[
Lists Home |
Date Index |
Thread Index
]
Elliotte Harold wrote:
> Sean McGrath wrote:
>
>
>> 1) The lack of sane, simple roundtrippability. I read in some XML, I
>> write it straight back out again. I loose stuff on the way. R u nuts?
>> And you call this a machine processible data format:-)
>
>
> In my experience if you really care about anything you lose (CDATA
> sections, entity references vs. characters, etc.), that's a serious
> code smell that indicates a major flaw in the stuff. Yeah, the stuff
> that comes out may not be as nicely formatted for humans reading it
> with more or a text editor, but often even that can be fixed with
> appropriate options on the serializer.
>
> I'm not sure exactly what this RIG thing you cited is, or what your
> use case is, but after skimming the spec, my nose is going nuts. This
> smells very bad. Why would anybody ever care about preserving
> attribute order? Why forbid CDATA sections? And section 12.1, "XML 1.0
> features that, if present in an instance, will cause RCF1 to RCF
> conversion to fail" includes "Presence of control characters in the
> U+0000 to U+001F range other than TAB (U+0009), CR (U+000D) and LF
> (U+000A)." These are not XML 1.0 features. If these things show up the
> document is malformed, irrespective of all the extra rules you're
> piling on top of raw XML. There's something deeply wrong here. :-(
The use case here is transmitting XML-based messages from one service to
another service on a Service Oriented Architecture and doing it in such
a way that (a) it is possible to be sure that a message routed "straight
through" has not been tampered with and yet (b) the XML is fully
visibile - not a lump of attatchment goo - for the purposes of
intelligent routing.
Equally important is the fact that an intermediating service can
add/modify/delete content from the XML instance without doing damage to
the untouched parts of the instance.
Sean
--
http://seanmcgrath.blogspot.com
|
