Development

Why CouchDB?

September 21, 2011

This article was contributed by Joe 'Zonker' Brockmeier.

Why should developers use CouchDB? It all comes down to replication, says Benjamin Young. Data is "lonely," trapped in proprietary formats and cloud silos, or remote applications that require dependable Internet connections. When the power goes out, when you want to work on the plane, or when you just want control over your data — Young says you want CouchDB.

Young was speaking at the Strange Loop 2011 conference on Tuesday, September 20 in St. Louis. The mission? To convince some of the 900-plus developers at the conference to consider CouchDB as database to power their applications. It wouldn't be very convincing if Young wasn't a fan himself, of course. Young's history with CouchDB goes back to work with BlueInk, a content management system for building simple sites for small companies. BlueInk used CouchDB, which caught the attention of CouchOne (now Couchbase, following a merger of CouchOne and Membase). BlueInk was released as open source a year ago when Young joined CouchOne.

What is CouchDB?

After giving a bit of his history, Young moved on to talking about CouchDB and what it actually is. CouchDB is a document storage database written in Erlang. It is open source, and part of the Apache project — though several different implementations are offered by a handful of companies like Couchbase. It's one of the current crop of NoSQL, or coSQL if you prefer, databases.

Documents in CouchDB are stored in JavaScript Object Notation (JSON) format, plus attachments if you have binary data (photos, for example) to put into the database. Queries are written using JavaScript. It's a distributed database that replicates statelessly. That is, as Young describes, "it can be interrupted and doesn't care." CouchDB counts the number of changes that have been made since it last connected to another instance, and picks up where it left off.

Young described a scenario of deploying an application using CouchDB in rural Africa. The users might only have Internet a couple of hours per day in rural Africa. If they were using a cloud-based application that depends on a constant connection "they'd be completely screwed". Using CouchDB, the data can be easily synced during the window of connectivity without any hassle. The database doesn't have to keep a constant state, CouchDB operates on the principle that it will be "eventually consistent".

As the scenario Young described implies, CouchDB is not limited to running on a server. It will run on a desktop, server, or even on a mobile device. Any device running CouchDB can communicate and send data to another instance of CouchDB, so replication can be peer-based or follow a server-client model. Replication can also be filtered, so that an instance only receives some of the information relevant to its local application. For instance, Young mentioned you might use CouchDB for distributing shipping information — the primary instance of CouchDB getting all of the order information, and filtering the appropriate orders to the users who are filling some of the orders.

Another argument for CouchDB? Young says "you already know the API," because it's basically HTTP. CouchDB provides a RESTful API with GET, PUT, POST, DELETE, as well as a non-standard extension to HTTP, COPY. The COPY API does what one would expect: it duplicates the contents and attachments of a document to a new document using a new name, and without requiring it to be retrieved first.

Web 2.5: Work Locally, Sync Globally

The big reason for choosing CouchDB according to Young? Replication. Like many in the open source community, Young doesn't express a lot of love for cloud services — which he describes as "managed and terms-of-serviced and fascist". That's not to say Young doesn't want to collaborate with his friends and co-workers. Far from it, but Young argues for a return to a "decentralized and very organic" Web. He calls it "Web 2.5" (because "Web 3.0 is already taken").

Web 2.0 gave us "cloud powers" but took away ownership of data. "For cloud powers we traded ownership, privacy, security, safety, stability... we need to get all those things back, to get that we need replication. Not only to move data, but applications". Web 2.5, says Young, is a situation where you have the browser and server "and then the .5 lives in the cloud, and that's a service that watches for changes in your CouchDB." Data a user chooses to share can still be synced up to cloud services, but it lives on the users' machine as well. "I as a person shouldn't have to keep everything that is me in a little package in the cloud... I can work locally on my laptop and then if I need to share I can go somewhere people are."

Get Started

Young suggests that developers interested in CouchDB start at Iris Couch as the fastest way to get started. He also suggested checking out the Apache project, and its mailing lists. For more immediate questions and feedback, Young noted that there was a decent community of CouchDB developers on Freenode on the #couchdb channel as well as in the #couchbase channel.

Comments (2 posted)

Brief items

Quotes of the week

Making software flexible, safe, and correct is hard work. Wish I still had my magic wand from kindergarten.

-- Greg Ward

(function(){var k=[];return function j(){k.push(i);j();}})()();

On every call, the function j attempts to add the contents of variable i out of the global scope to the enclosed array k. The eagle eyed will notice that I have not included a safety check here so that if i is undefined, the code will error out and the recursion of j will fail. This is intentional. The variable i represents input from other people, and I want to remember that I should always seek the advice and opinions of those around me. I know that I cannot grow in isolation, hence the catastrophic consequences of failing to garner the contributions of others. As I mentioned above, the k array represents knowledge or experience, and this part of the code also reminds me that I should always learn from the input I elicit.

Jim Sangwine

Not spending as much time sitting in meetings and fighting with other vendors is one of the competitive advantages PostgreSQL development has vs. the "big guys". There needs to be a pretty serious problem with your process before adding bureaucracy to it is anything but a backwards move. And standardization tends to attract lots of paperwork. Last thing you want to be competing with a big company on is doing that sort of big company work.

-- Greg Smith

Comments (none posted)

GNU Health 1.3.3 released

GNU Health is a hospital information system with modules for electronic medical record, hospital information systems, and health information systems. The 1.3.3 release is out; beyond a name change (from GNU Medical) and improved installation scripts, it includes: "major improvements in the Laboratory module. Now we have automatic visual alerts, possibility to exclude analytes from the analysis, qualitative and quantitative testing."

Full Story (comments: none)

Greedy register allocation in LLVM 3.0

Readers interested in the grungy details of compilers may enjoy this LLVM project blog entry on their new register allocation algorithm. "The new basic allocator does away with linear scan's dependence on visiting live ranges in linear order. Instead, it uses a priority queue to visit live range in order of decreasing spill weight. The active list used for interference checks is replaced with a set of live interval unions. Implemented as a B+ tree per physical register, they are an efficient way of checking for interference with already assigned live ranges. Unlike the active list, live interval unions work with any priority queue order."

Comments (7 posted)

Openshot 1.4

Version 1.4 of the OpenShot video editor has been announced. Changes include timeline improvements, a "more stable" effects engine, new color correction tools, more animations and transitions, and more. (LWN reviewed OpenShot in February).

Comments (none posted)

Newsletters and articles

Development newsletters from the last week

Bzr team weekly notes (September 20)
Caml Weekly News (September 20)
LibreOffice development summary (September 21)
PostgreSQL Weekly News (September 18)

Comments (1 posted)

Bray: Preparing for [Android] Handsets

~~Tim Bray~~ Scott Main notes that the upcoming Ice Cream Sandwich (ICS) release will support more form factors compared to the current Honeycomb. "Early this year, Honeycomb (Android 3.0) launched for tablets. Although Honeycomb remains tablets-only, the upcoming Ice Cream Sandwich (ICS) release will support big screens, small screens, and everything in between. This is the way Android will stay from now on: the same version runs on all screen sizes. Some Honeycomb apps assume that they'll run only on a large screen, and have baked that into their designs. This assumption is currently true, but will become false with the arrival of ICS, because Android apps are forward-compatible - an app developed for Honeycomb is compatible with a device running ICS, which could be a tablet, a phone, or something else."

Comments (18 posted)

Page editor: Jonathan Corbet
Next page: Announcements>>