Re: [xml] how do I...

[Sebastian Kuzminsky <seb highlab com>]

Daniel Veillard <veillard redhat com> wrote:
> On Tue, May 24, 2005 at 02:47:12PM -0600, Sebastian Kuzminsky wrote:
> > I've got a couple of processes that trade XML messages over the network.
> > The sender writes each XML message on a single newline-terminated line.
> > The receiver uses select and read, and reads until it gets a '\n',
> > then passes the line to xmlParseMemory and validates the resulting doc
> > with xmlValidateDtd.
> >
> >
> > This works well, but it's a little annoying to have to use '\n' as the
> > message separator.  What I'd really like is to let the sender spread
> > its message out over several lines if it wants, and have the receiver
> > detect the end-of-message without any gross hacks.
>
>   By the definition of XML this is not possible. Packing multiple
> XML document on a single stream without out of band markers is a frequent
> but huge design flaws. The demonstration is obvious for anybody who
> read the 2 first pages of the XML standard:
>
>    http://www.w3.org/TR/REC-xml/#sec-well-formed
>
>    First production of the XML specification:
>    [1] document ::= prolog element Misc*
>
>  Misc* means there is no potential limit to the number of Misc element at 
> the end, and not finding one is a fatal error. 
>  The direct result from this is that the parser must be told that the document
> is finished. And libxml2 API being strictly conformant does not offer APIs
> for what you want. 
>  I strongly suggest you redesign your network format to include markers
> or documents size in the pipe, the current state sounds broken.


Ok, thanks Daniel.


My network format already includes an end-of-document marker which never
appears inside the document ('\n'), so I guess I'm standards-compliant,
if only by dumb luck.  :)


-- 
Sebastian Kuzminsky
"Marie will know I'm headed south, so's to meet me by and by"
-Townes Van Zandt