[
https://issues.apache.org/jira/browse/KAFKA-15754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17781151#comment-17781151
]
Colin McCabe commented on KAFKA-15754:
--------------------------------------
I ran `kafka-storage.sh random-uuid` 10,000 times and got the following
distribution of first characters:
{code}
IDs starting with 0 : 166
IDs starting with 1 : 174
IDs starting with 2 : 135
IDs starting with 3 : 172
IDs starting with 4 : 155
IDs starting with 5 : 154
IDs starting with 6 : 152
IDs starting with 7 : 172
IDs starting with 8 : 170
IDs starting with 9 : 166
IDs starting with A : 147
IDs starting with B : 161
IDs starting with C : 172
IDs starting with D : 158
IDs starting with E : 164
IDs starting with F : 164
IDs starting with G : 146
IDs starting with H : 156
IDs starting with I : 166
IDs starting with J : 172
IDs starting with K : 177
IDs starting with L : 143
IDs starting with M : 171
IDs starting with N : 144
IDs starting with O : 157
IDs starting with P : 162
IDs starting with Q : 144
IDs starting with R : 157
IDs starting with S : 161
IDs starting with T : 158
IDs starting with U : 174
IDs starting with V : 166
IDs starting with W : 166
IDs starting with X : 159
IDs starting with Y : 165
IDs starting with Z : 161
IDs starting with _ : 159
IDs starting with a : 145
IDs starting with b : 169
IDs starting with c : 166
IDs starting with d : 171
IDs starting with e : 162
IDs starting with f : 154
IDs starting with g : 132
IDs starting with h : 152
IDs starting with i : 136
IDs starting with j : 166
IDs starting with k : 159
IDs starting with l : 156
IDs starting with m : 154
IDs starting with n : 155
IDs starting with o : 154
IDs starting with p : 158
IDs starting with q : 141
IDs starting with r : 165
IDs starting with s : 154
IDs starting with t : 162
IDs starting with u : 146
IDs starting with v : 161
IDs starting with w : 164
IDs starting with x : 154
IDs starting with y : 164
IDs starting with z : 154
{code}
No IDs were generated with a first character of `-`, as expected.
> The kafka-storage tool can generate UUID starting with "-"
> ----------------------------------------------------------
>
> Key: KAFKA-15754
> URL: https://issues.apache.org/jira/browse/KAFKA-15754
> Project: Kafka
> Issue Type: Bug
> Affects Versions: 3.6.0
> Reporter: Paolo Patierno
> Assignee: Paolo Patierno
> Priority: Major
>
> Using the kafka-storage.sh tool, it seems that it can still generate a UUID
> starting with a dash "-", which then breaks how the argparse4j library works.
> With such an UUID (i.e. -rmdB0m4T4–Y4thlNXk4Q in my case) the tool exits with
> the following error:
> kafka-storage: error: argument --cluster-id/-t: expected one argument
> Said that, it seems that this problem was already addressed in the
> Uuid.randomUuid method which keeps generating a new UUID until it doesn't
> start with "-". This is the commit addressing it
> [https://github.com/apache/kafka/commit/5c1dd493d6f608b566fdad5ab3a896cb13622bce]
> The problem is that when the toString is called on the Uuid instance, it's
> going to do a Base64 encoding on the generated UUID this way:
> {code:java}
> Base64.getUrlEncoder().withoutPadding().encodeToString(getBytesFromUuid());
> {code}
> Not sure why, but the code is using an URL (safe) encoder which, taking a
> look at the Base64 class in Java, is using a RFC4648_URLSAFE encoder using
> the following alphabet:
>
> {code:java}
> private static final char[] toBase64URL = new char[]{'A', 'B', 'C', 'D', 'E',
> 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T',
> 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i',
> 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x',
> 'y', 'z', '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '-', '_'}; {code}
> which as you can see includes the "-" character.
> So despite the current Uuid.randomUuid is avoiding the generation of a UUID
> containing a dash, the Base64 encoding operation can return a final UUID
> starting with the dash instead.
>
> I was wondering if there is any good reason for using a Base64 URL encoder
> and not just the RFC4648 (not URL safe) which uses the common Base64 alphabet
> not containing the "-".
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
