Re: Extension candidates

Dimitre Novatchev wrote:
> On Fri, Dec 12, 2008 at 8:37 AM, John Snelson <[email protected]> wrote:
> 
>>> Therefore, having a reference forever does not mean having the
>>> referenced object alive forever -- no major problem here.
>> That's not really true. The fact that nothing changes is an illusion that
>> the language environment plays on the program by carefully managing it's
>> access to the outside world (disk, network, etc.). It's entirely possible
>> that an XML file on disk could change during the time a program is running.
> 
> We have to draw the line somewhere: There could be an earthquake in SF
> or an ice storm in NE, or even a meteorite could collide with the
> Earth...
> 
> I think it is quite off-topic here to speak even about transactions.
> 
> It is totally up to an XSLT processor or its optimizer whether to keep
> something in memory, or to discard it and reassemble it later on
> as-needed basis. AFAIK, this is already the case with existing
> language processors. Therefore, to think of a "reference" only as a
> memory address is a very narrow approach that misses important cases
> of "reference".

I think you're missing the point. Generating plain old strings from 
nodes with a guarantee of being able to reverse the mapping will always 
leak resources of one sort or another, be it memory, disk space, file 
descriptors etc. Given that, I asked for a compelling use case for 
requiring this feature.

John

-- 
John Snelson, Oracle Corporation            http://snelson.org.uk/john
Berkeley DB XML:            http://oracle.com/database/berkeley-db/xml
XQilla:                                  http://xqilla.sourceforge.net