Re: How to create a DBMS from scratch?

Christopher Browne <cbbrowne_at_acm.org> wrote in message news:<arf24j$hosci$1_at_ID-125932.news.dfncis.de>...
> The world rejoiced as emmettwa_at_onebox.com (Emmett) wrote:
> > Anthony.Youngman_at_ECA-International.com (Anthony Youngman) wrote in message news:<9a993dee.0211190404.1d0e356c_at_posting.google.com>...
> > I figure the problems being referred to in this post have to do with
> > making XPath queries available on data - not in storing XML.
>
> That's probably about the size of it.
>
> There's precisely nothing preventing people from storing XML data in
> an RDBMS VARCHAR field.
>
> The challenge comes in treating the set of data as a hierarchy and
> invoking hierarchy-related queries in an intelligent way.
>
> There are two fundamental issues vis-a-vis SQL:
> a) SQL queries return sets that are not inherently ordered, which
> means that there's not a totally natural mapping of ordered
> hierarchical data onto the DBMS;
>
> b) A set is not a tree, which is kind of the converse of a). SQL
> queries return sets, so that you can't get a tree out of a query.
>
> You can't put in a tree /as a tree/, and can't get it out /as a
> tree/... That makes using XML a bit troublesome.

Precisely ... which is why I said "Look at Pick". It is ABSOLUTELY TRIVIAL, given a XML record, to declare a Pick "table", import the record set, and then declare indices, search, whatever.

Apart from the terminology, the Pick mindset and the XML mindset are very similar, almost to the point of a simple 1-to-1 topological equivalence.

Okay, I'd need to study XML, but if I was wanting to manipulate XML my first instinct would be to pull it into a Pick database, manipulate it with the database tools, and then chuck it out as XML again. Now who in their right mind would even contemplate such an approach, unless the conversion between paradigms was a no-brainer? And I'm pretty confident that, even including the conversion costs, I could run an *extremely* tight race against "native XML" tools, maybe (indeed, probably) even beating them for speed on large or complex datasets.

Cheers,
Wol Received on Wed Nov 20 2002 - 17:02:20 CET