couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Randall Leeds (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (COUCHDB-1219) Send data
Date Fri, 15 Jul 2011 21:01:00 GMT

    [ https://issues.apache.org/jira/browse/COUCHDB-1219?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13066202#comment-13066202
] 

Randall Leeds commented on COUCHDB-1219:
----------------------------------------

There's a lot more to this than just making an HTTP client, and it doesn't require using Node.
couchjs implements the JavaScript view server on an embedded SpiderMonkey. It's written in
C. It has optional bindings to libcurl and a simple HTTP client is already made because it's
used for running the javascript tests from the command line. That client is not exposed in
any of the javascript functions CouchDB runs, but specifically exposed by the test harness.
It's easy enough to inject into the global namespace of any other JavaScript we run.

However, the complexity is more at a Couch API level and not at the level of creating this
HTTP client. For example, we would *not* want to expose the HTTP client to a map or reduce
function because that's removing the isolation that encourages determinism in views. We could,
possibly, expose it to a _list, or a _show is a object in the global namespace. We'd probably
want to allow it to be globally disabled, so hosted services like IrisCouch don't have to
worry about their servers being abused.

So, exposing a basic HTTP client is easy. Exposing a *good* one that a hosted CouchDB provider
can trust to obey CORS, not hog server resources, etc, is probably more tricky. Finding a
place to expose it is the hardest part for me and always sends me into a spiral of designing
the 2.0 view server. I'm not going to bikeshed that here, though I'm open to talking about
it on the lists. 

> Send data
> ---------
>
>                 Key: COUCHDB-1219
>                 URL: https://issues.apache.org/jira/browse/COUCHDB-1219
>             Project: CouchDB
>          Issue Type: New Feature
>            Reporter: Johnny Weng Luu
>            Priority: Minor
>
> With CouchDB the application server landscape has changed. The database can now handle
almost all logic that is data related, eg. validation, updating, formatting etc.
> It has a HTTP server for incoming requests but one thing I find missing is a HTTP client
that can push data to other servers.
> What I have to do right now for sending data to an external service is to setup a (node.js)
server that listens on the _changes feed.
> When there is a document meant for being delivered through email, node.js will send a
HTTP request to the (sendgrid) email service.
> I think it would make sense if CouchDB could make HTTP requests on document changes,
meaning we can eliminate that extra HTTP client layer just for sending the data.
> Why only have a HTTP server for data but not a HTTP client for data.
> I think this would make CouchDB even more suitable for couchapps.
> Now the app can be offline. We send an email and it will be sent by couchdb when Internet
connectivity is there.
> Also now the whole app can be replicated and every couchdb instance can make it's own
HTTP requests and we no longer have to have a node.js server just to support basic HTTP requests.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message