couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jens Alfke (JIRA)" <>
Subject [jira] [Commented] (COUCHDB-1824) Official documentation of replication algorithm?
Date Sat, 28 Sep 2013 18:03:02 GMT


Jens Alfke commented on COUCHDB-1824:

This document has a lot of grammatical errors; would you like a detailed proofreading list?
Or actually it would make things easier if you put the document on the wiki where I could
edit it.

Technical issues:

>1. Assign an unique identifier to the source Database. Most of the time it will be the

No, that won't work. The identifier has to take into account other characteristics of the
replication like the filter name and parameters, and the optional list of doc IDs to replicate.
Otherwise multiple replications between the same two endpoints will collide and mess up each
other's state.

It's also a good idea for security to add in a randomized value that's only known to the local
server, so the document ID isn't guessable by others (since there are no read or write access
controls on _local documents.)

>2. Save this identifier in a special Document named _local/<uniqueid> on the Target

"Save this identifier" is misleading. You don't save the identifier, you use the identifier
as the document ID. What you save is the last remote sequence ID.

Also, the checkpoint gets saved at both the source and target, and its contents need to match,
to detect either database being rolled back to a backup, for instance.

>3. Get the Source changes feed

This is missing significant details that were in my document, like the use of style=all_docs
(without which you can't find conflicting revisions.)

>5. ... Also don't forget to get attachments that aren't already stored at the target.

Way too vague. See my doc for the details.

>After the group of revision is stored on the Target Database, save the new Checkpoint
on the Source database.

You need to describe _what_ to save in the checkpoint.

>Even if some revisions have been ignored the sequence should be take in consideration
for the Checkpoint.

I actually can't figure out what this sentence is trying to say. The grammar makes no sense.


* My note starting "Replication does not transfer obsolete revisions of documents, only the
current ones..." is fairly significant and it would be good to add similar content to your

* The reference to the TouchDB iOS wiki is obsolete; please change it to
. Thanks.
> Official documentation of replication algorithm?
> ------------------------------------------------
>                 Key: COUCHDB-1824
>                 URL:
>             Project: CouchDB
>          Issue Type: Documentation
>          Components: Documentation
>            Reporter: Nathan Vander Wilt
>            Assignee: Alexander Shorin
>             Fix For: 1.5.0
> Though it's in some ways an internal detail, it might be nice to provide a canonical
description of CouchDB's replication protocol (algorithm, really) in the documentation. See
links at:

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message