RE: [xml-dev] Nested Documents (was: XML 2.0)

["Michael Kay" <mike@saxonica.com>]

Thanks, but I still don't understand what you mean by "more tightly linked". You're just repeating the assertion that a final end tag is good for me because it enables me to detect more errors - which I have already explained is the wrong tradeoff in my case. In fact, if my log file really is broken for some reason, then I would rather be able to read most of it than none of it.

 

Michael Kay

 

---

From: Richard Salz [mailto:rsalz@us.ibm.com]
Sent: 27 February 2008 18:07
To: COUTHURES Alain
Cc: Michael Kay; xml-dev@lists.xml.org
Subject: Re: [xml-dev] Nested Documents (was: XML 2.0)

> I don't understand what you mean by "more tightly linked"...

In a single-rooted document, <a>...</a>, when I see that final closing bracket I can be pretty sure that I've got everything.  (Yes, trailing comments and PI's I handwave aside.)  If the input stream ends before that closing bracket, then I know something went wrong.  My XML consuming application can pretty much treat the input as a stream of bytes and not care where how or why something went wrong; at least at the first level.  See the bracket, things ok; don't see the bracket, something broke.

Now, let's imagine a multi-rooted document like <a>...</a><b>...</b>.  How do I know when I hit the end? Suppose the network drops after the </a>?  If you're using something like TCP socket API, you have no idea how many bytes were in-flight when the connection dropped.  (TCP is realiable only when its working; it's failures modes are pretty pitiful.)  Suppose I'm reading a log file and there's no free space on the disk but I do see the closing </b>.  Is that coincidence -- did the producer just get lucky and take the last bits of the disk, or am I missing <c>...</c> and friends?

There's no way to know, in a generic way, without doing very specific things depending on when where why and how the XML is transported.  Right now, you don't have to.  Yes, that's a simplification, 80/20, whatever you want to call it.  I'd call it a trade-off and say the current design is the right one. :)

It has issues in the real world, too.  Suppose you're writing a book and someone inserts a pagebreak between <a> and <b> in a multi-root document.  Are they gonna get confused or misled if they're not explicitly told to turn the page?

        /r$

--
STSM, DataPower Chief Programmer
WebSphere DataPower SOA Appliances
http://www.ibm.com/software/integration/datapower/