OASIS Open Document Format for Office Applications (OpenDocument) TC

Hi,

I’ve been thinking about the list-id/list-override proposal the whole weekend. Since we’re going to vote today and list-id/list-override I just want to summarize my main objectives:

1. Backward compatibility.
The list-id changes the actual number of a paragraph. The list-id makes is impossible for old readers compute the right number.

2. Roundtrip fidelity.
The algorithm described by Oliver to convert between list-id and numbered-paragraph contains non-deterministic choices. So the conversion between list-id and numbered-paragraph will be non-deterministic which results in a very bad user experience. (My opinion :-))

3. List-id and list-override solve the same use cases.
My understanding is that the list-id and list-override approached solve exactly the same problem and that list-override is better suited to address backward compatibility.

For reasons of 1-3 I can not support the current proposal which mixes list-ids and list-overrides.

To be a good TC member I updated the pure list-override proposal according to some concerns from Oliver and would like to offer this for vote too.

And Thomas: My objection to the list-id has nothing to do with personal issues; like not being able to accept a good compromise. Guaranteed. It’s just that I’ve been persuaded by ODF users on conferences etc. that stability of the file format is the most important thing. This is the only reason I object to the list-id approach. But I value your believe as a developer that a list-id is more cool than a list-override on a technical level but --- in my opinion --- it harms the stability of ODF.

Thanks for the patience and the professionalism in discussion and thanks for considering my concerns. 

~Florian

On Monday 12 March 2007 14:13, Florian Reuter wrote:
> I’ve been thinking about the list-id/list-override proposal the whole
> weekend. Since we’re going to vote today and list-id/list-override I just
> want to summarize my main objectives:
>
> 1. Backward compatibility.
> The list-id changes the actual number of a paragraph. 

Thats false. No idea why you would think that.

> The list-id makes is 
> impossible for old readers compute the right number.

Thats false. Even without the list-id old readers can not compute the right 
number. Its not specified how to do that.

> 2. Roundtrip fidelity.
> The algorithm described by Oliver to convert between list-id and
> numbered-paragraph contains non-deterministic choices. So the conversion
> between list-id and numbered-paragraph will be non-deterministic which
> results in a very bad user experience. (My opinion :-))

We since clarified the ideas a lot, and the algorithm may be adjusted.  But, 
please be more specific when you give your opinion.  I'd like to know if is 
an informed opinion.

> 3. List-id and list-override solve the same use cases.

I think you still misunderstand if you think this.

> For reasons of 1-3 I can not support the current proposal which mixes
> list-ids and list-overrides.

I can understand that, unfortunately you have misunderstood the issues and 
failed to open any dialogue over corrections we pointed out to your 
interpretation over the last week. I find it highly unfortunate that we 
arrive at a point where its a "yes"/"no" match because there is no open 
dialogue over the actual issues at hand.

> I’ve been persuaded by ODF users on conferences etc. that
> stability of the file format is the most important thing. This is the only
> reason I object to the list-id approach. 

Then please answer the email from Michael this morning which tells you how 
your idea that its not backwards compatible is incorrect.

I, and others, have opened a dialogue with you over this issue several times, 
and the only thing I see next is another email with the same issues being 
repeated that have been corrected in the reply.
Can you please continue a dialogue by replying to the issues addressed in the 
various mails on this subject?

So far it's more like we're entered in a 4x100m relay, and you're doing the 
backstroke.  That's not working.
-- 
Thomas Zander

On Monday 12 March 2007, Florian Reuter wrote:
> 1. Backward compatibility.
I don't think backward compatibility matters -that- much for numbered-paragraph that we can't change any behavior.
Numbered-paragraph was added at koffice's request, and was never implemented in OOo, and it turns out
that it was under-specified. So I don't see the point in trying to clarify a vague specification in a way that preserves
a theoretical backwards compatibility with implementations that can't possibly exist.
We should rather clarify it in a way that actually makes sense and is useable by implementations.

And if that means adding list-id, then let's add list-id, and let's not care about the non-existant problem
of backward compatibility. Implementations that actually use numbered-paragraph already (like koffice)
can take care of it at the application level (using the old algorithms if no list-id is used, and respecting
list-id if set).

-- 
David Faure, faure@kde.org, sponsored by Trolltech to work on KDE,
Konqueror (http://www.konqueror.org), and KOffice (http://www.koffice.org).

David Faure wrote:
> On Monday 12 March 2007, Florian Reuter wrote:
>> 1. Backward compatibility.
> I don't think backward compatibility matters -that- much for numbered-paragraph that we can't change any behavior.
> Numbered-paragraph was added at koffice's request, and was never implemented in OOo, and it turns out
> that it was under-specified. So I don't see the point in trying to clarify a vague specification in a way that preserves
> a theoretical backwards compatibility with implementations that can't possibly exist.
> We should rather clarify it in a way that actually makes sense and is useable by implementations.
> 
> And if that means adding list-id, then let's add list-id, and let's not care about the non-existant problem
> of backward compatibility. Implementations that actually use numbered-paragraph already (like koffice)
> can take care of it at the application level (using the old algorithms if no list-id is used, and respecting
> list-id if set).
> 

I agree to this statement.


Regards, Oliver.