You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Eric Evans (JIRA)" <ji...@apache.org> on 2010/05/04 18:02:58 UTC

[jira] Issue Comment Edited: (CASSANDRA-1022) separate assignment of current keyspace from login()

    [ https://issues.apache.org/jira/browse/CASSANDRA-1022?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12863858#action_12863858 ] 

Eric Evans edited comment on CASSANDRA-1022 at 5/4/10 12:02 PM:
----------------------------------------------------------------

{quote}
Eric, please remember http://thread.gmane.org/gmane.comp.db.cassandra.user/1038/focus=1404 where you asked me to merge login() with setKeyspace() and now they are pulled apart again. What has changed? Shouldn't we be insisting on the "one keyspace per connection" concept? Or has the philosophy changed?
If set_keyspace() is to be used, it should be the one returning an AccessLevel. login() should return void. In other words, the AccessLevel is per-keyspace, not per-user.
{quote}

I remember; what has changed is (anecdotal) experience with how people are (or more importantly are not) using this, and how well the interface is holding up over time (read: it's not).

Personally,  I am  stronger in my convictions now that we should _not_ be rolling our own AAA, that this is not The Way, and I am seeking to make this (still experimental) API as optional as possible while working toward something better, (AVRO-341).

{quote}
Also, login() should no longer throw AuthorizationException since it doesn't do authorization, while set_keyspace() should throw it.
{quote}

{{set_keyspace()}} is the required call now, {{login()}} works exactly as before only it uses the keyspace specified in {{set_keyspace()}} instead of having one passed in (and {{set_keyspace()}} invalidates any previous {{login()}}).

This was the whole point of this issue, to divorce the function of assigning a keyspace from authentication.

{quote}
Please let me know if I need to implement these changes or if someone else will do it or if there are any questions.
{quote}

I don't see any changes that need to be implemented here.

      was (Author: urandom):
    {quote}
Eric, please remember http://thread.gmane.org/gmane.comp.db.cassandra.user/1038/focus=1404 where you asked me to merge login() with setKeyspace() and now they are pulled apart again. What has changed? Shouldn't we be insisting on the "one keyspace per connection" concept? Or has the philosophy changed?
If set_keyspace() is to be used, it should be the one returning an AccessLevel. login() should return void. In other words, the AccessLevel is per-keyspace, not per-user.
{quote}

I remember; what has changed is (anecdotal) experience with how people are (or more importantly are not) using this, and how well the interface is holding up over time (read: it's not).

Personally,  I am now much stronger in my convictions now that we should _not_ be rolling our own AAA, that this is not The Way, and I am seeking to make this (still experimental) API as optional as possible while working toward something better, (AVRO-341).

{quote}
Also, login() should no longer throw AuthorizationException since it doesn't do authorization, while set_keyspace() should throw it.
{quote}

{{set_keyspace()}} is the required call now, {{login()}} works exactly as before only it uses the keyspace specified in {{set_keyspace()}} instead of having one passed in (and {{set_keyspace()}} invalidates any previous {{login()}}).

This was the whole point of this issue, to divorce the function of assigning a keyspace from authentication.

{quote}
Please let me know if I need to implement these changes or if someone else will do it or if there are any questions.
{quote}

I don't see any changes that need to be implemented here.
  
> separate assignment of current keyspace from login()
> ----------------------------------------------------
>
>                 Key: CASSANDRA-1022
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1022
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.7
>            Reporter: Eric Evans
>            Assignee: Todd Blose
>            Priority: Minor
>             Fix For: 0.7
>
>         Attachments: CASSANDRA-1022.patch
>
>
> With the completion of CASSANDRA-714, it is now a requirement that login() be called, even when using the AllowAllAuthenticator (effectively disabling auth), since this is how the current/connected keyspace is set. These two disparate functions (assigning keyspace and authentication) should be disentangled.
> I propose that the keyspace argument be removed from calls to {{login()}}, and that a new method ({{use_keyspace(string)}}?), be added.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.