Download and Read online, DOWNLOAD EBOOK, [PDF EBOOK EPUB ], Ebooks

download, Read Ebook EPUB/KINDE, Download Book Format PDF

Structural Information and Communication

Complexity 20th International Colloquium SIROCCO
2013 Ischia Italy July 1 3 2013 Revised Selected
Papers 1st Edition Andrea Clementi

OR CLICK LINK
https://textbookfull.com/product/structural-
information-and-communication-complexity-20th-
international-colloquium-sirocco-2013-ischia-
italy-july-1-3-2013-revised-selected-papers-1st-
edition-andrea-clementi/
Read with Our Free App Audiobook Free Format PFD EBook, Ebooks dowload PDF
with Andible trial, Real book, online, KINDLE , Download[PDF] and Read and Read
Read book Format PDF Ebook, Dowload online, Read book Format PDF Ebook,
[PDF] and Real ONLINE Dowload [PDF] and Real ONLINE
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Structural Information and Communication Complexity

27th International Colloquium SIROCCO 2020 Paderborn
Germany June 29 July 1 2020 Proceedings Andrea Werneck
Richa
https://textbookfull.com/product/structural-information-and-
communication-complexity-27th-international-colloquium-
sirocco-2020-paderborn-germany-june-29-july-1-2020-proceedings-
andrea-werneck-richa/

Structural Information and Communication Complexity:

25th International Colloquium, SIROCCO 2018, Ma'ale
HaHamisha, Israel, June 18-21, 2018, Revised Selected
Papers Zvi Lotker
https://textbookfull.com/product/structural-information-and-
communication-complexity-25th-international-colloquium-
sirocco-2018-maale-hahamisha-israel-june-18-21-2018-revised-
selected-papers-zvi-lotker/

Brain Inspired Computing International Workshop

BrainComp 2013 Cetraro Italy July 8 11 2013 Revised
Selected Papers 1st Edition Lucio Grandinetti

https://textbookfull.com/product/brain-inspired-computing-
international-workshop-braincomp-2013-cetraro-italy-
july-8-11-2013-revised-selected-papers-1st-edition-lucio-
grandinetti/

Communication Technologies Information Security and

Sustainable Development Third International Multi topic
Conference IMTIC 2013 Jamshoro Pakistan December 18 20
2013 Revised Selected Papers 1st Edition Faisal Karim
Shaikh
https://textbookfull.com/product/communication-technologies-
information-security-and-sustainable-development-third-
international-multi-topic-conference-imtic-2013-jamshoro-
Distributed Computer and Communication Networks 17th
International Conference DCCN 2013 Moscow Russia
October 7 10 2013 Revised Selected Papers 1st Edition
Vladimir Vishnevsky
https://textbookfull.com/product/distributed-computer-and-
communication-networks-17th-international-conference-
dccn-2013-moscow-russia-october-7-10-2013-revised-selected-
papers-1st-edition-vladimir-vishnevsky/

Hybrid Systems Biology Second International Workshop

HSB 2013 Taormina Italy September 2 2013 and Third
International Workshop HSB 2014 Vienna Austria July 23
24 2014 Revised Selected Papers 1st Edition Oded Maler
https://textbookfull.com/product/hybrid-systems-biology-second-
international-workshop-hsb-2013-taormina-italy-
september-2-2013-and-third-international-workshop-
hsb-2014-vienna-austria-july-23-24-2014-revised-selected-
papers-1st-editi/
Information Security Applications 14th International
Workshop WISA 2013 Jeju Island Korea August 19 21 2013
Revised Selected Papers 1st Edition Yongdae Kim

https://textbookfull.com/product/information-security-
applications-14th-international-workshop-wisa-2013-jeju-island-
korea-august-19-21-2013-revised-selected-papers-1st-edition-
yongdae-kim/

Information Security and Cryptology ICISC 2013 16th

International Conference Seoul Korea November 27 29
2013 Revised Selected Papers 1st Edition Hyang-Sook Lee

https://textbookfull.com/product/information-security-and-
cryptology-icisc-2013-16th-international-conference-seoul-korea-
november-27-29-2013-revised-selected-papers-1st-edition-hyang-
sook-lee/

Simulation and Modeling Methodologies Technologies and

Applications International Conference SIMULTECH 2013
Reykjavík Iceland July 29 31 2013 Revised Selected
Papers 1st Edition Mohammad S. Obaidat
https://textbookfull.com/product/simulation-and-modeling-
methodologies-technologies-and-applications-international-
conference-simultech-2013-reykjavik-iceland-
 Thomas Moscibroda
Adele A. Rescigno (Eds.)

Structural Information
LNCS 8179

and Communication
Complexity
20th International Colloquium, SIROCCO 2013
Ischia, Italy, July 2013
Revised Selected Papers

123
Lecture Notes in Computer Science 8179
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Alfred Kobsa
University of California, Irvine, CA, USA
Friedemann Mattern
ETH Zurich, Switzerland
John C. Mitchell
Stanford University, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
Oscar Nierstrasz
University of Bern, Switzerland
C. Pandu Rangan
Indian Institute of Technology, Madras, India
Bernhard Steffen
TU Dortmund University, Germany
Madhu Sudan
Microsoft Research, Cambridge, MA, USA
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbruecken, Germany
Thomas Moscibroda Adele A. Rescigno (Eds.)

Structural Information
and Communication
Complexity
20th International Colloquium, SIROCCO 2013
Ischia, Italy, July 1-3, 2013
Revised Selected Papers

13
Volume Editors
Thomas Moscibroda
Microsoft Research Asia
System Algorithms Research Group
No. 5, Dan Ling Street, 100080 Beijing, China
E-mail: moscitho@microsoft.com
Adele A. Rescigno
Università di Salerno
Dipartimento di Informatica
Via Ponte don Melillo, 84084 Fisciano, SA, Italy
E-mail: arescigno@unisa.it

ISSN 0302-9743 e-ISSN 1611-3349

ISBN 978-3-319-03577-2 e-ISBN 978-3-319-03578-9
DOI 10.1007/978-3-319-03578-9
Springer Cham Heidelberg New York Dordrecht London

Library of Congress Control Number: 2013952950

CR Subject Classification (1998): F.2, C.2, G.2, E.1, C.2.4

LNCS Sublibrary: SL 1 – Theoretical Computer Science and General Issues

© Springer International Publishing Switzerland 2013
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and
executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication
or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location,
in ist current version, and permission for use must always be obtained from Springer. Permissions for use
may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution
under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of publication,
neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or
omissions that may be made. The publisher makes no warranty, express or implied, with respect to the
material contained herein.
Typesetting: Camera-ready by author, data conversion by Scientific Publishing Services, Chennai, India
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
 Preface

The 20th International Colloquium on Structural Information and Communi-

cation Complexity (SIROCCO 2013) took place in Ischia, Italy, for three days
starting July 1, 2013.
SIROCCO is devoted to the study of communication and knowledge in dis-
tributed systems from both qualitative and quantitative viewpoints. Special
emphasis is given to innovative approaches and fundamental understanding in
addition to efforts to optimize current designs. The typical areas include dis-
tributed computing, communication networks, game theory, parallel computing,
social networks, mobile computing (including autonomous robots), peer-to-peer
systems, communication complexity, fault-tolerant graph theories, and random-
ized/ probabilistic issues in networks.
This year, 67 papers were submitted in response to the call for papers, and
each paper was evaluated by at least three reviewers. The Program Committee
selected 28 papers for presentation at the colloquium and publication in this
volume after in-depth discussions. The SIROCCO Prize for Innovation in Dis-
tributed Computing was awarded this year to Andrzej Pelc from the University
of Quebec for his contributions to the understanding of distributed computing. A
laudatio summarizing his many and important innovative achievements appears
in these proceedings.
We further congratulate the recipients of the 2013 SIROCCO Best Student
Paper award. This year, two papers were selected to share the award. Specif-
ically, the 2013 SIROCCO Best Student Paper Award was given to Lakshmi
Anantharamu for her paper “Broadcasting in Ad Hoc Multiple Access Channels”
(with Bogdan Chlebus) and Sebastian Kniesburges and Andreas Koutsopoulos
for their paper “A Deterministic Worst-Case Message Complexity Optimal So-
lution for Resource Discovery” (with Christian Scheideler).
The collaboration of the Program Committee members and the external re-
viewers enabled completion of the process of reviewing the papers and discussing
them in less than four weeks. We thank them all for their devoted service to the
SIROCCO community. We thank the authors of all the submitted papers; with-
out them we could not have prepared a program of such quality. We thank
Gennaro Cordasco for his assistance as publicity chair, and Yang Chen for serv-
ing as the submission chair. We also thank the keynote and invited speakers
Andrea Richa, Cyril Gavoille, and Fabian Kuhn. The preparation of this event
was guided by the SIROCCO Steering Committee, headed by Shay Kutten.
VI Preface

We are indebted to Luisa Gargano and Ugo Vaccaro for their assistance with
local arrangements during the colloqium. We gratefully acknowledge the financial
support of the Dipartimento di Informatica from Università di Salerno.

July 2013 Thomas Moscibroda

Adele A. Rescigno
 Organization

Program Committee
Ittai Abraham MSR Silicon Valley, USA
Chen Avin Ben-Gurion University, Israel
Keren Censor-Hillel Technion, Israel
Wei Chen Microsoft Research Asia
Ajoy Datta University of Nevada, USA
Gabriele Di Stefano University of L’Aquila, Italy
Yuval Emek ETH Zurich, Switzerland
Qiang-Sheng Hua Tsinghua University, China
David Ilcinkas Université Bordeaux, France
Toshimitsu Masuzawa Osaka University, Japan
Friedhelm Meyer Auf der
Heide University of Paderborn, Germany
Thomas Moscibroda Microsoft Research Asia
Calvin Newport Georgetown University, USA
Adele Rescigno Università di Salerno, Italy
Nicola Santoro Carleton University, Canada
Ulrich Schmid TU Vienna, Austria
Siddhartha Sen Princeton, USA
Jukka Suomela University of Helsinki, Finland
Subash Suri UC Santa Barbara, USA

Additional Reviewers
Aspnes, James Czyzowicz, Jurek
Auletta, Vincenzo D’Angelo, Gianlorenzo
Biely, Martin D’Emidio, Mattia
Borokhovich, Michael Das, Shantanu
Bouzid, Zohir De Marco, Gianluca
Burman, Janna Delling, Daniel
Casteigts, Arnaud Dereniowski, Dariusz
Chalopin, Jérémie Devismes, Stéphane
Charron-Bost, Bernadette Dobrev, Stefan
Cheilaris, Panagiotis Drees, Max
Cicalese, Ferdinando Friggstad, Zachary
Cicerone, Serafino Függer, Matthias
Cohen, Asaf Ganesh, Ayalvadi
Cord-Landwehr, Andreas Gargano, Luisa
Cordasco, Gennaro Georgiou, Chryssis
VIII Organization

Gu, Yue Pardo Soares, Ronan

Gu, Zhaoquan Parter, Merav
Hirvonen, Juho Pasquale, Francesco
Inoue, Michiko Peters, Joseph
Izumi, Taisuke Pietrzyk, Peter
Jung, Daniel Pignolet, Yvonne-Anne
Kakugawa, Hirotsugu Podlipyan, Pavel
Kamei, Sayaka Prencipe, Giuseppe
Keller, Barbara Rajsbaum, Sergio
Kesselheim, Thomas Robinson, Peter
Kling, Peter Rybicki, Joel
Kowalski, Dariusz Scalosub, Gabriel
Kralovic, Rastislav Scheideler, Christian
Kranakis, Evangelos Schwarz, Manfred
Kulkarni, Sandeep Suchan, Karol
Kutten, Shay Sundell, Håkan
Köhler, Sven Vaccaro, Ugo
Kößler, Alexander Wang, Yajun
Labourel, Arnaud Weinstein, Omri
Lamani, Anissa Widder, Josef
Larkin, Daniel Winkler, Kyrill
Le Merrer, Erwan Yamauchi, Yukiko
Lin, Xiao Yang, Siyu
Manne, Fredrik Yu, Dongxiao
Markou, Euripides Yuan, Wen
Moses, Yoram Zhang, Jialin
Navarra, Alfredo Zhang, Le
Nowak, Thomas Zémor, Gilles
Ooshita, Fukuhito
 Laudatio
2013 SIROCCO Prize for Innovation in Distributed
Computing awarded to Andrzej Pelc

David Peleg
on behalf of the award committee

It is a pleasure to award the 2013 SIROCCO prize for innovation in distributed

computing to Andrzej Pelc. The award is given for his significant contribution to
communication paradigms for information dissemination. Let us briefly review
this as well as some other major contributions and achievements of Andrzej,
with a focus on those related to his many SIROCCO papers. (Notably, Andrzej
coauthored 13 SIROCCO papers during the past two decades).
Much of Andrzej’s research concentrated on algorithmic aspects of distributed
computation and communication networks. In particular, he has contributed
extensively to the study of communication paradigms designed to disseminate
information in communication networks, such as broadcasting (i.e., sending a
message from one source to every node in the network), gossiping (i.e., broad-
casting from every node in parallel), multicasting and related paradigms. His
work focused on developing algorithmic techniques for performing these tasks
efficiently, by complexity measures such as time and (number and size of) mes-
sages, in different network architectures and under different assumptions, and
while attempting to ensure a variety of desirable properties, most notably fault-
tolerance. In addition, he studied the effects of having only partial knowledge
of the network topology on the performance of broadcast algorithms in message
passing networks and more recently in radio networks of different types (such as
arbitrary topology, geometric graphs or unit disk graphs). Moreover, he recently
published an authoritative book on these topics. Related to this line of work are
Andrzej’s SIROCCO papers [1–3], for example.
Another central theme of Andrzej’s research involved the question of over-
coming communication failures. He addressed this question in many different
settings and under different assumptions on the nature of possible failures, such
as random (dependent or independent) failures in nodes, links or transmissions,
benign or malicious (Byzantine) faults, and globally or locally bounded faults.
His studies served to broaden our knowledge and systematically organize the
arsenal of tools available to us in this area. Related to this line of work are
Andrzejs SIROCCO papers [4–6].
Andrzej was involved in studying fault-resilience in other contexts as well.
Let me mention one area related to fault-tolerance where he made significant
contributions, namely, the problem of system-level fault diagnosis in multipro-
cessor systems. This problem concerns a process where processors in the system
can test each other for failures. It is assumed that fault-free testers correctly
X D. Peleg

identify the fault status of tested processors, while faulty testers can give arbi-
trary test results. The goal is to develop algorithms for identifying correctly the
status of all processors, assuming that the number of faults does not exceed a
given upper bound. Andrzej’s work explored static and adaptive solutions at-
tempting to minimize the number of tests or the probability of error for various
system topologies and under different failure models. Related to this line of work
is Andrzej’s SIROCCO paper [7].
Recently, Andrzej started studying algorithmic problems in systems of au-
tonomous mobile agents or robots, which involve identical memoryless units
residing in the nodes of a network or a terrain. Most notable is his work on the
task of network exploration, which requires a mobile agent with small memory
to explore an unknown network, i.e., traverse all its nodes and edges (possi-
bly returning to the starting node), with no a priori knowledge of the network
topology. The feasibility of this task, and the efficiency of exploration algorithms,
depend on the model assumptions and on the class of allowed graph topologies.
Andrzej studied this problem in different settings on trees and general graphs,
paying attention mostly to the issues of memory requirements and resilience to
failures. A second central problem studied extensively by Andrzej in this area is
the rendezvous problem in networks, where two mobile agents, located in nodes
of an unknown network, have to meet at some common location. Other problems
he studied include the tasks of gathering the agents in one place and of searching
for a “black hole” (namely, a destructive node) in the network. Related to this
line of work are Andrzej’s SIROCCO papers [8–11].
In summary, Andrzej’s impressive technical achievements, his numerous in-
fluential contributions to efficient and failure-resistant algorithms in communi-
cation networks, and his leadership role in the research community in the field,
make him a most highly deserving candidate for the prize.

References
1. Krzysztof Diks, Andrzej Lingas, and Andrzej Pelc. An Optimal Algorithm for
Broadcasting Multiple Messages in Trees. In Proceedings of the 4th SIROCCO,
69–80, 1997.
2. Dariusz R. Kowalski and Andrzej Pelc. Time of Radio Broadcasting. In Proceedings
of the 10th SIROCCO, 195-210, 2003.
3. David Ilcinkas, Dariusz R. Kowalski, and Andrzej Pelc. Fast Radio Broadcasting
with Advice. In Proceedings of the 15th SIROCCO, 291-305, 2008.
4. Andrzej Pelc. Fast Fault-tolerant Broadcasting and Gossiping. In Proceedings of
the 2nd SIROCCO, 159-172, 1995.
5. Andrzej Pelc. Efficient Fault Location with Small Risk. In Proceedings of the 3rd
SIROCCO, 292-300, 1996.
6. Michel Paquette and Andrzej Pelc. Optimal Decision Strategies in Byzantine En-
vironments. In Proceedings of the 11th SIROCCO, 245-254, 2004.
7. Evangelos Kranakis, Andrzej Pelc, and Anthony Spatharis. Optimal Adaptive
Fault Diagnosis for Simple Multiprocessor Systems. In Proceedings of the 5th
SIROCCO, 82-97, 1998.
 2013 SIROCCO Prize for Innovation in Distributed Computing XI

8. Paola Flocchini, David Ilcinkas, Andrzej Pelc, and Nicola Santoro. Remember-
ing without Memory: Tree Exploration by Asynchronous Oblivious Robots. In
Proceedings of the 15th SIROCCO, 33-47, 2008.
9. Jurek Czyzowicz, David Ilcinkas, Arnaud Labourel, and Andrzej Pelc. Asyn-
chronous Deterministic Rendezvous in Bounded Terrains. In Proceedings of the
17th SIROCCO, 72-85, 2010.
10. Samuel Guilbault and Andrzej Pelc. Gathering Asynchronous Oblivious Agents
with Local Vision in Regular Bipartite Graphs. In Proceedings of the 18th
SIROCCO, 162-173, 2011.
11. Samir Elouasbi and Andrzej Pelc. Time of Anonymous Rendezvous in Trees: De-
terminism vs. Randomization. In Proceedings of the 19th SIROCCO, 291-302,
2012.
 Meeting in Networks
(Abstract of Award Lecture)

Andrzej Pelc
Université du Québec en Outaouais
Canada

Abstract. Two or more mobile entities, called agents or robots, start-

ing at distinct initial positions, have to meet. This task is known in the
literature as rendezvous and has many applications, both in everyday
life and in computer science. Among many alternative assumptions that
have been used to study the rendezvous problem, two most significantly
influence the methodology appropriate for its solution. The first of these
assumptions concerns the environment in which the mobile entities nav-
igate: it can be either a terrain in the plane, or a network modeled as
an undirected graph. The second assumption concerns the way in which
the entities move: it can be either deterministic or randomized. In this
talk we survey recent results on deterministic rendezvous in networks.
Adversarial Models for Wireless Communication
(Abstract of Keynote Lecture)

Andrea Richa

Arizona State University, U.S.A.

Abstract. In this talk, we present some recent work on adversarial mod-

eling of wireless communication. We use an adaptive adversary to model
the hard to predict physical interference, as well as other disruption
in communication caused by temporary obstacles, mobility, background
noise, co-existing networks, jammers, etc.
In particular, we focus on adversarial models for jamming. We present
simple, local-control medium access control (MAC) protocols for wireless
networks that are provably robust against adaptive adversarial jamming.
Our protocols are orthogonal to physical layer protocols that rely on a
broad spectrum, and can be used in conjunction with those or in net-
works where a broad spectrum is not available (e.g., sensor networks).
We present a summary of our work in this area, going from single-hop
wireless networks to multihop wireless networks modeled under SINR
(signal-to-noise ratio model), and from more standard adaptive adver-
sarial models for the jammer(s) to a more realistic adversarial model
where, in addition to knowing the protocol and its entire history, the
jammer also has some knowledge about the action of the nodes at the
current time step. Our protocols are energy efficient, and require only
very limited amount of knowledge about the jammer and the network.
We also present simulation results that further validate our theoretical
bounds.
We also address other recent work by the theoretical community on
applications of adversarial modeling in wireless computing that focus on
different paradigms (e.g., broadcasting, etc.).

The work on adversarial modeling of wireless jamming is joint work with Christian
Scheideler (U. of Paderborn, Germany), Stefan Schmid (TU Berlin Telekom Labs),
Jin Zhang (Google), Adrian Ogierman (U. of Paderborn) and Baruch Awerbuch (John
Hopkins University).
 Labeling Schemes with Forbidden-Sets
(Abstract of Invited Talk)

Cyril Gavoille

Université of Bordeaux, France

Abstract. The goal of labeling schemes is to understand how much

information must be attached to the nodes of a network (formalized as
labels) to solve a graph problem assuming the answer can be determined
solely on the basis of the labels of the nodes invoked in the query. In
this talk, I give a short survey on an extension of labeling schemes that
can answer graph problems where some of the nodes may be turn off (or
forbidden).
 Distributed Computation in Directed
and Dynamic Networks
(Abstract of Invited Talk)

Fabian Kuhn University of Freiburg, Germany

Abstract. We consider simple distributed data aggregation and infor-

mation dissemination problems such as computing the minimum or the
sum of a bunch of values or broadcasting multiple messages to all nodes in
a network. In standard, undirected networks, these tasks are well studied
and can be solved by simple distributed algorithms in time proportional
to the diameter of the network. In my talk, I will discuss the complexity
of such fundamental problems in networks with unidirectional links and
in networks with dynamic topology. We will see that in absence of stable,
bidirectional links, also the most basic distributed computation and in-
formation dissemination tasks become challenging, leading to a number
of fascinating new research questions.
 Table of Contents

Dynamic Networks Algorithms

Distributed Community Detection in Dynamic Graphs . . . . . . . . . . . . . . . . 1
Andrea Clementi, Miriam Di Ianni, Giorgio Gambosi,
Emanuele Natale, and Riccardo Silvestri

Exploration of the T -Interval-Connected Dynamic Graphs: The Case

of the Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
David Ilcinkas and Ahmed Mouhamadou Wade

A Characterization of Dynamic Networks Where Consensus Is

Solvable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Étienne Coulouma and Emmanuel Godard

Algorithms 1
Self-adjusting Grid Networks to Minimize Expected Path Length . . . . . . 36
Chen Avin, Michael Borokhovich, Bernhard Haeupler, and Zvi Lotker

On Advice Complexity of the k-server Problem under Sparse Metrics . . . 55

Sushmita Gupta, Shahin Kamali, and Alejandro López-Ortiz

Connected Surveillance Game . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Frédéric Giroire, Dorian Mazauric, Nicolas Nisse,
Stéphane Pérennes, and Ronan Soares

Online Algorithms
Non-Additive Two-Option Ski Rental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Amir Levi and Boaz Patt-Shamir

Competitive FIB Aggregation for Independent Prefixes: Online Ski

Rental on the Trie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Marcin Bienkowski and Stefan Schmid

A Nonmonotone Analysis with the Primal-Dual Approach:

Online Routing of Virtual Circuits with Unknown Durations . . . . . . . . . . 104
Guy Even and Moti Medina
XVIII Table of Contents

Social Networks Systems

Self-organizing Flows in Social Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Nidhi Hegde, Laurent Massoulié, and Laurent Viennot
Performance/Security Tradeoffs for Content-Based Routing Supported
by Bloom Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Hugues Mercier, Emanuel Onica, Etienne Rivière, and Pascal Felber

Influence Diffusion in Social Networks under Time Window

Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Luisa Gargano, Pavol Hell, Joseph Peters, and Ugo Vaccaro

Distributed Algorithms
Analysis of Fully Distributed Splitting and Naming Probabilistic
Procedures and Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Yves Métivier, John Michael Robson, and Akka Zemmari

A Deterministic Worst-Case Message Complexity Optimal Solution

for Resource Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Sebastian Kniesburges, Andreas Koutsopoulos, and
Christian Scheideler
Maintaining Balanced Trees for Structured Distributed Streaming
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Frédéric Giroire, Remigiusz Modrzejewski, Nicolas Nisse, and
Stéphane Pérennes

Robots
Rendezvous of Two Robots with Constant Memory . . . . . . . . . . . . . . . . . . . 189
Paola Flocchini, Nicola Santoro, Giovanni Viglietta, and
Masafumi Yamashita

Pattern Formation by Mobile Robots with Limited Visibility . . . . . . . . . . 201

Yukiko Yamauchi and Masafumi Yamashita
Optimal Gathering of Oblivious Robots in Anonymous Graphs . . . . . . . . 213
Gabriele Di Stefano and Alfredo Navarra

Wireless Networks
Probabilistic Connectivity Threshold for Directional Antenna Widths . . . 225
Hadassa Daltrophe, Shlomi Dolev, and Zvi Lotker

Broadcasting in Ad Hoc Multiple Access Channels . . . . . . . . . . . . . . . . . . . 237

Lakshmi Anantharamu and Bogdan S. Chlebus
 Table of Contents XIX

Profit Maximization in Flex-Grid All-Optical Networks . . . . . . . . . . . . . . . 249

Mordechai Shalom, Prudence W.H. Wong, and Shmuel Zaks

Algorithms 2
Measuring the Impact of Adversarial Errors on Packet Scheduling
Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Antonio Fernández Anta, Chryssis Georgiou, Dariusz R. Kowalski,
Joerg Widmer, and Elli Zavou

Optimal Buffer Management for 2-Frame Throughput Maximization . . . . 274

Jun Kawahara and Koji M. Kobayashi

Dynamically Maintaining Shortest Path Trees under Batches

of Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
Annalisa D’Andrea, Mattia D’Emidio, Daniele Frigioni,
Stefano Leucci, and Guido Proietti

Algorithms 3
Simultaneous Consensus vs Set Agreement: A Message-Passing-Sensitive
Hierarchy of Agreement Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Michel Raynal and Julien Stainer

Steiner Problems with Limited Number of Branching Nodes . . . . . . . . . . . 310

Dimitri Watel, Marc-Antoine Weisser, Cédric Bentz, and
Dominique Barth

Exact and Approximate Algorithms for Movement Problems

on (Special Classes of) Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Davide Bilò, Luciano Gualà, Stefano Leucci, and Guido Proietti

Maximum Distance Separable Codes Based on Circulant Cauchy

Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
Christian Schindelhauer and Christian Ortolf

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

 Distributed Community Detection
in Dynamic Graphs
(Extended Abstract)

Andrea Clementi1 , Miriam Di Ianni1 , Giorgio Gambosi1 , Emanuele Natale1 ,

and Riccardo Silvestri2
1
Università Tor Vergata di Roma
{clementi,diianni,gambosi}@mat.uniroma2.it, emanatale@gmail.com
2
Sapienza Università di Roma
silvestri@di.uniroma1.it

Abstract. Inspired by the increasing interest in self-organizing social

opportunistic networks, we investigate the problem of distributed detec-
tion of unknown communities in dynamic random graphs. As a formal
framework, we consider the dynamic version of the well-studied Planted
Bisection Model dyn-G(n, p, q) where the node set [n] of the network
is partitioned into two unknown communities and, at every time step,
each possible edge (u, v) is active with probability p if both nodes be-
long to the same community, while it is active with probability q (with
q << p) otherwise. We also consider a time-Markovian generalization of
this model.
We propose a distributed protocol based on the popular Label Prop-
agation Algorithm and prove that, when the ratio p/q is larger than nb
(for an arbitrarily small constant b > 0), the protocol finds the right
“planted” partition in O(log n) time even when the snapshots of the dy-
namic graph are sparse and disconnected (i.e. in the case p = Θ(1/n)).

Keywords: Distributed Computing, Dynamic Graphs, Social Oppor-

tunistic Networks.

1 Introduction

Community detection in complex networks has recently attracted wide atten-

tion in several research areas such as social networks, communication networks,
biological systems [14]. Understanding the community structure of a complex
network is a challenging crucial issue in several applications (good surveys on
this topic can be found in [2,12,20]). A modern application scenario (the one
this paper is inspired from) is that of Opportunistic Networks where recent
studies show that social-aware protocols provides efficient solutions for basic
communication tasks [5].

Partially supported by Italian MIUR under the PRIN 2010-11 Project ARS
TechnoMedia.

T. Moscibroda and A.A. Rescigno (Eds.): SIROCCO 2013, LNCS 8179, pp. 1–12, 2013.

c Springer International Publishing Switzerland 2013
2 A. Clementi et al.

The static Planted Bisection Model [3,4,13] (or Stochastic Blockmodel, as it

is known in the statistics community [15]) is a popular framework to formalize
the problem of detecting communities in random graphs.
The (Static) Planted Bisection Model: Centralized Algorithms. The
(static) Planted Bisection Model is defined as a static random graph G(n, p, q)
(with p, q ∈ (0, 1) such that q << p) where the node subset [n] = {1, 2, . . . , n}
is partitioned into two equal-sized unknown communities V1 and V2 and each
possible edge (u, v) is included with probability p if u and v both belong to the
same community while it is included with probability q otherwise1 . The goal
here is to identify the unknown partition.
In [13] and, successively, in [10,19], some efficient centralized algorithms have
been presented for the above problem. Such algorithms are based on centralized,
expensive procedures such as simulated annealing and spectral-graph computa-
tions: all of them require the full knowledge of the graph adjacency matrix and,
moreover, they work on static graphs only.
Community Detection in Opportunistic Networks. Recent studies in op-
portunistic networks focus on the impact of the agent social behavior on
some basic communication tasks such as routing and broadcasting [5]. Recently,
this issue has been investigated in an emerging class of opportunistic networks
called Intermittently-Connected Mobile Networks (ICMNs) [23]: such networks
are characterized by wireless links, representing opportunities for exchanging
data, that sporadically appear among network nodes (usually mobile radio de-
vices). So-called social-aware communication protocols rely on the reasonable
intuition that, since mobile devices are carried by people who tend to form com-
munities, members (i.e. nodes) of the same community are used to communicate
with each other much more often than nodes from different communities. Ex-
periments on real-data sets have widely shown that identifying communities can
strongly help in improving the protocol performances [5]. It thus follows that
community detection in ICMNs is a crucial issue.
As observed above, several centralized community-detection methods have
been proposed in the literature that may result useful for offline data analysis
of mobile traces. However, it is a common belief that next-future technologies
will yield a dramatic growth of self-organizing ICMNs where the network proto-
cols work without relying on any centralized server. In this new communication
paradigm, it is required that community detection is performed in a fully dis-
tributed way. To this aim, in this paper we consider an algorithmic solution to
community detection in ICMNs that relies on the epidemic mechanism known
as Label Propagation Algorithms [18,21].
The Dynamic Planted Bisection Model. In order to capture the high dy-
namicity of ICMNs, we consider the natural dynamic version of the G(n, p, q)
model. A dynamic graph is a probabilistic process that describes a graph whose
topology changes with time: so it can be represented by a sequence G = {Gt =
1
Observe that when p = q the random graph model is the well-known Erdös-Rényi
model.
 Distributed Community Detection in Dynamic Graphs 3

([n], Et ) : t ∈ N} of graphs with the same set V = [n] of nodes, where Gt is the
snapshot of the dynamic graph at time step t.
The dynamic version of the Planted Bisection Model, denoted as dyn-G(n, p, q),
consists of a dynamic graph where n is the number of nodes while p = p(n) and
q = q(n) are the edge-probability functions. At every time step t, each edge (u, v)
is included in Et with probability p if both u and v belong to the same commu-
nity Vi (i = 1, 2) while it is included with probability q otherwise (this model can
also be seen as a non-homogeneous version of the dynamic Erdös–Rényi graph
model [5]). So, the dynamic state (on/off) of an edge over the time is a random
variable having Bernoully distribution with parameter p or q, respectively. Of
course, relevant simplifications have been assumed in this model. However, in [6],
experimental validations have shown that some real ICMNs exhibit some crucial
connectivity properties (such as hop-diameter) which are well-approximated by
sparse dynamic Erdös–Rényi graphs.
A strong simplifying assumption in the dynamic Erdös–Rényi graph model is
time independence: the graph topology at time t is fully independent from the
topology at time t − 1. Edge Markovian Evolving Graphs (in short edge-MEG)
were first introduced in [7] as a generalization of the dynamic Erdös–Rényi graph
model that captures the strong dependence between the existence of an edge at
a given time step and its existence at the previous time step. An edge-MEG is
a dynamic random graph G(n, p↑ , p↓ , E0 ) = {Gt = ([n], Et ) : t ∈ N} defined
as follows. Starting from an initial random edge set E0 , at every time step,
every edge changes its state (existing or not) according to a two-state Markovian
process with probabilities p↑ and p↓ . If an edge exists at time t, then, at time
t + 1, it disappears with probability p↓ . If instead the edge does not exist at
time t, then it will come up at time t + 1 with probability p↑ . We observe that
the setting p↓ = 1 − p↑ yields a sequence of independent Erdös–Rényi random
graphs, i.e., dynamic Erdös–Rényi graphs, with edge probability p = p↑ . Edge-
MEGs have been adopted as concrete models for several real dynamic networks
such as faulty networks [8], peer-to-peer systems [22], mobile ad-hoc networks
[22]. Furthermore, Edge-MEGs have been considered by Whitbeck et al [23] as
a concrete model for analyzing the performance of epidemic routing on sparse
ICMNs. In this paper, we consider the Edge-MEG as a mathematical model
for ICMNs. The dynamic Planted Bisection Model can be easily generalized
in order to include edge-MEGs: here, we have two edge-probability parameter
pairs (p↑ , p↓ ) and (q↑ , q↓ ) between two nodes u and v depending on whether they
both belong to the same community or not. So, if both u and v belongs to the
same community then the edge (u, v) is governed by the 2-state Markov chain
with parameters (p↑ , p↓ ) otherwise the edge is governed by the 2-state Markov
chain with parameter (q↑ , q↓ ). We assume that q↑ << p↑ and, according to the
parameter tuning performed in [23], it turns out that the best fitting to real
scenarios is achieved by setting p↓ , (and q↓ ) as absolute constants.
The algorithmic goal in the dyn-G(n, p, q) model is to design a fully-distributed
protocol that computes a good (node) labeling for the dynamic graph G: we say
that a function Z : V → {1, 2} is a good labeling for G if Z labels each community
4 A. Clementi et al.

with a different label: ∀i, k ∈ {1, 2} ∀u ∈ Vi ∀v ∈ Vk : Z(u) = Z(v) ↔ i = k.

Nodes are entities that share a global clock and know (a good approximation
of) the number n of nodes but it is not required they have distinct IDs. Initially,
each node does not know anything about communities while, at every time step,
it can exchange information with its current neighbors.
In [16], some greedy protocols are tested on specific sets of real mobility-trace
datas. By running such protocols, every node constructs and updates its own
community-list according to the length and the rate of the contacts observed so
far by itself and by the nodes it meets. So, the protocol exploits the intuition
that communities are formed by nodes that use to meet each other often and
for a long time. However, no analytical result is given for such heuristics that,
moreover, require nodes to often update and transmit relatively large lists of
node-IDs.
Label Propagation Algorithms. A well-studied community-detection strat-
egy is the one known as Label Propagation Algorithms (LPA) [21]. This strategy
is based on a simple epidemic mechanism which can be efficiently implemented
in a fully-distributed fashion since it requires easy local computations: it is thus
very suitable for opportunistic networks such as ICMNs. In its basic version,
some distinct labels are initially assigned to a subset of nodes; at every step,
each node updates its label (if any) by choosing the label which most of its (cur-
rent) neighbors have (the majority label); if there are multiple majority labels,
one label is randomly chosen. Several versions of LPA-based protocols have been
tested on a wide range of social networks [11,18,21]: such works experimentally
show that LPA-based protocols work quite efficiently and are effective in provid-
ing almost good labeling. Based on extensive simulations, [21] empirically shows
that the average convergence time of the (synchronous) LPA-based protocols is
bounded by some logarithmic function on n. Clearly, the goal of the protocol is
to converge to a good labeling for dyn-G(n, p, q). Despite the simplicity of LPA-
based protocols, very few analytical results are known on their performance over
relevant classes of graphs. As observed in [17], it seems hard to derive, from
empirical results, any fundamental conclusions about LPA behavior, even on
specific families of graphs. Recently, [11] provided a semi-synchronous version of
the LPA-based protocol and formally prove that it guarantees finite convergence
time on any static graph. In [17], an LPA-based protocol has been analyzed on
the Planted Partition Model for highly-dense topologies. In particular, their anal-
ysis considers the static model G(n, p, q) with p = Ω(1/n1/4− ) and q = O(p2 ).
In this restricted static case, it is shown that the protocol converges in con-
stant expected time and conjectured a logarithmic bound for sparse topologies.
In general, providing analytical bounds on the convergence time of LPA-based
protocols over relevant classes of networks is an important open question that
has been proposed in several papers arising from different areas [11,17,21].
Our Algorithmic Contribution. We provide an efficient distributed LPA-
based protocol on the dynamic Planted Bisection Model dyn-G(n, p, q) with ar-
bitrary p > 0 and q = O(p/nb ) where b > 0 is any arbitrarily small constant.
 Distributed Community Detection in Dynamic Graphs 5

Our protocol yields  with high probability2 (in short w.h.p.) a good labeling in
O max{log n, pn } time. The bound is tight for any p = O(1/n) while it is
log n

only a logarithmic factor larger than the optimum for the rest of the parameter
range (i.e. for more dense topologies). For the first time, we thus formally prove a
logarithmic bound on the convergence time of an LPA-based protocol on a class
of sparse and disconnected dynamic random graphs (i.e. for p = Θ(1/n)). The lo-
cal labeling rule adopted by the protocol is simple and requires no node IDs: the
only exchanged informations are the labels. Our protocol can be easily adapted
in order to construct a good labeling in the presence of a larger number of equal-
sized communities (provided that this number is an absolute constant) and, more
importantly, it also works for the Edge-MEG model G(n, p↑ , p↓ , q↑ , q↓ , E0 ) in the
parameter range q↑ = O(p↑ /nb ), where b is any positive  constant. In the lat- 
ter model, the completion time is w.h.p. bounded by O M · max log n, log n
p↑ n
where M is a bound on the mixing time of the two 2-state Markov chains gov-
erning the edges of the dynamic graph. It is known that (see for example [7])
  
1 1
M = O max , , log n
p↑ + p↓ q↑ + q↓
Observe that, when p↓ and q↓ are some arbitrary positive constants and p↑ =
Ω(1/n) (this case includes the “realistic” range derived in [23]), then M =
O(log n) and the bound on the completion time becomes O(log2 n). This bound
is only a logarithmic factor larger than the optimal labeling time in the case of
sparse topologies, i.e., when p↑ = Θ(1/n). We run our protocol over hundreds of
random instances according to the dyn-G(n, p, q) model with n varying from 103
to 106 . Besides a good validation of our asymptotical analysis, the experiments
show further positive features of the protocol. Our protocol is indeed tolerant to
non-homogeneous edge-probability functions. In particular, the protocol almost-
always returns a good labeling in Bernoullian graphs where the edge probability
is not uniform, i.e., for each pair (u, v) of nodes in the same community, the
parameter pu,v is suitably chosen in order to yield irregular sparse graphs. A
detailed description of the experimental results together with all technical proofs
can be found in the full version of the paper [9]

2 The Protocol and Its Analysis

In this extended abstract, we consider the dynamic graph dyn-G(n, p, q) under
the following restrictions: the parameter p is known by every node; there are only
2 communities V1 and V2 , each of size n/2 (n is an even number); the labeling
process starts with (exactly) two source nodes, s1 ∈ V1 that is labeled by z1 and
s2 ∈ V2 that is labeled by z2 with z1 = z2 . The parameters p and q belong to
the following ranges
1 d log n p 
 p  and q = O , for some constants d > 0 and b > 0. (1)
n n nb
2
As usual, we say an event holds with high probability if it holds with probability at
1
least 1 − nΘ(1) .
6 A. Clementi et al.

Such restricions make the description easier, thus allowing us to focus on the
main ideas of our protocol and of its analysis. In the full version [9], we show
how to obtain the general result stated in the Introduction.
The protocol relies on the simple and natural properties of LPA. Starting from
two source nodes (one in each community), each one having a different label, the
protocol performs a label spreading by adopting a simple labeling/broadcasting
rule (for instance, every node gets the label it sees most frequently in its neigh-
bors). Since links between nodes of the same community are much more frequent
than the other ones, we can argue that the good-labeling will be faster than the
bad-labeling (in each community, the good labeling is the one from the source of
the community while the bad labeling is the one coming from the other source).
However, providing a rigorous analysis of the above process requires to cope
with some non-trivial probabilistic issues that have not been considered in the
analysis of information spreading in dynamic graphs made in previous papers
[1,7,8]. Let us consider any local labeling rule that depends on the label config-
uration of the (dynamic) neighborhood of the node only. At a given time step,
there is a subset Ic ⊆ [n] of labeled nodes and we need to evaluate the probabil-
ities Pg (Pb ) that a non-labeled node gets a good (bad) label in the next step.
After an initial phase, there is a non-negligible probability that some nodes will
get the bad label. Then, such nodes will start a spreading of the bad label-
ing at the same rate of the good one. Observe also that good-labeled nodes may
(wrongly) change their state as well, so, differently from a standard single-source
broadcast, the epidemic process is not monotone with respect to good-labeling.
It turns out that the probabilities Pg and Pb strongly depend on the label-
balance between the sizes of the subsets of well-labeled nodes and of the badly-
labeled ones in the two communities. Keeping a tight balance between such values
during all the process is the main technical goal of the protocol. In arbitrary label
configurations over sparse graph snapshots, getting “high-probability” bounds
on the rate of new (well/badly) labeled nodes is a non-trivial issue: indeed,
it is not hard to show that, given any two nodes v, w ∈ [n] \ Ic , the events
“v will be (well/badly-)labeled” and “w will be (well/badly-)labeled” are not
independent. As we will see, such issues are already present in this “restricted”
case. A first important step of our approach is to describe the combination
between the labeling process and the dynamic graph as a finite-state Markovian
process. Then, we perform a step-by-step analysis, focusing on the probability
that the Markovian Process visits a sequence of states having “good-balance”
properties.
Our protocol applies local rules depending on the current node’s neighborhood
and on the current time step only. The protocol execution over the dynamic
graph can be represented
 by the following
 Markovian Process: for any time
(t) (t) (t) (t)
step t, we denote as k1 , k2 , h1 , h2 ; Et the state reached by the Markovian
(t)
Process where ki denotes the number of nodes in the i-th community labeled by
(t)
label zi at time step t and hi denotes the number of nodes in the i-th community
labeled by label zj at time step t, for i, j = 1, 2 and j = i. In particular, the
Markovian Process works as follows
 Distributed Community Detection in Dynamic Graphs 7

   
(t) (t) (t) (t) dyn-G(n, p, q) (t) (t) (t) (t) protocol
... → k1 , k2 , h1 , h2 ; Et −→ k1 , k2 , h1 , h2 ; Et+1 −→
 
protocol (t+1) (t+1) (t+1) (t+1) dyn-G(n, p, q)
−→ k1 , k2 , h1 , h2 ; Et+1 −→ ...
The main advantage of this description is the following: observe the process in
any fixed state and consider the set of nodes U still having no label. Then it is not
hard to verify that, in the next time step, the events {“node v gets a good/bad
label”, v ∈ U }, are mutually independent. This will allow us to prove strong-
concentration bounds on the label-balance discussed above for a sufficiently-long
sequence of states visited by the Markovian Process, thus getting a large fraction
of well-labeled nodes in each community within a short time; this corresponds
to a first protocol stage called fast spreading of the good labels.
Unfortunately, this independence property does not hold among labeled nodes
of the same community, let’s see why in the next simple scenario. Assume that
the rule is the majority one, consider two nodes u and v having the same label
z at time t, and assume the event E = “node u will keep label z at time t + 1”
holds. Then the event “(u, v) ∈ Et | E” is more likely and, thus, according to the
majority rule, the event “v gets label z | E” is more likely as well. This clearly
shows a key-depencence in the label spreading.
In order to overcome this issue, our protocol allows every node to change
its first label-updating rule only after a spreading stage of suitable length (we
will see later this stage is in fact formed by 3 consecutive phases): we can thus
analyze the spreading of the good labeling (only) on the current set of unlabeled
nodes (where stochastic independence holds) and prove that the process reaches
a state with a large number of well-labeled nodes. After this spreading stage,
labeled nodes (have to) start to update their labels according to some simple rule
that will be discussed later. In the full version [9], we prove this saturation phase
has logarithmic convergence time by providing a simple and efficient method to
cope with the above discussed stochastic dependence.

2.1 A Restricted Setting: Formal Description

The protocol works in 5 consecutive temporal phases: the goal of this phase parti-
tion is to control the rate of new labeled nodes as function of the expected values
(t) (t)
reached by the random variables (r.v.s) ki , hi (at the end of each phase). In-
deed, when such expected values reach some specific thresholds, the protocol
and/or its analysis must change accordingly in order to keep the label configura-
tion well-balanced in the two communities during all the process and to manage
the stochastic depencence described above.
At any time step t, we denote, for each node v ∈ Vi , the number of zi -labeled
neighbors of v as Niv (t), for i = 1, 2. Given a node v ∈ V , the set of its neighbors
at time t will be denoted as Γt (v). For the sake of brevity, whenever possible we
will omit the parameter t in the above variables and, in the proofs, we will only
analyze the labeling in V1 , the analysis for V2 being the same.
Another random document with
no related content on Scribd:
to work well; so that Andy was now able to attend strictly to
business, and not spend most of his time trying to keep an
equilibrium.
The biplane had kept on rising, even after Frank brought his craft on
an even keel. He wondered what the meaning of this movement
could be. At the moment all he could think of was that Puss believed
he would be able to make better direct time if he kept just fifty feet
above the line of flight adopted by his rival.
The shouts became dimmer as they left the aviation field further in
their wake. Still they knew that every eye was focused on their
movements, and hundreds of glasses brought into use to note every
little movement of the two rival craft.
Frank seemed just as calm and collected as though he had been
going off on an ordinary little trip, to give the engine a warming-up.
From where he sat he could watch the working of the wonderful little
Kinkaid engine; for in a monoplane it is always secured before the
pilot. Some aviators incline to call this an advantage, because with a
biplane the engine must of necessity be back of the navigator.
He and Andy could converse without trouble, should the spirit move
them. True, with the little engine doing its liveliest, and the muffler
not throttled in the least, there arose a necessity for raising the voice
a trifle in order to be heard even a foot away; but Andy had good
lungs.
“Frank, they’re holding the advantage!” he exclaimed, when they had
been moving along for another minute, and heading almost straight
for the summit of the high mountain.
Looking down, Andy could see the trees of the forest far below. A
thousand feet, yes, possibly twelve hundred, they had risen without
making use of the usual method of “boring” for altitude. It was really
higher than as yet they had dared venture. Still, strange to say, Andy
did not feel the least particle of fear in connection with any possible
tumble.
His anxiety was concentrated upon the chances of the biplane
leading them all the way along the journey, just as though it were a
procession, and Puss the grand marshal.
“I know it,” replied Frank, without showing any concern.
“But we must break into their pace better than this, if we hope to
win!” declared Andy, who had taken a look upward, to see the
grinning face of Sandy Hollingshead turned down toward them, just
as though he already scented victory in the air.
“Don’t you worry, Andy!” Frank simply said.
“Are you holding back anything?” demanded the other, eagerly.
“A little. We can do better. Just wait, and trust me!”
Andy breathed more easily. When Frank spoke like that it always
gave him a new lease on hope. That came of knowing his cousin so
well, and having such perfect confidence in his sterling abilities.
When Frank Bird said “Have faith, and trust me to pull through,”
Andy was ready to believe almost anything could happen.
So he once more turned his eyes on the cap of the great mountain
which they were approaching at a rapid rate.
Old Thunder Top was indeed an imposing spectacle as seen from a
level. Of course, in the eyes of real mountaineers, the elevation
would have been a trifle, and they might have spoken of it as a mere
“foothill”; but to all loyal Bloomsbury boys it was always referred to
with respect, as the “Mountain”! Perhaps the fact of those queer cliffs
preventing any one from reaching the summit added to the
admiration with which it was gazed upon. Had the boys been
accustomed to picnicking upon that top whenever they pleased, it
must have lost much of its grandeur.
Frank had judged wisely.
“We are on a direct line with the top, don’t you think?” he asked
presently.
“As near as I can say, we are,” replied the other, as the monoplane
veered just a trifle when a gust of wind, coming from beyond the
peak, met her, and Frank manipulated his aerial steed after a clever
fashion he had inaugurated.
“If anything, a little higher,” continued Frank.
“Yes, that’s a fact,” assented Andy, with another keen look.
“That’s what I wanted. Sure you’ve got that flag handy, are you?”
went on the pilot of the speeding monoplane.
“Just you get there, and you’ll see how soon I jump out and wave it!”
declared Andy, with a vim.
Frank looked up.
The biplane still led, though by a narrow margin. At the same time, if
the relative distance were maintained to the close of the race, Puss
and Sandy would be able to land on the level plateau of the
mountain top a few seconds before them, and that would win out.
And Frank had now to decide in his active mind whether the vital
second had not arrived when he meant to release the little extra
speed he had been so jealously husbanding ever since they started.
He believed that Puss had opened his throttle to its widest extent
from the beginning, and would not have any reserve force left.
Knowing the reckless nature of his rival gave him this idea. If it
proved true, there was nothing to be feared, and they could
outdistance the biplane without difficulty.
The mountain top was now near enough for them to see the
formation of the rock. It was a matter of considerable moment
whether there was a level space large enough to allow of a landing
in safety; for an aeroplane cannot stop in twenty feet of clear ground
when going at speed.
The sun was now starting down its western journey, and
unfortunately it interfered to some extent with a clear view. Frank had
even thought of this. It was of tremendous importance to him that he
knew just what he had to expect when he attempted to land on top of
that pile of rock; and in order to assist his eyesight he had fastened a
pair of good field glasses, so that he could glue his eyes to them for
a couple of seconds, while the monoplane was shooting along in a
direct course.
The result satisfied him that his little scheme had paid, for he made a
mental photograph of the plateau, and noted just which side offered
the better advantage in the way of a landing place.
But now the decisive second was at hand when he must release his
little reserve speed, and send the monoplane on a trifle faster.
Much depended on the result. If Puss had been wise enough to do
the same thing he would be able to retain the advantage which he
now possessed, perhaps even add to the gap between them.
And so it was with more or less trepidation that Frank gave his
throttle its very last push.
“Oh!” exclaimed Andy, who of course felt the little jump which this
new impetus gave the already drumming motor.
Of course the ever watchful Sandy, from his eyrie above, would
immediately discover this maneuver on the part of the rival
aeroplane, and communicate the news to his companion.
What then? Would the pilot of the biplane simply follow suit, and thus
increase the speed of his craft? That would be the logical deduction,
if only Puss had any surplus in reserve.
Andy was on the watch, for that was a part of his business. Having
little else to do, since Frank managed the engine and the tail rudder
entirely, he was expected to discover, and report, everything that
might bear in the least on their chances.
And Andy immediately gave utterance to a low cry of delight. Even
had he not spoken a single word Frank would have known full well
that they were now rapidly closing the little gap that up to now had
stood between the monoplane and its larger rival.
“We’re going to run past them, Frank!” exclaimed Andy, doubtless
quivering with concentrated nervousness and delight. “Already we’ve
cut their lead down by half! Oh! don’t poor old Sandy looked scared
now! We’ve got them on the run, Frank, as sure as you live!”
But Frank made no answer. Perhaps a slight smile, as of pleasure,
may have crossed his set face. Only too well did he know that when
Puss Carberry was concerned, a fellow could never be positive of
having won until the line were actually crossed; and even then it was
his favorite stunt to claim “foul!”
To tell the truth, Frank would be very much easier in his mind when
once they were clear of that hovering biplane, whose Gnome engine
was banging away just above them as though scores of guns were
being discharged in rapid succession.
The suspicion that had flashed athwart his mind earlier in the race
now returned in double force; he feared lest those reckless rivals,
ready to take the most desperate chances rather than confess to
defeat, would attempt one of their customary mean tricks.
That may have been why, in the very beginning, Puss had insisted
upon keeping at a higher level than the other aeroplane! It gave him
the privilege of seeing how his rival might be coming on, without
craning his neck. It also opened up an opportunity for something to
drop, of course accidentally, just when the smaller air craft was
forging ahead!
Frank drew a long breath. He knew that the crisis of the race was
now upon him. The speed of the biplane had not increased by even
a fraction, which fact proved plainly that Puss had not held anything
in reserve.
Then it looked very much as though Puss and his chum were bound
to be beaten, unless they adopted some underhand tactics, trusting
to the distance, and the little haze encountered at this height, to
screen their despicable action from the eyes of those who looked
through all those glasses.
And Andy too must have feared something of the sort, for he was
keeping his eyes fastened on the biplane, now almost directly
overhead. Frank knew that he must meet the sudden emergency, if
one arose, with quickness, if he meant to prevent a catastrophe. He
was resolutely determined not to slow down, and allow the others to
gain a victory they had not earned; that was not Frank Bird’s way.
“Oh! he’s going to drop something on us, Frank!” cried Andy,
suddenly.
“Who is—Sandy?” demanded the other. “Give me a push as it leaves
his hand!”
Andy did not understand, but he was in the habit of minding what
Frank said; and three seconds later he brought his elbow sharply
against the pilot’s side.
Sandy had let go above, allowing the bulky object to fall through
space!
 CHAPTER XXII.
WELL WON!

Instantly Frank closed the throttle, and shut off all power!
It was taking a big chance; but there was nothing else to be done.
No matter what it was Sandy had let slip, expecting that it would fall
upon the monoplane, to at least cause consternation, and in some
way lessen the speed of the smaller craft, Frank did not mean that it
should strike them, if he knew it.
Of course their speed instantly slackened; not much, perhaps, but
just enough to allow of a miss in the calculations of the unscrupulous
Sandy.
Some object whizzed past, just in advance of the now descending
monoplane. Immediately it went by, Frank, under the belief that the
danger was now over, once more carefully opened the throttle.
Joy! the faithful little Kinkaid answered to the call, and began to
renew its former volleying. Once more they were going along swiftly,
though a bit lower than when the sudden emergency had caused
such prompt work on the part of the wide-awake pilot.
Frank shot a look upward.
The biplane had not diminished its speed an iota all this while. Puss
was attending to his part of the business, leaving all other matters in
the care of his well groomed assistant.
Both of them were leaning forward, staring down and backward at
the monoplane. Even at that distance Frank could see that their
faces were as white as chalk, as though the enormity of what they
had done now burst upon them. Perhaps they may even have felt a
spasm of relief at that moment, because the sand bag which had
been dropped had missed its intended target, thanks to Frank’s
ready wit.
Now the monoplane seemed to be pushing forward with more speed
than ever, as if bent on making up for lost time. And Frank was
rising, too, for he knew he must of necessity find himself above the
crown of the mountain, when ready to alight.
“What was that they dropped?” he asked of Andy.
“Looked like a sand bag,” replied the other; “but whatever would they
be doing with such a thing in a biplane?”
“That was what I wanted to know,” replied Frank, “when I saw it tied
there with a cord; and Puss explained that he and Sandy were not
quite heavy enough. Said their experiments had proved the biplane
could make faster time with a little more weight!”
“He just lied!” burst out the indignant Andy. “A hundred to one they
took that sand bag up with them on purpose to drop it on us if we
tried to pass. And look how he kept hovering up there. That gives
him away, I tell you!”
“Perhaps he got the idea from hearing how that other sand bag
came down on our lumber pile, nearly squashing us while we slept!”
observed Frank.
“Well, he only had one, didn’t he?” questioned Andy, showing
considerable nervousness; for they were now once more directly
under the biplane.
“Only one, so the trick can’t be duplicated,” answered Frank,
confidently.
“I wouldn’t put it past that sneak Sandy, to let a monkey wrench drop
on us, if he could lay hands on one,” cried Andy; and then raising his
voice he shouted: “Hey! don’t you dare try that trick again! Accidents
don’t happen twice in succession; and they’ll hang you for murder if
anything knocks us out. They can see everything that goes on up
here!”
Possibly this was stretching it pretty lively; but all the same Andy
meant to frighten Sandy, so that he would not dream of following up
a blunder by a second miserable attempt.
“It’s too late, anyhow!” said Frank, with a vibration in his voice that
might be caused by anticipated triumph.
“Yes, we’re passing them, as sure as you live! Look at the poor old
biplane dropping out of the race, Frank! Why, it might just as well
stand still as try to keep up with this dandy little airship, once you pull
the throttle wide open! We’ve got ’em beat to a frazzle, I tell you!
Goodbye, fellows. We’ll wait for you on top of old Blitzen and
Thunder! Sorry, but somebody’s got to eat the drumsticks of the
turkey!”
Andy was feeling immensely relieved. The monoplane no longer
ranged under its larger opponent. Superior speed, backed by careful
management, had given them the lead. And as Andy declared, it
looked as though the race might end in a real Garrison finish, the
one behind shooting to the front when on the home stretch.
No matter what they would have liked to do, Puss and Sandy were
now helpless to hinder the triumphal arrival of their rivals on top of
the mountain. Everything depended on the success that might attend
Frank, when making his drop. Should he make a bad job of it, and
shoot beyond the other edge of the plateau, possibly after all the
others might be the first to land. It was their only hope.
Frank knew what he had before him. He was keenly alive to the
chances of making a poor landing. And like a wise general he had
anticipated all such things before now, even practicing stopping
within a certain limited space when going at full speed.
“We’re high enough, all right, Frank!” cried Andy reassuringly, as
they swooped down toward the top of the ominous cliffs that had
always barred their gaining lodgment on the crown of Old Thunder
Top.
“Yes, no doubt about that, Andy,” returned the other, confidently.
“Now, be ready for your part. Remember, not to blunder, or we may
lose out yet. They are coming hotfooted after us, you know!”
“I’ll remember. You can trust me, Frank!”
Really, Andy was showing commendable grit and steadiness as the
termination of the fierce race through the upper currents of the air
drew nearer and nearer its termination. There was hope that in time
he might conquer that nervousness of his, and play his part as a
worthy successor to his famous father, the professor.
Like a great bird they sailed straight for the plateau marking the flat
top of the elevation. Frank could even see the nest of sticks and
grass that marked the home of the two great kings of the air, the bald
eagles, now circling around overhead, and evidently greatly excited
at the coming of these astonishing creatures, with their loud
crackling voices.
“Say, you don’t think they’ll tackle us, and knock us off the rocks?”
cried Andy, who had also been taking notice of the wheeling birds,
now swooping down, and anon rising higher on outspread pinions.
“Keep an eye on ’em!” was all Frank could say; for just then they
were close to the outer edge of the plateau, and his entire attention
had to be focused upon what was before him, since one little
misjudgment might bring about the ruination of his plans, however
admirably fashioned.
Andy had already clutched the little pole to which the National
emblem was fastened, so that not a second might be lost in giving it
to the breeze, once his feet touched the plateau. But his anxiety was
sufficient to cause him to reach to the tool box, and extract a rather
long alligator-jaw wrench, which he had in his mind as the most
suitable weapon of defense, in case of an emergency, in which one
of those old pirates of the air figured.
Angry shouts came from the rear. Of course it was the very last
despicable little scheme of the baffled plotters, by which they hoped
to disconcert Frank enough to cause him to make a bad landing, so
that they might come swinging along in time to fly the flag first.
But Frank was not built that way. It would have to be something
much greater than a few harmless hoots, to cause him to lose his
head, especially when so very important a result depended on his
nice judgment.
He had calculated to a fraction of a foot just how far above the
plateau the monoplane was situated, so that when he shut off the
engine they would drop lightly just where he figured.
And Andy knew how to apply the drag brake, so as to haul up in a
short distance.
All the same it must have been a moment of extreme anxiety to both
of the daring young aviators. They had victory within their grasp, and
in another few seconds it would be clinched and riveted, when their
flag flew from the crown of the now conquered Old Thunder Top, that
had so long defied all attempts at mastery.
Just as Frank had figured the monoplane glided down after the
engine was stopped, and touched the rocks as gently as ever he had
come to earth, running along on the three bicycle wheels, jolting over
the rough surface, yet gradually coming to a standstill, as the brake
got in its work.
Indeed, the aeroplane had not actually come to a stop before Andy
was out of his seat, and wildly flaunting the flag that had been given
him by the head of the sports committee. He knew that every eye far
away was riveted on the spot, and that since the biplane was still
afloat, those who had glasses could readily see how the other air
craft had landed first, and hence won the race.
Of course Andy shouted like a young cowboy; he would hardly have
been human not to have found some such outlet for the pent-up
emotions that were threatening to suffocate him.
And naturally enough, those victorious whoops must have been so
like gall and worm-wood to the disheartened pair just about to alight
on the plateau, a quarter of a minute after the victors had taken
possession.
Frank knew too that there must be the added consciousness of
having attempted a nasty trick, and failed! There is possibly no
meaner feeling that can overwhelm a boy than to realize that he has
tried to down a rival through trickery, that must have been apparent
to many eyes, and failed.
But Frank’s was a generous nature. Even then he was resolved not
to press the charge against his defeated rivals. No harm had
resulted from the contemptible endeavor to delay or injure them; and
doubtless already Puss must regret that he had ever allowed himself
to conspire with Sandy to carry it out. Surely he could not have
realized what a terrible thing it was they had attempted. Let it go as
an accident then; but all the same Frank was bound to make sure
that he did not again sail the upper currents under any sort of an air
craft which either Puss Carberry or Sandy Hollingshead piloted.
The others managed to alight on the plateau, though their
momentum was enough to have carried them over the other edge
had not Frank, who had left his own machine, laid hold and held the
biplane back.
Puss looked white and confused. Sandy, on the other hand, scowled,
and clenched his hands menacingly, as though so sore over his
defeat that he was almost tempted to rush on the cheering bearer of
the flag, and have it out with him there on the very apex of Old
Thunder Top.
Possibly the sight of that long alligator-jaw wrench which Andy still
clutched in his right hand may have deterred the belligerent Sandy,
though his face continued to work spasmodically, as though he might
be saying things not at all complimentary to the object of his
aversion.
Suddenly Frank gave utterance to a shout. Faintly on the air came
the uproarious cheering of the tremendous throng, away down
yonder on the aviation field, as they saw the humiliation of the once
proud Thunder Top; but it was not in connection with this that Frank
gave tongue.
“Look out!” he cried, “the eagle!”
Sandy Hollingshead happened to be the one picked out by the angry
bird, upon whom to first try his claws and beak. The boy turned at
Frank’s cry, and just managed to throw his arm up to screen his face
from the attack. But the heavy bird struck him with tremendous force
so that Sandy was hurled over upon the rocks, and more or less
bruised and cut.
Andy ran toward him, bent on defending the prostrate lad from any
further attack on the part of the enraged feathered king of the air.
Then he stopped short, gaped at something that lay there on the
rocks, having fallen undoubtedly from one of Sandy’s coat pockets
when he was sent sprawling; and with a shrill laugh Andy snatched
the object up in his hand.
“Frank, looky here would you! Just think of me finding it up on Old
Thunder Top!”
And Frank stared, as well he might, for his chum was holding up the
missing little aluminum monkey wrench for which he had so long
searched everywhere.
 CHAPTER XXIII.
PROVEN GUILTY—CONCLUSION.

“Hurrah! found at last! Didn’t I tell you I’d run it down sooner or later,
Frank? And just to think that this sneak had it all the while; grabbed it
some time when perhaps it fell out of my pocket. It’s the greatest
thing ever! I’m glad I came up here!”
So Andy kept on crying, to the secret amusement of his cousin.
Evidently the other found more real joy in the sudden and
unexpected recovery of his missing monkey wrench, than in the
great victory which the little monoplane had won.
“Look out! There come both of them, Andy! Drop flat!” he yelled, as
he saw the circling eagles start to swoop down again.
Andy just saved himself by following directions, for one of the eagles
barely missed him. Sandy was sitting up, and rubbing the back of his
head, where it had come in contact with the hard rock. He appeared
half dazed, and evidently there was little use demanding any
explanation as to how the precious tool chanced to be in his
possession. Truth to tell, Andy never did find out, and had to jump at
conclusions.
The great birds continued to wheel and dart at the intruders, so that
all of the boys were soon engaged in defending themselves.
“They think we mean to rob their nest of the two eaglets you can see
there,” was Frank’s explanation. “Perhaps if we go over to the other
side of the plateau they may haul off, and let us embark again. I
wouldn’t like to hurt them, boys.”
“And I’d kill the whole outfit, if I had my way,” grumbled Sandy,
whose clothes were torn and marked with blood, where the sharp
talons of the furious bird had clawed along his person.
“Oh! well, we’ll leave you here to clean ’em out, if you say so,”
remarked Puss, who was himself anxious to get down from that
dizzy height as soon as possible, and feeling ugly toward all
creation, as fellows who make a bad mess of things usually are.
“Not much you don’t,” said Sandy quickly. “I’m going when you get
good and ready, bet your life on it. Wouldn’t ketch me staying up
here alone. Wow! even if I had a rope long enough to reach down,
I’d be afraid to chance it. Come along, Puss, we ain’t got no call to
stay here any longer. Let’s vamose.”
The biplane was the first to start off, and Frank was a little nervous
as to whether the thing could be successfully navigated in so short a
space. But nothing went wrong, and presently those who manned
the other aeroplane also took their places and made the trial.
The flag had been left fluttering in the breeze, Andy having fixed the
short pole in a crevice of the rocks, where he could wedge it fast.
With the aid of any fairly decent glasses it could be seen from town;
and would doubtless serve to stimulate many boys in the endeavor
to accomplish some similar feat of daring.
The eagles were still soaring in great circles, now rising, and again
swooping down on their broad pinions. Frank even feared that they
might take a notion to strike the strange bird that had dared invade
their eyrie home; but evidently the eagles had come to the wise
conclusion that they need fear nothing from the visit of the two
aeroplanes, for they followed them but a short distance, to return,
and perching on a crag give utterance to what might be called a
victorious scream.
“Say, what d’ye think of that?” demanded Andy, laughing as the
sound floated to them while speeding along. “They reckon they’ve
licked us, good and plenty.”
“Well,” said Frank, quickly, “so they have in one sense, for we gave
up the field to them. But looks to me as though Puss and Sandy
somehow don’t want to return to the aviation field. They’re veering
off as if they meant to go home.”
“Humph! guess that’s the best thing they could do anyhow, after
what happened!” grunted Andy.
“Meaning that sand bag they let drop?” remarked his cousin. “If I
were you, Andy, I wouldn’t say anything about that, unless asked.
Perhaps it was an accident, and they didn’t mean to do us any
harm.”
“Accident! You know just as well as I do, Frank Bird, that it was
meant, every time,” exploded the impulsive Andy. “It’s just the kind of
dirty trick Puss and his cowardly shadow are always playing on
those they don’t like.”
“Well, could you swear to it?” asked Frank.
“On general principles, yes I could,” answered the other, shaking his
head in an obstinate fashion.
“Then you saw Sandy unfasten the cord, or cut it loose?” Frank went
on.
“No—no, I can hardly go as far as that. He seemed to be handling
the bag, and I just guessed what he had in mind,” Andy admitted.
“Well, since we couldn’t prove our assertion it would be better to
keep mum on the subject. They’ll hatch up a story, and swear they
were just going to cast the bag over-board, thinking they might hit up
a faster pace, and didn’t see us below. You ought to know Puss
Carberry by this time; did you ever see him wanting a good excuse
for anything he did? And he can put on such an innocent face, too.
Let it drop, Andy. We won, and can afford to be generous, you
know.”
Andy could never stand out against this convincing tone of Frank’s.
“Oh! all right, if you say so, Frank, though I think you’re by long odds
too easy on the skunks. Why, if that bag had struck us in a certain
way, we might be as dead as herrings long before now. Makes me
shiver every time I look down. And after a fall of more than a
thousand feet, a fellow wouldn’t look good at his own funeral. But
since you say forget it, I’ll try to.”
When they hovered over the big field there was a whirlwind of shouts
that must have been pleasant music to these two young victorious
air voyagers returning from their recent exploit.
The next half hour was filled with plenty of excitement all around.
Frank had to guard his precious little monoplane from the crowds of
curious and applauding people who had witnessed their plucky race.
And the silver cup was indeed a beauty, well worth all the effort they
had put into their work. No one was more extravagant in praise than
Colonel Josiah Whympers, who toddled around with crutch and
cane, telling everybody he met what wonderful things Andy and
Frank were going to do some day. While most people were of the
opinion that he “put the cart before the horse” when using those two
names in that fashion, still they could forgive him, because Andy was
naturally everything to the doting old man.
Of course after that it was demanded that the Bird boys give a few
exhibition flights, just to let the gaping crowd see to what an
astonishing degree the modern aviator could guide his novel craft
through the air.
So Frank ascended to a height of nearly fifteen hundred feet, boring
his way upward after a fashion much in vogue among these pilots
who lead the world in aerial navigation; after which he descended in
spirals, being averse to attempting the risky stunt known as
volplaning, until he had learned the ropes better.
But it was all a grand circus for the thousands who viewed these
wonderful feats for the first time. And great was the uproarious
applause that greeted the young aviators after they had landed
again.
Before evening came the Bird boys once more went up, and headed
for the home field, tired but satisfied.
Dr. Bird had insisted that Frank come home for the night, since he
had been away so very long now.
“I guess there’s no danger about the monoplane,” Frank remarked,
as they locked the doors, and Andy for the twentieth time drew out
his recovered little monkey wrench to examine it carefully. “You know
Chief Waller nabbed those two men, Jules and Jean, and has them
locked up tight. Besides, now that the race is over, Puss and Sandy
will have no reason to want to injure our machine.”
“Perhaps not,” said Andy, “but Colonel Josiah ain’t going to take any
risks. He told me he had hired a watchman to sleep here in the shed
every night, just as long as we want. I’m going to hang around and
wait for him. I don’t trust Puss or his crony one little bit.”
“Well,” said Frank, as he prepared to depart on his wheel, “we’ve
had a grand day of it, old fellow; and I doubt if we ever see such a
great time again.”
“Just what I was thinking,” replied Andy, half regretfully, as though he
felt badly because all pleasant things must have an end. “There’ll be
no more races for us to win, and things will get mighty humdrum,
unless something turns up shortly.”
Little did either of the Bird boys, fresh from their victory of the air,
dream of the astonishing adventures that were soon to fall to their
portion, beside which those they had experienced, as narrated
between the covers of this book, would appear almost insignificant.
In good time the reader may be taken into our confidence, and
allowed to share in the knowledge of those stirring times that is in
our possession.
A few days later Frank and Andy happened to be among a group of
boys gathered on the campus in front of the high school building.
Although school had long since been dismissed for the summer
vacation, still the boys often congregated here by the famous
Bloomsbury school fence, to talk over things in general, such as
interested lads in a country town.
Baseball matters were being discussed, and the possibilities of a
good football season in the Fall. Frank and Andy were not so deeply
interested in these matters as usual though they did not see fit to tell
their friends just why.
Frank had been watching for an opportunity to carry out a little
scheme he had in mind, and which he had talked over with Andy,
Elephant Small, Larry Geohegan, and one or two other good fellows.
“Here he comes, Frank!” said Andy finally, as Puss Carberry and his
eternal shadow, Sandy Hollingshead, were seen approaching from
the direction of town.
Just as they were passing Larry stepped forward.
“I say, Puss, does this belong to you?” and he held out a card—none
other than the one which had been found in the hangar of the
monoplane the day after that trick of cutting the canvas of the planes
had been accomplished.
Puss was for once taken off his guard.
“Why, yes, I believe it does, Larry,” he said, immediately pulling out a
pack of fine cards. “You know I brought these up with me from the
city. See, it has the Indian on the back, and the words ‘Red Hunter.’
I’ll run them over, and see if the jack of spades is missing.”
He did so in an adept manner that told how accustomed he was to
handling such things.
“You see, it is missing,” he said triumphantly, “so I’ll thank you for
returning my black jack to me. Where did you pick it up, Larry?”
“Oh! you’re not indebted to me for its return,” declared Larry, turning
up his nose in disgust. “Frank here found it; he can tell you just
where.”
And Puss grew fairly scarlet, he hardly knew why himself, as he
turned his gaze upon the accusing face on the one whom he had
done so much to injure.
“You dropped it out of your pocket the night you visited our hangar,
and cut the canvas of our monoplane wings to flinders. I have been
saving it for you. Thank you, Puss, for admitting that you were the
author of that dirty trick,” and Frank turned his back on the confused
rogue.
Unable to frame a reply, Puss and his crony walked hastily away.
And before night the whole of Bloomsbury knew of what they had
been guilty; because Larry and Elephant refused to keep it to
themselves.
But it was not to be expected that this would cause such fellows as
Puss Carberry or Sandy Hollingshead to see the error of their ways.
On the contrary, it was only apt to make them the more bitter against
the Bird boys; and in time to come they would wish more than ever
that they could find some way by means of which they might injure
those who had so skillfully guided their little air craft to victory in the
race to the crest of Old Thunder Top.
Whether that opportunity would ever come, as well as many other
things in the line of adventure which were fated to befall the Bird
boys, must be left to another volume, which the reader, who has
followed our venturesome young aviators thus far, will be pleased to
know has already been issued under the title of “The Bird Boys on
the Wing; or, Aeroplane Chums in the Tropics.”
The End.