[Moims-dai] Todays Telecon: PDI on PDI

D or C Sawyer Sawyer at acm.org
Tue Jun 5 17:14:25 UTC 2018

Dear All,

This is my third attempt to try to make my points clear and I think I’m improving.  You will be the judge.

OAIS provides the view, clearly shown in Figure 2-3, of the Producer role submitting SIPs to the Archive.  In figure 2-4 the SIPs are used to create the Content Information, and as Mark notes, then the PDI for that Content Information can be obtained/created and associated.  In this sense, the PDI is supplementary to the Content Information and would not exist but for the Content Information.  The result is an AIP for storage in the Archive. In the above scenario, which is at the heart of OAIS, there is no implication that the PDI described above would be taken to be Content Information.  They are clearly distinct categories of information with likely different sources and with the PDI being subservient to the Content Information.  This is the top level view that sets the context for understanding the role played by the Producer and the relationships among the types of information objects.

David raises the issue that, well, the high level model view of figure 2-3 doesn’t say how the PDI is preserved and this is a concern.  But note that it doesn’t say how the Content Information is preserved either.  It simply says they are preserved as part (contained) of the AIP.  So at this high level, they are all understood to be preserved as part of a specific AIP.  I think this is a fair understanding that most people will agree with.

Expanding the model of an AIP, we get figure 4-10 which defines an Information Object.  This applies to the Content Information because it is defined to be an Information Object.  We see that it consists of a Data Object and its Representation Information. 

 Further expanding inside the AIP with figure 4-11 we see the Representation Information detailed as being composed of three types of Representation Information and with a recursion relation that ends at some point (not germane for this discussion). 

 Still inside the AIP, and further clarifying the nature of PDI and Representation Information, we have figure 4-12 that describes both of them as Information Objects, which means they also have Data Objects and Representation Information.   The result at this point is a set of data objects, all logically inside a single AIP, whose number is limited because the Representation Information recursion is limited. They are preserved by virtue of their logical inclusion inside the AIP.

Now the question is asked, since we thought that PDI was important in the preservation of the Content Information, which is the Content Data Object and its Representation Information, why isn’t it also important to associate secondary PDI with the original PDI?  Well the answer is, it might or might not be depending on the Archive and the specific original PDI.  If it is important, how should that be conceived and shown?  Original PDI was associated with the Content Information.  By analogy, secondary PDI would be associated with original PDI.  This would be shown as a recursive relationship, but all still within the same AIP.  We could include a diagram showing this recursion of PDI on PDI.  This recursion ends when the Archive decides that it no longer makes sense to continue it.  It may be different for the different components of PDI, which are shown in table 4-1 and listed in the text.  The PDI diagram could be broken out further with these components, but I’m not sure it is worth it.  Since they are classed as Information Objects, they could also be broken out in the Representation Information dimension but I see no need and this recursion is limited anyway.

The point that I hope is becoming clear is that there is no need to reuse Content Information with all its implications from the high level model of an Information Package and apply it to PDI.  This only promotes communication difficulties as the past two telcons and intervening messages have demonstrated.  In fact, we should always try to avoid reusing the same terminology in different contexts where the meaning is different to some degree. That is certainly the case when wanting to apply Content Information to PDI.  In short, there is no need.  

I recommend we put a simple diagram of PDI recursion into Section 4.2 and briefly talk about limiting the recursion by what the Archive finds is practical in each case.

As a corollary, there is no need to remove PDI from being applied to Representation Information as is currently being proposed.  David mentioned that he finds most archives are not applying PDI to Representation Information, but from a conceptual model perspective this is no justification as I, for one, have given several examples where it is relevant.  If one want to see PDI on PDI made more explicit, then at least the possibility of PDI on Representation Information needs to be retained.  There is absolutely no good reason to remove it from the conceptual model now that (I Hope!) were not trying to apply the AIP to PDI.  There are instead very relevant reasons to keep it.


More information about the MOIMS-DAI mailing list