Re: [office-comment] Section 1.5 - implementations SHOULD use thenamespace prefixes given.

["David A. Wheeler" <dwheeler@dwheeler.com>]

>>>do you know an application that actually could be able to support the 
>>>OASIS Open Office format, but does not support namespaces?
>>
>>Yes - KOffice :-)
>>
>>We currently look for "office:styles" etc., not for "the element named prefix:styles
>>where prefix was associated with urn:oasis:names:tc:openoffice:xmlns:office:1.0",
>>as proper namespace support would mandate.
...
> I was just proven wrong: Qt's DOM implementation supports namespaces just fine.
> 
> It's the KOffice code which was written without this in mind, and with hardcoded
> prefixes instead. I guess we will convert it to "do the right thing".

You're proving my point: it's quite easy/common to implement a reader
that doesn't use XML namespaces. We already have one!

But more importantly: I suspect that most implementations
of readers will NOT be full-up office suites, just like most
HTML processors are not full-up web browsers.
Instead, most readers will probably be small specialized tools
built to extract certain data in exotic ways.
It's much easier to implement small tools if you give them
fewer complications.  For most, handling namespaces, while
desirable, isn't critical.

By all means, use namespaces, and encourage their use.
But if you can encourage "standard" prefixes, that should be used
unless there's some reason not to, it's more likely
that one-off tools will work.  One-off tools are actually
quite common, so let's remove a common source of errors for them.

And besides... humans have to read this too.  If you insert
random character names for the XML namespaces, a full implementation
of namespaces is happy... but the humans having to debug bad
data files sure aren't.  Please, create a common convention; that
will make it easier for the poor humans who have to pick up
the pieces.

--- David A. Wheeler