Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!usc!elroy.jpl.nasa.gov!ames!pasteur!ucbvax!mtxinu!sybase!tim@sybase.com
From: tim@sybase.com (Tim Wood)
Newsgroups: comp.databases
Subject: Re: Full Text Database Products
Keywords: full text storage retrieval
Message-ID: <8514@sybase.sybase.com>
Date: 27 Feb 90 18:27:08 GMT
References: <1990Jan30.204658.1822@comp.vuw.ac.nz> <1990Feb19.235547.8915@comp.vuw.ac.nz> <768@dgis.dtic.dla.mil> <8457@sybase.sybase.com> <771@dgis.dtic.dla.mil>
Sender: news@Sybase.COM
Reply-To: tim@sybase.com (Tim Wood)
Organization: Sybase, Inc.
Lines: 37

In article <771@dgis.dtic.dla.mil> jkrueger@dgis.dtic.dla.mil (Jon) writes:
>tim@ohday.sybase.com (Tim Wood) writes:
>>One can index BLOBs ...  by maintaining an identifier column value
>>of a simple datatype for each row containing a BLOB.  The indexing 
>>on the simple datatype is bound to be faster and more reliable than
>>the one-off indexing code a user writes for his/her BLOB datatype.
>I can show counterexamples (they happen to be textual datatypes).
>So it's not "bound to be".  Might be on average.  What it's bound to
>be is more subvertible, less safe, less productive.  ...

I'm interested in the counterexamples.  I question the
efficiency and robustness of directly indexing several-MB images/documents
in the DBMS with user-written code.  I don't understand the connection
between your points on integrity and productivity, except that less
robust code leads to more bugs, hence lower integrity and productivity.

>>Moreover, one can change the "ordering" of the BLOBs by changing 
>>the identifier values, whereas with directly-indexed BLOBs one would need
>>to change the type-specific indexing algorithm.  This amounts to changing 
>>part of the datatype (or object type) definition.  
>Sounds like a bad idea to me.  I prefer object ordering to be defined
>by the object.  E.g. you mind if I change ordering of ints?  Say by one 
>application without other applications' knowledge?

In making my point I was assuming that the object definition (in particular,
indexing order) might be subject to change.  For identifier columns, 
integrity rules could be defined to prevent changing the BLOB ordering,
although a strong binding between the BLOB and the i.d. columns would be 
missing.  The points you are making about integrity are valid.  I'm still 
learning about this topic, but I don't have a clear sense of the need or 
practicality of full access methods support on abstract data types.
-TW
---
Sybase, Inc. / 6475 Christie Ave. / Emeryville, CA / 94608	  415-596-3500
tim@sybase.com          {pacbell,pyramid,sun,{uunet,ucbvax}!mtxinu}!sybase!tim
		This message is solely my personal opinion.
		It is not a representation of Sybase, Inc.  OK.