[LINK] XML is evil

[Richard Chirgwin]

Craig Sanders wrote:
> On Tue, Nov 18, 2008 at 04:09:54PM +1100, Richard Chirgwin wrote:
>
>   
>> The problem arises when someone who *doesn't* want to dabble in data
>> formats, like me, finds himself confronted with 80 MB of XML document,
>> and no particular help from the origin of the document to make it
>> accessible to any tool other than where it was created. So it is, I
>> agree with someone earlier, the originators of XML datasets who act like
>> ninnies, by assuming that some "other" tool somewhere out there will
>> take care of users and we can treat the dataset like an infinite sandbox.
>>     
>
> they're not "acting like ninnies". they're using XML as a data transfer
> tool, exactly as it is intended to be.
>
> that assumption (i.e. that there will be some other tool) is the right
> assumption to make. the ability to make and rely on that assumption is
> part of the reason for XML.
>   
Where ninniedom comes to the fore is where someone says "here's the tool
to parse our XML / send it to MySQL / whatever", while someone in
another part of the development team says "let's add a new tag", and the
parser doesn't work; which sends the user back to the raw XML file.
Which, for the particular file in question (the Open Street Map planet
file for Australia), means "you can use this in our application, you can
learn to parse the file by hand, or you can forget it".

When the file is big, opening it in (say) Firefox to get the "no style
information" complaint but the document tree isn't particularly useful.
Making sure I have all the permutations, even just for an OSM segment
for Botany, by reading the document tree? Nope...

So it comes down to "why is this silly"? and, IMO, it's where XML is the
wrong tool for the job.

RC
> The XML file itself contains enough meta-data (field-names, attributes,
> etc) to parse it correctly and extract data even without the DTD (which
> is, BTW, referred to in the XML file), and the associated DTD provides
> further information for more sophisticated tools like editors and
> validators..
>
> craig
>
>