RE: CMS tutorials

JC,

An alternative is to storing files as BLOBS and content as text or XHTML/XML
is to store files, images, etc. into the directory structure and the rest of
the content into the RDBMS or a combination of both.  I've seen this work
perfectly well - while still keeping in mind the open architecture approach.

Best,

Graham

-----Original Message-----
From: J C Lawrence [mailto:[email protected]] 
Sent: Sunday, December 21, 2003 7:31 PM
To: graham.zahoruiko-aO66/[email protected]
Cc: 'Andrew.Bean'; cms-list-/[email protected]
Subject: Re: [cms-list] CMS tutorials 


On Sun, 21 Dec 2003 18:06:06 -0500 
J Graham Zahoruiko <J> wrote:

> In my note I stressed the need to leverage the Open RDBMS as the point 
> of integration for all content systems (CMS) combined with an Open 
> publishing model (no API - use your own tools, scripting languages, 
> application servers, etc.).

This assumes that an RDBMS is an appropriate and beneficial storage model
for your back end data.  This can be true, is often true, but is not
necessarily true.  The relational table model is but one possible view of
data and is frequently enough a lousy fit for CMS data (eg RDBMS constraints
and requirements for is-type-of models).  I generally find that a mix of
flat file-system data (essentially path-named blobs), RDBMS and OOBMS do
what I need.  RDBMS systems also tend to be extremely fragile and
expensively brittle once elements of data and schema versioning enter the
picture.  A near guaranteed tarbaby recipe, yet an area that data-property
partitioned systems (using the filesystem or an OODBMS hierarcy or a custom
DBMS) can handle fairly pleasantly.

> The NET/NET is, the RDBMS is for integration & storage, the CMS is for 
> managing content (creation, workflow, versioning), and the Portal is 
> leveraged for publishing information in a specific format for = 
> consumption.

None of the current RDBMSes handle blobs well.  None handle anything other
than trivial partitioning schema worth a damn (just try and have
data-property defined partitioning schema under Oracle).  Often these are
simple areas where filesystem-based methods work well.  Dynamic type systems
where you don't know the library of possible types or their relations in
advance work very badly in RDBMS worlds, struggle but can get by in standard
OODBMS systems, and tend to work very nicely in runtime morphic systems like
the ZODB.

-- 
J C Lawrence
---------(*)                Satan, oscillate my metallic sonatas.
[email protected]               He lived as a devil, eh?
http://www.kanga.nu/~claw/  Evil is a name of a foeman, as I live.

--
http://cms-list.org/
please trim your posts.