The document discusses the challenges and innovations in real-time collaborative editing using Conflict-free Replicated Data Types (CRDTs) within peer-to-peer (P2P) networks. It highlights the limitations of current centralized systems and proposes a decentralized model that allows for collaborative editing without being reliant on a fixed number of sites. The new model aims to enhance performance, resist censorship, and reduce infrastructure costs, making collaborative editing more efficient and scalable.

1. Real-Time
Collaborative Editing
With CRDTs

2. InfoQ.com: News & Community Site
• Over 1,000,000 software developers, architects and CTOs read the site world-
wide every month
• 250,000 senior developers subscribe to our weekly newsletter
• Published in 4 languages (English, Chinese, Japanese and Brazilian
Portuguese)
• Post content from our QCon conferences
• 2 dedicated podcast channels: The InfoQ Podcast, with a focus on
Architecture and The Engineering Culture Podcast, with a focus on building
• 96 deep dives on innovative topics packed as downloadable emags and
minibooks
• Over 40 new content items per week
Watch the video with slide
synchronization on InfoQ.com!
https://www.infoq.com/presentations/
crdt-tachyon-collaborative-editing

3. Purpose of QCon
- to empower software development by facilitating the spread of
knowledge and innovation
Strategy
- practitioner-driven conference designed for YOU: influencers of
change and innovation in your teams
- speakers and topics driving the evolution and innovation
- connecting and catalyzing the influencers and innovators
Highlights
- attended by more than 12,000 delegates since 2007
- held in 9 cities worldwide
Presented at QCon San Francisco
www.qconsf.com

4. Real-Time
Collaborative Editing
With CRDTs

5. Teletype for Atom

6. Make it as easy to code together
as it is to code alone.

9. Boulder to Gambettola - 8795 km
~30ms at the speed of light

11. a b c a b c

12. a b c a b c
insert(x, 1)
a x b c
insert(y, 2)
a b y c

13. a b c a b c
insert(x, 1)
a x b c
insert(y, 2)
a b y c
insert(x, 1)
a x b y c

14. a b c a b c
insert(x, 1)
a x b c
insert(y, 2)
a b y c
insert(x, 1)
a x b y c
insert(y, 2)
a x y b c

15. a b c a b c
insert(x, 1)
a x b c
insert(y, 2)
a b y c
insert(x, 1)
a x b y c
insert(y, 2)
a x y b c
a x y b c != a x b y c

16. 0 0 a b c

17. 0 0 a b c
00
01
insert(x, 1)
a x b c

18. 0 0 a b c
00
01
insert(x, 1)
a x b c
00
insert(y, 2)00
insert(y, 2)0 0
a b c

19. 0 0 a b c
00
01
insert(x, 1)
a x b c
00
insert(y, 2)00
insert(y, 2)0 0
a b c
insert(x, 1)insert(y, 2)
✨
01
✨
✨
insert(y, 3)
IT( , )

20. 0 0 a b c
00
01
insert(x, 1)
a x b c
00
insert(y, 2)00
insert(y, 2)0 0
a b c
insert(x, 1)insert(y, 2)
✨
01
✨
✨
insert(y, 3)
IT( , )
11
01 insert(y, 3)
a x b y c

24. Decentralized
Multi-User Undo
String-Wise Operations
Consistently Low Overhead

25. "Unfortunately, implementing OT sucks. There's a
million algorithms with different tradeoffs, mostly
trapped in academic papers. The algorithms are really
hard and time consuming to implement correctly. Wave
took two years to write and if we rewrote it today, it
would take almost as long to write a second time.”
– Google Wave Engineer

26. a b c a b c

27. a b c a b c
insert(x, ✨)
a x b c

28. a b c a b c
insert(x, ✨)
a x b c
insert(y, ")
a b y c

29. a b c a b c
insert(x, ✨)
a x b c
insert(y, ")
a b y c
insert(x, ✨)
a x b y c

30. a b c a b c
insert(x, ✨)
a x b c
insert(y, ")
a b y c
insert(x, ✨)
a x b y c
insert(y, ")
a x b y c

31. a b c a b c
insert(x, ✨)
a x b c
insert(y, ")
a b y c
insert(x, ✨)
a x b y c
insert(y, ")
a x b y c
a x b y c a x b y c

32. CRDTs
Conflict-Free Replicated Data Types

33. Data Consistency for P2P Collaborative Editing
Gérald Oster
Institute for Information Systems
ETH Zurich
osterg@inf.ethz.ch
Pascal Urso, Pascal Molli, Abdessamad Imine
Université Henri Poincaré, Nancy 1
LORIA
{urso,molli,imine}@loria.fr
ABSTRACT
Peer-to-peer (P2P) networks are very efficient for distribut-
ing content. We want to use this potential to allow not only
distribution but collaborative editing of this content. Exist-
ing collaborative editing systems are centralised or depend
on the number of sites. Such systems cannot scale when
deployed on P2P networks. In this paper, we propose a
new model for building a collaborative editing system. This
model is fully decentralised and does not depend on the
number of sites.
Categories and Subject Descriptors
C.2.4 [Computer-Communication Networks]: Distrib-
uted Systems—Distributed applications; H.5.3 [Informa-
tion Interfaces and Presentation]: Group and Organiza-
tion Interfaces—Collaborative computing, theory and models
General Terms
Algorithms, Design, Human Factors
Keywords
CSCW, Collaborative editing, Optimistic replication, Con-
currency control
1. INTRODUCTION
Currently, peer-to-peer systems demonstrated how they
can ensure scalable content distribution. In their survey [4],
Androutsellis-Theotokis et al. wrote:
“Peer-to-peer content distribution systems rely
on the replication of content on more than one
node for improving the availability of content,
enhancing performance and resisting censorship
attempts.”
We want to reuse these characteristics not only for content
distribution but also for content editing. Currently, P2P
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
CSCW’06, November 4–8, 2006, Banff, Alberta, Canada.
Copyright 2006 ACM 1-59593-249-6/06/0011 ...$5.00.
networks mainly distribute immutable contents, we want to
distribute updates on this content and manage collabora-
tive editing on it. We are convinced that if we can deploy a
group editor framework on a P2P network, we open the way
for P2P content editing. It means that all existing collab-
orative editing applications such as CVS and Wiki can be
redeployed on P2P networks and take advantage of availabil-
ity improvements, performance enhancements and censor-
ship resistance of P2P networks. For instance, Wikipedia [2]
is currently a collaborative encyclopædia that has collected
more than 4,700,000 articles in more than 200 languages.
Wikipedia has more than 50 million of page requests per day.
200,000 changes are made every day [1]. However, Wikipedia
needs a costly infrastructure to handle the load. Hundreds
of thousands of dollars are spent every year to fund the in-
frastructure. A P2P massive collaborative editing system
would allow to distribute the service, tolerate failures, im-
prove performances, resist to censorship and share the cost
of the underlying infrastructure.
Collaborative editing systems such as CVS or Wikis are
currently centralised and cannot be adapted to peer-to-peer
networks. Collaborative systems based on the operational
transformation approach [7, 23] can be decentralised.
However, existing algorithms such as GOTO [23], ABT[16]
and SOCT2 [21] rely on vector clocks to detect concurrent
operations. OT approach supposes that each operation is
immediately executed locally, stored in a local log and then
broadcast to other sites in order to be re-executed and stored
in their logs. A vector clocks is associated to each opera-
tion. A vector clocks [17] is an array of logical clocks, one
clock per site. It is used to detect the happened-before rela-
tionship and therefore the concurrency between operation.
It causes no problem if the number of sites is fixed and low
but if the number of sites grows, the size of the vector clocks
is unbounded. Thus, messages exchanged between sites will
grow as well as the size of local operation. Also, the time ef-
ficiency of operation on vectors clocks will decline as vectors
clocks grow. Clearly, vectors clocks prevent these algorithms
to scale and represent a serious bottleneck for their deploy-
ment on P2P networks.
In this paper we propose a new model called WOOT for
building group editors that is suitable for dynamic P2P sys-
tems. Compared to existing decentralised group editor mod-
els, the number of sites involved in group editing is not a
variable.
The remainder of this paper is organised as follows: Sec-
tion 2 describes the WOOT approach and details the con-
sistency model. Section 3 presents a formal definition of
WOOT
2006
Without
Operational
Transformation

34. a b c
1:1 1:2 1:3
globally unique id
site-id:sequence-number

35. a b c
1:1 1:2 1:3
1 1:2
2:1
1:1

36. a b c
1:1 1:2 1:3
1 1:2
2:1
1:1 1:2
3:1
1:32

37. a b c
1:1 1:2 1:3
1 1:2
2:1
1:1 1:2
3:1
1:32
4:1
1:2 3 3:1

38. 1:31:21:1
cba
1:31:21:1
cba

39. 1:31:21:1
cba
1:31:21:1
cba
1:1
2:1
1:21 2 1:3
3:1
1:2
2
3:1 1:31:21:1
cba
2:1
1
1:31:21:1
cba

40. 1:31:21:1
cba
1:31:21:1
cba
1:1
2:1
1:21 2 1:3
3:1
1:2
2
3:1 1:31:21:1
cba
2:1
1
1:31:21:1
cba
1:1
2:1
1:21
2
3:12:1
1
1:31:21:1
cba

41. 1:31:21:1
cba
1:31:21:1
cba
1:1
2:1
1:21 2 1:3
3:1
1:2
2
3:1 1:31:21:1
cba
2:1
1
1:31:21:1
cba
1:1
2:1
1:21
2
3:12:1
1
1:31:21:1
cba
1:2
3:1
1:32
a b c
1:1 1:2 1:3
1
2:1 3:1
2

42. 1:21:1
ba

43. 1:21:1
ba
1 1:2
1:1
1:1

44. 1:21:1
ba
1 1:2
1:1
1:1 2 1:2
2:1
1:1

45. 1:21:1
ba
1 1:2
1:1
1:1 2 1:2
2:1
1:1 1:1
3:1
1:23

46. 2a b

47. 2a b
3

48. 2a b
3

49. 2a b3

50. 2a b3
1

51. 2a b3
1

52. 2a b31

53. a b a b a b

54. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b

55. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b

56. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1

57. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b

58. 1
2
a b
3

59. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b

60. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b
a 3 1 2 b

61. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b
a 2 3 1 b

62. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b
a 2 3 1 b
a 1 2 b
a≺2≺b

63. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b
a 2 3 1 b
a 1 2 b
a≺2≺b
a≺3≺1

64. a b a b a b
a 2 b
a≺2≺b
a 1 b
a≺1≺b
a 1 b
a≺1≺b
a 3 1 b
a≺3≺1
a≺2≺b
a 2 3 1 b
a 1 2 b
a≺2≺b
a≺3≺1
a 3 1 2 b

65. 1
2
a b
3

66. 1
2
a b
3✕

67. 1
2
a b
3✕

68. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b

69. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b

70. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b
1 a≺2≺b

71. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b
1 a≺2≺b
a 3 1 2 b

72. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b
1 a≺2≺b
a 3 1 2 b
a≺2≺b1
a 1 2 b

73. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b
1 a≺2≺b
a 3 1 2 b
a≺2≺b1
a 1 2 b
2 a≺3≺1

74. a b a b a b
a 1 b
a≺1≺b1 1 a≺2≺b
a 2 b
1
a≺3≺12
a 1 b
a≺1≺b
a 3 1 b
1 a≺2≺b
a 3 1 2 b
a≺2≺b1
a 1 2 b
2 a≺3≺1
a 3 1 2 b

75. String-Wise
WOOT
2014

76. START
1:1
END
0:0 0:1
hello world

77. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there

78. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there !
3:1
1:1, 11 0:1, 0

79. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there !
3:1
1:1, 11 0:1, 0
START
1:1, 0
END
0:10:0 2:1
3:1
1:1, 5
hello, there world!

80. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there !
3:1
1:1, 11 0:1, 0
START
1:1, 0
END
0:10:0 2:1
3:1
1:1, 5
hello, there world!
1:1, 9 3:1, 0
2:2

81. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there !
3:1
1:1, 11 0:1, 0
START
1:1, 0
END
0:10:0 2:1
3:1
1:1, 5
hello, there world!
1:1, 9 3:1, 0
2:2
ms
3:1, 0
4:1
1:1, 11

82. START
1:1
END
0:0 0:1
hello world
1:1, 51:1, 5
2:1
, there !
3:1
1:1, 11 0:1, 0
START
1:1, 0
END
0:10:0 2:1
3:1
1:1, 5
hello, there world!
1:1, 9 3:1, 0
2:2
ms
3:1, 0
4:1
1:1, 11
hello, there worldms!START
1:1
END
0:10:0 2:1 1:1, 5 4:1 3:12:2

83. hello, there worldms! ENDSTART

84. hello, there worldms! ENDSTART

95. hello, there worldms!START
1:1
END
0:10:0 2:1 1:1, 5 4:1 3:12:2
tail
1:1,3
head
1:1,8

97. teletype-crdt

98. teletype-client
teletype-crdt

99. teletype-client
teletype-crdt
teletype-server

100. teletype
teletype-client
teletype-crdt
teletype-server

105. Future Directions

106. teletype.atom.io

107. Watch the video with slide
synchronization on InfoQ.com!
https://www.infoq.com/presentations/
crdt-tachyon-collaborative-editing