Subject Re: [Firebird-general] IBM moves the database goalposts - xml related
Author Lester Caine
Paulo Gaspar wrote:

> That is NOT what I was talking about.

Just pointing out that real USERS of XML do not see it as a METHOD of
storing data and don't expect it to be used as such. You state that XML
is used to STORE structured and semi-structured data. My customers say
it is used to ACCESS data, that can be stored how you like - and
particularly in fast relational databases, so that results can be found
quickly.

> I was talking about searching patterns in semi-structured data and you
> are talking about using XML as vehicle for interaction among multiple IT
> systems.

IBM are looking to combine two approaches, SQL and XML and as far as I
can see expand on BLOB by providing another view on those elements. Some
sort of indexing has to be provided on documents, down to a word level
if required, and the last thing you want to do is search a 10Gb flat
file for all occurances of SMITH. So you need to relate the data to what
you want fast searches on. The documents store well as XML data, and can
be managed in that format very well, scanning and OCR produce well
ordered data, which with a little help can be indexed better by adding
additional XML tags, and THIS is what is then passed between systems,
but the stored data is the content of tags in tables which can be
indexed. ( You can even store the raw scanned image as an XML element -
but some how it think it's better to save that as an external file with
an XML tag of the file name ;) )

I think one of the problems here is the distinctions made between
relational and hierarchical data. All of my data is hierarchical, and I
store it in a relational database. I can look for a name, a town, a form
type, or whatever. The vast majority of data is still stored in files
because that is the best way to manage it, and I could use normal
scanning search tools to find data that I have not 'relationalised'. All
that IBM are proposing, as far as I can see, is to make those files an
XML archive with further XML tools to navigate it - sounds a bit like
Longhorn to me - back to what the FILE SYSTEM should provide.

From the IBM article
"Or, to be more accurate, it will not just be a relational database."
They are looking at Longhorn, and what it was supposed to provide, so
they can provide an alternative. We can't compete there, but we can
improve things like XML handling and perhaps provide an XML format BLOB?
That is probably the first stepping stone.

Jim - I think you are already addressing most of this in different ways
- Are we heading to a properly managed file structure?

--
Lester Caine
-----------------------------
L.S.Caine Electronic Services