[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Fwd: Re: (IDEA) High performance XML/HTTP web database with performanceIndex (in Java)
Here is a message that was originally posted to the Java Apache mailing list
and the private answer of Marc Fleury (EJBoss). I would like to hear ideas
about this.
Thanks,
Falko
---------- Forwarded Message ----------
Subject: Re: (IDEA) High performance XML/HTTP web database with performanceIndex (in Java)
Date: Tue, 12 Oct 1999 15:06:36 -0700
From: "marc.fleury" <marc.fleury@ejboss.org>
Kevin,
sorry I answer privately, I don't want to hog the list while the vote is
going on (you know ethics and stuff).
The problem you raise is very close to one we have in EJBoss.
Let me explain. We are convinced that SQL and relational databases have
a huge impedence mismatch for the web and EJB in particular. The more
vocal EJB advocates dismiss the RDBMS altogether for the webOS. We are
also convinced that for people that write applications on the web, the
actual schema of the database should be transparent, i.e. you want to
"store" objects not necessarily deal with row and tables.
This is fine and dandy but the main problem we encounter is that the
finders required in the EJB spec (i.e. find all the destination object
in a travel app that have the "china" field) is very hard to do in OO.
Namely OQL might not be the most efficient way to do it. XML search is
*much* more attractive. You don't need to know the complete structure
AND logic of the object (it's class) to run searches on its persistence
representation (xml file).
One possibility raised by Falko (main developer behind Ozone from
Sofwarebuero) is persistent DOM representations (would match the future
java serialization engine output, XML as you know). Then we could do
searches from teh java layers on this persistent XML file (happens to be
an enterprise bean but could be anything as you point out). In short
this is a great idea, and we would use a finite subset of it, which
could be a first step.
SO this is more a FYI and a first "touch base" with some EJBoss folks .
regards
marc
>
> (overview)
>
> When the Internet started to experience massive growth one of the ways
> designed to conquer its massive amount of information was to setup a
> "search engine" which would spider the Internet or Intranet and
> calculate word counts from the HTML presentation layer. Later,
> databases started being written to the web application spectrum that
> added a gateway to the structured data.
>
> Later the W3C in its infinite wisdom published the XML specification to
> help split the difference and a standard give structure to documents
> that can easily be extended.
>
> (idea)
>
> Currently Java lacks an index server. One that parses a URL or
> filesystem and generates a META-Index (approx 40% original size) of the
> content and allows users to find a document within a filesystem or
> website based on a query.
>
> This lacks any structure as the user might search on something like
> China and either get the Country or the type of dishes.
>
> What I think is needed is a blend of the two. There is a *lot* of
> existing data that is HTML based (and will be for a while) that needs to
> be indexed within a 100% Java environment. Adding XML support to a Java
> Index Server would allow the user to enumerate a list of known DTDs and
> run a query on <country>China</country> (or XQL/XOQL/DOM) and obtain the
> doc (possibly HTMl after XSL re-format).
>
> (goals)
>
> I would just like to get feedback. There are some smart people here. ;)
>
> Kevin
>
> --
> ----------------------------------------------------------
> To subscribe: java-apache-on@list.working-dogs.com
> To unsubscribe: java-apache-off@list.working-dogs.com
> Archives and Other: <http://java.apache.org/main/mail.html>
> Problems?: jon@working-dogs.com
-------------------------------------------------------
--
______________________________________________________________________
Falko Braeutigam mailto:falko@softwarebuero.de
softwarebuero m&b (SMB) http://www.softwarebuero.de