[ODFPlugtest] MSO does not read LO files

[Andreas Guelzow]

On Tue, 2011-06-28 at 12:39 -0700, Dennis E. Hamilton wrote:
> The question is, what is the easiest way to repair the damage.  If it
> is by raising an issue in the current Candidate OASIS Standard Public
> Review, I will reluctantly do that.  If we can instead offer guidance
> for implementers and reconcile the specification later, I would be
> happier to do that.

As an implementor, if I am faced with a choice between accepting
"guidance" and violating the specification, I will have to follow the
specification.  As a member of the OASIS ODF TC I am absolutely opposed
to issue "guidance" that violates the specification.

> 
> The standard is too brittle in this place, with no added value and the
> downside of making documents that are so "valid" rejected by
> down-level consumers that will likely process the document just fine
> once it is "corrected."

Which consumers? I am only aware of one implementor whose product (any
version) is so brittle to stumble over this change.

> 
> I do not believe this was a well-considered change (even though it is
> there because I opened my big mouth).
> 
> Originally, manifest:version was defined only for use on a
> <manifest:file-entry> to indicate the version of the artifact that the
> file entry corresponds to. 

Your initial arguments that the manifest should have version information
still applies. Before we dropping the requirement of this version we
should really consider whether we can always live without version info,
or what exactly the absence of the version attribute means. 

> 
> I know I commented somewhere that when I first saw manifest:version in
> a draft of Part 3, it naturally occurred to me that it was about the
> version of the manifest and so foolishly went to the next step: why
> not?
> 
> Showing how confusing my confusion can be, we had these two JIRA
> issues:
> 
> OFFICE-2243:
> <http://tools.oasis-open.org/issues/browse/OFFICE-2243>
>   manifest:version ill-specified and misleading

Does MSOffice stumble over this attribute or only over the attribute of
the name name to the manifest:manifest element?
> 

> OFFICE-2244:
> <http://tools.oasis-open.org/issues/browse/OFFICE-2244>
>  Provide Package-Level manifest:package-version attribute
>  *THIS IS ENTIRELY MY ERROR* 
>  ****How many times must I say this was a blunder?****

Why does the reasoning given not any longer apply. Are we sure that this
was indeed a blunder? [To me it still looks like that the choice made by
an implementor not to allow for future ODF versions is suddenly seen as
a problem for the specification.]

> 
> Here we made things even more brittle.  Nice idea in a self-evident
> sort of way. Terrible idea considering that it breaks down-level
> compatibility whether or not the document in hand will be processable
> just fine by an ODF 1.1 consumer were the attribute simply not used.
> This is not about what an ODF 1.2 consumer should do, but about the
> havoc introduced for no gain with respect to some ODF 1.1 consumers
> who take the ODF 1.1 manifest.xml schema as gospel.

Are we really sure that there cannot be situations where this version
attribute is important?

Andreas

> 
>  I doubt that this was ever discussed on the TC or noticed by a
> reviewer after it was introduced and first appeared in this form in a
> Public Review.
> 
>  The description in the issue states my concern.  I did not consider the down-level impact.
> 
>  (I don't know if anything happened about Michael's comment concerning the <dsig:document>.  I'm afraid to look.
> 

PS: I believe this discussion should really happen on the TC mailing
list.


-- 
Andreas Guelzow <aguelzow at pyrshep.ca>