[
https://issues.apache.org/jira/browse/CASSANDRA-10146?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14730492#comment-14730492
]
Benjamin Lerer commented on CASSANDRA-10146:
--------------------------------------------
I found that in {{ProtocolErrorTest}} the wrong exception was caught by
{{testNegativeBodyLength}} and {{testBodyLengthOverLimit}}. The direction was
setup wrongly in those test so the exception thrown was about that.
{{testNegativeBodyLength}} is infact useless because the length is an unsigned
int and by consequence cannot be negative. The checks for the bytes to discard
was apparently also wrong and no error was ever thrown.
I will also do a patch for 2.1 to fix those problems.
> Deprecate v1 and v2 protocol in 2.2, drop support in 3.0
> --------------------------------------------------------
>
> Key: CASSANDRA-10146
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10146
> Project: Cassandra
> Issue Type: Task
> Components: Core
> Reporter: Tyler Hobbs
> Assignee: Benjamin Lerer
> Labels: client-impacting, doc-impacting
> Fix For: 2.2.x, 3.0.0 rc1
>
>
> In 3.0, we would like to use frozen collections in the system keyspaces, and
> it seems likely that we will eventually want to use tuples or nested
> collections as well. Drivers that only support protocol versions 1 and 2
> will not be able to read these system keyspaces because they cannot decode
> those types.
> I think this is a good time to start deprecating and dropping support for old
> protocol versions. The v3 protocol was introduced in 2.1, so if we remove
> support for v1 and v2 in 3.0, that gives users two major versions to upgrade
> their drivers. Fortunately, upgrading drivers to a version that supports the
> v3 protocol is generally straightforward.
> The benefits of doing this are:
> * We can use new types in the system keyspaces
> * We can eliminate protocol-version-specific encoding and decoding of
> collections within Cassandra
> * Driver maintainers can eventually drop support for old protocol versions
> once all C* versions that support them are EOL
> To avoid a hard drop of v1 and v2 support in 3.0, I propose that we also
> officially deprecate these in 2.2. Unfortunately, we don't have
> protocol-level warnings until v4, so we can't use that to notify users of the
> deprecation, but the combination of a NEWS.txt entry, warning logs, and
> (potentially) driver-level warnings should suffice.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
