Provenance DM WD review - invalidatedAtTime description

[Ole Streicher]

Hi Mathieu,

to bring the specific problem here with "invalidated", here are two
examples:

1. Suppose you have a calibration activity that takes a calibration file
to produce some result. That result was then taken out and someone used
it in their publication.

At some point, the calibration file was superceded by the next version
(for the *same* input data, so with the same observation time),
invalidating the first moment from that moment on.

Some time later, someone comes and wants to reproduce the original
result with its original data, and to store that in the Provenance database.

That would require that he uses the *first* calibration file (which is
already invalid) in a *new* activity. Although he could do it (as long
as the first calibration file is available), he cannot record its
Provenance, since it contradicts to the constraining that an Entity
cannot be used after invalidation.

Reproducing a result (even with outdated calibration files) is however a
perfect use case.

2. Suppose someone wants to do some statistics on the changes made
between different versions of a calibration file (where some may be
invalidated). This is possible as long as the files are somehow
available. However, describing the provenance for the activity that
calculates the statistics is impossible, because of the same constraint.

Both examples are not limited to calibration entities.

That shows, that there is a difference between the validity period of a
file and the lifetime of an Entity:

 * A dataset must/should not be used for regular processing outside of
   its *validity time*.

 * an Entity cannot be used (for *any* purpose, even for display on the
   user's side) outside of its lifetime.

This difference has to made clear in the description of the attribute.
That was the intention of the addition, and (again) this came out as one
of our agreements in May.

Best regards

Ole

On 10.07.19 11:49, Mathieu Servillat wrote:
> Hi,
> 
> To be precise and base the discussion on facts rather than opinions, the
> previous corresponding attribute (PR 2018-10-15) was indeed:
> destructionTime - prov:invalidatedAtTime - datetime - date and time at
> which the entity was erased or invalidated
> 
> We then received comments during an RFC phase that questioned why some
> attribute names were different from the W3C PROV names. It is not my
> decision to change this attribute, but a general agreement between the
> authors, as is the decision to follow strictly the W3C PROV definitions
> when the concepts are exactly the same so as to avoid any confusion
> (i.e. all the core model, Activity, Entity, Agent, Used, WasGeneratedBy,
> WasAssociatedWith, WasAttributedTo). Further interpretation of this
> attribute in the context of an implementation can be discussed elsewhere.
> 
> Cheers,
> Mathieu
> 
> 
> 
> Le mer. 10 juil. 2019 à 11:29, Ole Streicher <ole at aip.de
> <mailto:ole at aip.de>> a écrit :
> 
>     Hi Mathieu,
> 
>     this attribute was originally called "destructionTime" in our model,
>     which perfectly describes its intention (see the previous PR).
> 
>     The name was changed by you to the W3C attribute name "invalidatedTime"
>     during the adaptation of the compromise model, specifically in commit
>     #5479 on Volute.
> 
>     On my responses, and in the May Interop discussion, I responded that the
>     name change may be misleading, since it does not completely reflect the
>     semantics of that attribute. The compromise here was to add a sentence,
>     that has nothing to do with the validity of the Entities.
> 
>     Since we are not really W3C compatible anymore (and need conversion name
>     tables elsewhere as well), maybe the clearer solution to go back to the
>     old attribute name here.
> 
>     Best regards
> 
>     Ole
> 
>     On 10.07.19 10:27, Mathieu Servillat wrote:
>     > Dear DMers,
>     >
>     > As it seems that the authors can also express comments on their text
>     > during the WD review, I have one concern with a sentence that was
>     > somehow added without my consent, and that I find wrong and misleading
>     > about the relation of invalidatedAtTime to validity. Here is the
>     text in
>     > the WD:
>     >
>     > Table 1 - invalidatedAtTime: date and time of the destruction,
>     > cessation, or expiry of the entity. The
>     > entity is no longer available for use (or further invalidation) after
>     > invalidation. This does not relate to
>     > the validity of the entity.
>     >
>     > The last sentence "This does not relate to the validity of the entity"
>     > brings confusion, as after invalidation, an entity is of course not
>     > valid anymore in some way (more specifically "no longer available for
>     > use", and probably most of the time destroyed). We could further
>     discuss
>     > elsewhere the relative validity of an entity for a given usage, but it
>     > is out of the scope of the DM WD, that simply provides an
>     irreversible,
>     > absolute validity check.
>     >
>     > This attribute is directly taken from the W3C PROV recommendation, and
>     > so is the description text, except for the addition of this sentence.
>     > Therefore, it would be best to stick to the W3C description to avoid
>     > confusion.
>     >
>     > Cheers,
>     > Mathieu
>     >
>     > --
>     > Dr. Mathieu Servillat
>     > Laboratoire Univers et Théories, Bât 18, Bur. 222
>     > Observatoire de Paris, Site de Meudon
>     > 5 place Jules Janssen
>     > 92195 Meudon, France
>     > Tél. +33 1 45 07 78 62
>     > --
>     >
> 
> 
> 
> -- 
> Dr. Mathieu Servillat
> Laboratoire Univers et Théories, Bât 18, Bur. 222
> Observatoire de Paris, Site de Meudon
> 5 place Jules Janssen
> 92195 Meudon, France
> Tél. +33 1 45 07 78 62
> --
>