RE: CCL proposal (quotes)

Well, I'll give the politically incorrect explanation.

Z39.58 was developed to standardize the command interface to bibliographic
services.  I'm talking about Dialog, ORBIT and BRS along with others.  It
included the logon and logoff and index browse commands as well as the
Search command.  It was never adopted by any of those systems (at least not
by the ones I used.)

Here at OCLC, we tried to use the Z39.58 search grammar and found it to be
ambiguous.  So, we had to patch it to make it work.

So, we have a flawed, unimplemented standard that was allowed to slip
gracefully into its grave at review time.

Ralph

-----Original Message-----
From: Mike Taylor [mailto:mike@tecc.co.uk]
Sent: Thursday, May 09, 2002 11:07 AM
To: rden@loc.gov
Cc: www-zig@w3.org
Subject: Re: CCL proposal (quotes)


> Date: Thu, 09 May 2002 10:28:27 -0400
> From: Ray Denenberg <rden@loc.gov>
> 
> > Ray,
> >
> > This _is_ a joke, isn't it?  Isn't it?  Huh?  Huh?
> 
> No, it's a serious suggestion.

Oh.

> It appears to me that there isn't much implementation of the
> IEEE1003.2 nor the Z39.58 regular expression.  The reason given in
> the IEEE case is that it's too complex. The counter suggestion has
> been to implement a subset, and the counter argument to that is
> "interoperability problems".  The reason given in the Z39.58 case is
> that it doesn't have sufficient functionality.

Surely both of these problems would be overcome by implementors who
really needed regexp functionality?

> I have a real distaste for the prospect of trying to patch up a
> Z39.58 definition (when Z39.58 doesn't even exist anymore).

I agree with Alan that we need to know _why_ Z39.58 has gone away.  If
it's been replaced -- either by a new ANSI/NISO standard, or by a
reference to one of ISO's, then we should just fix the relation-104
explanatory prose to be a reference to the appropriate new standard.
_If_ it turns out that there's no replacement, or that it's fatally
flawed, then that _might_ be the time to think about hand-crafting a
replacement.

(BTW., as I went to check the value of the Z39.58 regular expression
Relation attribute, I noticed that the internal links within
http://lcweb.loc.gov/z3950/agency/defns/bib1.html
still don't work, and that Position and Completeness attributes are
still not listed.)

> Given that there isn't much implementation investment in either (and
> if I'm wrong on that I expect that people will speak up) it seems
> reasonable to suggest [...]

... that there's not much real requirement for them?

> [...] that we define a regular expression for Z39.50, compatible
> with existing regular expressions, that encompasses the requirements
> of the Z39.50 implementor community.
> 
> So, what do you suggest, Mike?

Sorry, Ray, that I can't be more positive; but what I suggest is this:
that upgrading from the current set of four largely unimplemented
pattern-matching Relations to five largely unimplemented
pattern-matching Relations would not be a big win.

(And yes, I _know_ that stance is at odd with my request for an
SQL-"like" Relation attribute ... but I don't care :-)

 _/|_	 _______________________________________________________________
/o ) \/  Mike Taylor   <mike@miketaylor.org.uk>   www.miketaylor.org.uk
)_v__/\  "You question the worthiness of my code?  I should kill you
	 where you stand!" -- Klingon Programming Mantra

Received on Thursday, 9 May 2002 12:30:10 UTC