De meilleurs identifiants grâce au composant Symfony UID

Les identifiants, qu'ils soient techniques ou fonctionnels, sont essentiels car ils permettent d'immatriculer les différentes ressources que nous manipulons au sein de nos projets. Ils peuvent être représentés sous différentes formes : nombres entiers, chaînes de caractères, UUID, etc. Quel est le format le plus compréhensible ? Quel est le plus performant ? Au cours de cette conférence, je vous propose d'échanger sur ce thème passionnant, notamment en (re)découvrant le composant Symfony UID. À travers un retour d'expérience, j'aborderai également le sujet de la rétrocompatibilité dans le cadre de la migration d'identifiants existants.

Transcript

1. De meilleurs identifiants grâce au composant Symfony UID
2. Hello :-) Thomas Calvet - fancyweb
3. A “name” that identifies “something”
4. A value that uniquely identifies a resource
5. The resource unicity depends of the perimeter
6. An elementary and essential knowledge for developers
7. Identifiers everywhere
8. What makes a good identifier?
9. Is one identifier format always enough?
10. Functional identifiers vs technical identifiers
11. Integer identifiers
12. Easy to use, to understand, to generate (auto increment) and performant 👍
13. Lexicographically sortable 👍
15. Generable only by a central point (database) ⚠
16. Guessable (IDOR) /user/22 ⚠
17. Information leak ⚠
18. Provides no additional context 🤔
19. Random string identifiers
20. Generable offline, not guessable, can be prefixed 👍
21. Probably a little harder to use, to understand, to generate and a little less performant 🤔
22. Data fragmentation / not lexicographically sortable ⚠
23. UUIDs (Universally Unique IDentifiers)
24. 3 UUIDs (canonical format)
25. RFC 4122
26. A very very very big integer (128 bits)
27. Theoretically unique because of the number of combinations
28. Identify a UUID version
29. Time based UUIDs (version 1 and 6)
30. Not lexicographically sortable UUID v1
31. MAC address leak UUID v1
32. Lexicographically sortable UUID v6
33. Random node UUID v6
34. Time based UUIDs make good primary keys
35. Timestamp leak ⚠
36. Name based UUIDs (version 3 and 5)
37. Hash of a namespace and a name
38. Reproductible
39. Version 3 = md5 Version 5 = sha1 Prefer version 5
40. Random based UUID (version 4)
41. Only randomness
42. Version 4 UUIDs are good for random values (eg: tokens)
43. Version 2? ⚠
44. ULIDs (Universally unique Lexicographically sortable IDentifiers)
45. A ULID (canonical format)
46. Timestamp (48 bits) Randomness (80 bits) Close to a UUID v6
47. Lexicographically sortable
48. github.com/ulid/spec
49. “Compatible” with UUID
50. UUIDs and ULIDs are generable offline 👍
51. Storing UUIDs and ULIDs ⚠
52. UUIDs and ULIDs sum up 🤔
53. UUIDs vs ULIDs 🤔
54. Integers vs strings vs U[U|L]IDs 🤔
55. Using UUIDs and ULIDs in PHP
56. composer require symfony/uid
58. Integrated with other Symfony components
59. Not experimental anymore
60. Factories
61. Commands
62. Generate a UUID / ULID
63. Inspect a UUID / ULID
64. api.video study case
65. 2 letters prefix 128 bits of randomness to base 62 A current identifier
66. The same identifiers everywhere
67. User eXperience and database issues
68. A clear prefix A ULID converted to base 58 New functional identifier
69. ULIDs converted to UUID canonical format New technical identifier
70. In the code
71. Transforming new identifiers (video_XXX) to VideoIdentifier
75. Sum up
76. Issues solved vs increased complexity
77. Backward compatibility
78. Many possible solutions
79. Generating new identifiers for all existing resources
81. Transforming legacy identifiers (viXXX) to VideoIdentifier
84. And so many more interesting challenges ;-)
85. Thank you