the funds necessary for the
228
journey, were compelled to return home and perish. Thus an act
intended as a blessing proved an unmitigated curse, and modern
Roumania by this new measure has outstripped even mediaeval
Spain in cruelty. For the Spanish sovereigns, blinded by religious
bigotry, had yet given to the Jews the alternatives of conversion or
exile. Their Roumanian imitators, infatuated by racial fanaticism, will
not baptize the Jews, nor dare they banish them; but, like Pharaoh of
old, they virtually bid them stay and be slaves.
CHAPTER XXIII
ANTI-SEMITISM