Re: [Xsltforms-support] Problem with replace all

[Klotz, L. <Lei...@xe...>]

Alain,
    -----Original Message-----
    From: COUTHURES Alain [mailto:ala...@ag...] 
    Sent: Friday, January 29, 2010 9:16 AM
    To: Klotz, Leigh
    Cc: Thorsten Roggendorf; xsl...@li...
    Subject: Re: [Xsltforms-support] Problem with replace all
    
    Hello,
    > 
    > I think this is an XSLTForms limitation of submission
    > replace='all', but of course Alain would have to answer that.
    >
    
    The required instructions to interpret an XSLT
    processing-instruction (such as the one for applying xsltforms.xsl
    !) in the return of the submission are not yet implemented.
    
I still believe the <submssion method="get" replace="all"> could easily use
the same technique as the XSTLForms feature <submssion
method="xml-urlencoded-post"> and use the underlying HTML4 form
implementation for the GET instead of using XHR.  

This seems like a small amount of work, and would fix the main use case immediately.

    > I believe the problem is because replace='all' uses XHR, even when method='get'.
    
    XmlHttpRequest is the only way to manage errors, if any, from the
    initial XForms document.

I believe that having the success case of replace='all' work according to
spec is more important than having the failure case work.

Also, XForms 1.1 Test 11.4.9.b was recently changed to use replace='none' instead of
replace='all' because it seems likely that xforms-submit-error
handling is not possible in the replace='all' case in many processors, not just
Ubiquity and XSLTForms.

See http://lists.w3.org/Archives/Public/public-forms/2009Dec/0074.html
for discussion.

    > You can use an XSLTForms-only hack for POST, by doing
    > method="xml-urlencoded-post".  In this case, XSLTForms creates
    > an HTML4 form and serializes the XML into a single HTML4 field
    > named "postdata."  The disadvantge is that you must alter
    > server-side code to look for a URL Encoded POST containing a
    > parameter "postdata" and then extract that and parsing it as
    > XML.
    
    This hack was easy to implement. It guarantees that the browser
    knows it is a completely new page so memory leak should be under
    control. But there is no error management...
    
    By the way, at server-side, it is not necessary to check if the
    parameter is named "postdata" or anything else...
    

    > Personally, I think that XSLTForms could similarly implement submission method='get' 
    > when replace='all' by creating an HTML form element, and there would be no compensation 
    > needed and your example below would work.
    I consider that xf:load is better for that usage.
    >   Perhaps Alain has already explored this and found it unworkable for 
    > some reason, or perhaps it is just work that needs doing (and maybe 
    > you could help?)
    >   
    
    There is still plenty of work that needs doing in XSLTForms !
    
    Thanks!
    
    -Alain

Leigh.