[
https://issues.apache.org/jira/browse/COUCHDB-2497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14237949#comment-14237949
]
Robert Newson commented on COUCHDB-2497:
----------------------------------------
-1, _replicate is the only way to trigger a replication that doesn't cause
writes to a database, it's important. Especially as malformed replicator docs
that cause crashing replications can pound the _replicator hard, introducing
conflicts too.
_replicator being a database is the mistake, a serious operational one.
_replicate should have added "persistent":true as a flag, and all the
persistence part should have been hidden.
> Deprecate /_replicate endpoint
> ------------------------------
>
> Key: COUCHDB-2497
> URL: https://issues.apache.org/jira/browse/COUCHDB-2497
> Project: CouchDB
> Issue Type: Improvement
> Security Level: public(Regular issues)
> Components: HTTP Interface, Replication
> Reporter: Alexander Shorin
>
> We have two similar API to run replications. How about to reduce them single
> one? We cannot just return HTTP 301 on /_replicate to /_replicator for POST
> requests since in this case user must confirm the request submission whatever
> that means. But we can just reroute requests internally or try to use
> experimental [HTTP 308|http://tools.ietf.org/html/rfc7238] to deal with it.
> The motivation is the simplification of replication tasks management, since
> it doesn't simply to cancel active temporary replication as like as it could
> be done for persistent ones.
> Thoughts?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
