Re: Major features of 9.0? - Mailing list pgsql-advocacy

fw@deneb.enyo.de (Florian Weimer) writes:
> * Josh Berkus:
>
>>> Isn't hstore key/value pair data, rather than schema-less.
>>
>> Well, when the "NoSQL" people talk about "schemaless", that's what they
>> mean.
>
> Some of them have got arbitrarily nested documents involving
> sequences, booleans, sequences of string/document pairs, strings, and
> floats.  Positioning PostgreSQL's simple key/value support against
> that could be a PR mistake. 8-)

And that points at the problem that "NoSQL" isn't a uniform thing; it's
more a reaction against than a specific thing.

<http://www.dbms2.com/2009/12/12/legit-nosql-key-value-store/>
Here, it is pointed at as the conjunction of three things:
 a) Key/value stores (as some proposed here)
 b) Document managers (as you describe)
 c) DBMS imitations built atop something like Hadoop (which seems to
    revert into SQL)

Monash suggests that the "legitimate" part is the distributed key/value
stores.  Those using CouchDB won't be thrilled with that!  :-)

A perhaps-related spectrum of questions comes in here...
<http://www.dbms2.com/2009/12/11/nosql-q-and-a/>

I think we'd need to have an answer for the CouchDB users in order to
realistically make the claims.

Note that there's a new Ubuntu thing called "Ubuntu One"
<http://en.wikipedia.org/wiki/Ubuntu_One> which is a "cloud-based"
storage/sync service analogous to Apple's iDisk/MobileMe (probably the
most familiar thing of this sort).  You can sync a variety of sorts of
data, including the CouchDB databases containing contact info that can
be used by various applications, using a substrate called "Desktop
CouchDB" <https://launchpad.net/desktopcouch>.  See
<http://www.freedesktop.org/wiki/Specifications/desktopcouch> for more
details on how they're trying to use it.

What they're doing with CouchDB is fairly much "morally equivalent" to
what one would do using LDAP.  With the difference that what I've seen
of CouchDB seems at least *somewhat* usable, whereas any time I touch
LDAP I feel like I'd rather poke burning needles in my eyes...  :-(

In any case, let me reiterate...  I think we'd need to have a way to
duplicate the sort-of-schema-free model of CouchDB, and to manipulate it
reasonably conveniently.

Abstractly speaking, none of these things are truly "schemaless" - what
they have done is to decide on a particular structure:

 - KV pair databases have the super-rigid schema that you have a single
   table with key/value pairs stored in it

 - Document databases are more or less like any/all of:
    a) LDAP hierarchies
    b) Forests of XML documents
    c) Bloated Goats, I mean, Lotus Notes

It would be an interesting thing to have an API of some sort to enable
conveniently storing document databases.  That's not going to be in by
9.0 :-).

It would be well worthwhile to have discussion of this at PGCon.
--
let name="cbbrowne" and tld="acm.org" in name ^ "@" ^ tld;;
http://linuxdatabases.info/info/slony.html
"Fashion is a form of ugliness so intolerable that we have to alter it
every six months."  -- Oscar Wilde