incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Anderson <jch...@apache.org>
Subject Re: stale document references
Date Wed, 05 Aug 2009 23:13:22 GMT
On Wed, Aug 5, 2009 at 7:11 AM, william cinnebar<merz.online@gmail.com> wrote:
> I have some data I want to store.
> Initially i was using common rdbms and querying with sql because this method
> of storing data is popular and very well supported.
> Then i found that xml is a better way to explain/map/store the data.
> I have since found that json is better still and overcomes a lot of the
> problems that occur using xml.
>
> I like couchdb for json documents, js querying and the erlang.
> Iv read through much of the couchdb documentation.
>
> It seems from what iv read that it is better to minimise the document size
> due to the way couchdb incrementally performs updates to the data.
> It also seems there are methods to efficiently collate views of related
> documents and query the data.
>
> I have come across the assertion that documents should be self contained but
> what exactly is meant by this?
>
> Perhaps I have missed some important information?
>
> Isnt normal/denormal relative to any particular use case of a document?
> A document that is in a normalised state for one use may be in a
> denormalised state for another use without any change to the document.
> What would be achieved by duplicating data for each use case?
>
> At what point is a document rendered denormal in couchdb terms and isnt it
> just a matter of perspective?

Exactly. By self-contained it is meant as a contrast to the fully
normalized relational model. It's also meant not as a rigorous
criteria, just a way of thinking about things.

>
> How do couchdb document references go stale?

CouchDB documents have IDs, which are used for saving and loading
them, but aside from that there are no references. There is certainly
nothing that can go stale.

If you record the id of another document somewhere, and that document
is then deleted, the recorded id doesn't change. It's just like links
between pages on the web.

> Is the replication process unstable?

It should be robust and reliable.

> Does couchdb forget stuff for some reason that is not documented (or that i
> have missed?)

CouchDB doesn't forget stuff. It does, however, only really keep track
of the current version of a document. Compaction removes non-current
versions of documents. Also, old versions are not replicated.

>
> The couchapp/sofa model does not appear to be relevant to the model i am
> working with.
> There seems to be no reason in the core documentation why the model I am
> working with could not be implemented using couchdb.
>
> Also I apologise id like to request further clarity: what exactly is meant
> by the term 'standalone application' where couchapp/sofa is concerned?
>

A standalone application is one that can be deployed to just CouchDB,
and accessed via a browser. Most uses of CouchDB are not for
CouchApps, however, if you can make your application a CouchApp this
means that you users can access it via the web, or run it on a local
CouchDB, without installing any other software.

Hope that helps.

Chris



-- 
Chris Anderson
http://jchrisa.net
http://couch.io

Mime
View raw message