RE: [Qexo-general] re: XqlServlet and Communication between XQL s cripts

[Tom Reilly]

+1 on CallContext having the ability to override the current
directory.

It might also be considered to get rid of the notion of a "current
directory" and have the CallContext API be something like:

URL getResource(String)

One example where this would be handy is a web application wrapped
in war where the URL may point to a jar entry.

> -----Original Message-----
> From: Per Bothner [mailto:address@hidden 
> Sent: Wednesday, January 29, 2003 12:49 PM
> To: Ivelin Ivanov
> Cc: address@hidden
> Subject: Re: [Qexo-general] re: XqlServlet and Communication 
> between XQL scripts
> 
> 
> Per Bothner wrote:
> > Presumably URLs should be relative to the "web-app", but I'm not
> > sure whether they currently are.  I thought they were ...
> 
> Ivelin Ivanov wrote:
> > Just noticed that you actually use a special QEXO function 
> to prepend the
> > absolute path to the relative:
> >  request-servlet-path(). This in effect reads the content 
> of the file, but
> > will not evaluate it if it is an XQL script, because it 
> will not go through
> > the servlet engine.
> 
> I had thought perhaps the servlet engine would change the "current
> directory" for different web applications?  I guess not, since then
> it would have to modify the user.dir system property, which is a
> global property.
> 
> I though about this some.  I think we need semantics more-or-less
> like you propose - I'm assuming that means the relative URIs are
> resolved relative to the web application root?
> 
> However, there are problems with the implementation.  For one,
> it makes Document depend on servlets.  For another, I think
> we need something more general.  Specifically that the "current
> directory" should be part of the per-thread state.  In Kawa
> that means hanging it off the current CallContext.  If the
> CallContext does not have current directory set, then we need to
> fall back to the global user.dir.  There are some complications
> dealing with thread inheritance, and it might make sense to
> also have an optional per-Interpreter current directory.  But
> perhaps those refinements can wait.
> 
> I also think I should implement the fn:base-uri(node) function.
> I've though about how to represent this information in a TreeList,
> and as soon as I have a breather I'll implement it.  Then we can
> implement the two-operand 'document', which will also help with
> this problem, especially if we also add a 'current-base-uri'
> (or whatever) function.
> -- 
>       --Per Bothner
> address@hidden   http://www.bothner.com/per/
> 
> 
> 
> _______________________________________________
> Qexo-general mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/qexo-general
>