W3C home > Mailing lists > Public > xmlschema-dev@w3.org > December 2007

RE: Validate regular expressions

From: <noah_mendelsohn@us.ibm.com>
Date: Thu, 6 Dec 2007 21:36:00 -0500
To: "Michael Kay" <mike@saxonica.com>
Cc: "'Sascha Mantscheff'" <922492@gmx.de>, xmlschema-dev@w3.org
Message-ID: <OFC7E8E592.C2552A7C-ON852573AA.000D699B-852573AA.000E1C16@lotus.com>

Michael Kay writes:

> These arguments are not without merit - if schemas were 
> computationally complete, would you be able to trust them? My 
> own view, however, is that we should give users the power they 
> need to handle this kind of requirement.

That's really only part of the concern that some of us have about 
"unlocking that power".  The deeper concern, for me at least, is that the 
community hasn't been able to agree on an interoperable means of 
expressing those extensions.  Shall we all agree to write and distribute 
them in C#?  With what APIs into the processor, what assumptions about how 
to get at buffers to be validated, etc, etc.?  I'm not too sure my 
employer would be too happy about C#, I suspect Microsoft would prefer we 
didn't use Java, and many of us who also build very high performance C 
implementations wouldn't find either to be satisfactory.

Clearly, it's not just that powerful languages involve some risk:  it's 
that if my company, for example, defines an extension and supports it in 
Xerces, the chances that interoperable implementations will promptly 
appear in all other conforming processors are very, very low, unless the 
extension is dirt simple and tremendously valuable, in which case, 
everyone will indeed rush to clone it.  In more typical cases, users will 
start to write references to extensions into their schemas, and will then 
find that those schemas don't work with many tools, and when sent to 
business partners.  Worse, I fear that certain vendors might be tempted to 
build families of extensions that their tools would generate by default. 
Then we'd see a proliferation of schemas that were technically conforming, 
but for which control of the specifications was not suitably "open". We've 
seen this sort of thing before, and I think it's a much bigger problem for 
schemas than for, say, XQuery.  If I write a bunch of queries and find 
that when I mail them to some other company they don't work, well that's a 
disappointment, but how often to I mail around complex XQueries? 
Occasionally.   Conversely, exchanging schemas across organizations is a 
core use case.  That's why I'm so much more sensitive about this in the 
case of schemas than I am for some other systems like XQuery.


Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
Received on Friday, 7 December 2007 02:34:30 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:56:13 UTC