[ODFPlugtest] MSO does not read LO files

[Andreas Guelzow]

On Mon, 2011-06-27 at 20:42 -0400, robert_weir at us.ibm.com wrote:
> > From: Hanssens Bart <Bart.Hanssens at fedict.be>

> > Fixing the error is exactly what a company / project should do, 
> otherwise
> > we're back at mimicking bugs like in the pre-bubble HTML -days ...
> > 
> 
> This is different than "quirks mode" in browsers.  That was the problem 
> when a CSS property would work one way in one browser and another way in 
> another browser.  So it was impossible to make a web page that would work 
> everywhere because of varying behavior associated with the use of a 
> property.
> 
> The manifest version is not really an example of that.  The ODF 1.2 draft 
> doesn't specify any behavior related to this attribute.
> 
> I appreciate that the property is mandatory according to the schema.  So 
> live has give us two sucky alternatives:
> 
> 1) Write out the attribute and have Office 2010 scare the bejeebers out of 
> users with a statement that their file is corrupt.
> 
> or
> 
> 2) Leave the attribute out and face the wrath of Alex Brown's blog when he 
> validates the documents and finds they are missing an attribute.  But the 
> file loads in Office 2010 without error.
> 
> 
> > SO IMHO what MS-Office perhaps should do is give a little warning about
> > it reading an ODF 1.2 while only supporting 1.1 instead of the current
> > error message. And OOo should also set this attribute correctly
> > 
> 
> That would be an ideal solution.  But until that change is made in Office 
> 2010, I don't think that having users be told that their ODF document is 
> corrupt is particularly good marketing for LO or ODF.  No user is going to 
> blame Office.  They are going to blame the person who sent them the 
> document.
> 

Of course there is a third solution, that we have control over: do not
introduce any new attributes that are _required_. Having this required
attribute forces implementations to choose between writing invalid files
(unacceptable in my mind) or create files that some programs are
unwilling to read.

Of course any program should be able to recognize that the file in
question is from a newer version of ODF. So stating that the file is
corrupt and needs to be "repaired" is just the wrong thing to do.
Of course OOo also has the habit to state that it is "repairing" files
when in fact it doesn't do any such thing. (We experienced that with
Gnumeric generated files when the underlying library we used set an
incorrect flag in the zip wrapper and OOo talked about repairing the
file when it just ignored that flag.)

I am wondering what has happened to the adage to be as generous as
possible when reading a file (but as particular as possible when writing
it.) 

Andreas

-- 
Andreas Guelzow <aguelzow at pyrshep.ca>