{"id":1862,"date":"2024-11-07T09:00:00","date_gmt":"2024-11-07T09:00:00","guid":{"rendered":"https:\/\/exponentialdecay.co.uk\/blog\/?p=1862"},"modified":"2025-12-01T16:59:53","modified_gmt":"2025-12-01T16:59:53","slug":"dont-implement-premis-represent-it","status":"publish","type":"post","link":"https:\/\/exponentialdecay.co.uk\/blog\/dont-implement-premis-represent-it\/","title":{"rendered":"Don\u2019t implement PREMIS (re)present it"},"content":{"rendered":"

I got a response<\/a> to my paper PREMIS Events Through an Event-Source Lens<\/span>.<\/span><\/p>\n

There are two strange choices made by this response. I’ll touch on the more personal one at the end, but first, what does the response say?<\/span><\/p>\n

<\/p>\n

It\u2019s not entirely clear.\u00a0<\/span><\/p>\n

If the response says that, \u201cit is a choice to implement PREMIS?\u201d<\/em> And that \u201cPREMIS can be implemented in different ways?\u201d<\/em> \u201cand that it\u2019s technology agnostic\u201d<\/em> Then yes, 100% that\u2019s basically the driver for my original paper and once you read it holistically, instead of dissecting it and cherry-picking points, you will probably read it that way as well. <\/span><\/p>\n

As I wrote in my first blog response<\/a> to the publication of my paper in 2023,<\/span> Tessella\u2019s Rob Sharpe\u2019s 2013 presentation was an important reference point for me and we\u2019ll revisit it below, but Rob labors that PREMIS is technology agnostic and can be represented in other formats, and since 2013 I haven\u2019t seen enough conversation or discussion about that, and I wanted to amplify that message by looking at PREMIS in an event-sourced model as an aggregation.\u00a0<\/span><\/p>\n

If there\u2019s something more substantive in the PREMIS Editorial Committee’s (EC) response, then I feel it\u2019s lost in its own stylistic choices (to focus on what I <\/span>might<\/span><\/i> have been saying rather than taking a show don\u2019t tell approach to clarifying their more salient points.). <\/span><\/p>\n

I wonder if it might have been handled differently? I am pretty easy to find these days, and so reaching out to clarify any of my thinking might have been one way; perhaps there was a way to collaborate on a response; perhaps most of of the EC\u2019s concerns (if there are any) could have been handled with a joint editorial note in the original paper to clarify that my words are not<\/em> an authoritative source on PREMIS, rather, PREMIS (events) were largely a vehicle to describe more the benefits of an event-sourced architecture and that you still need to consider and interpret<\/em> the PREMIS documentation and guidance for yourself before implementing it in your own solutions.<\/span><\/p>\n

Going a different direction<\/span><\/h1>\n

The essence of the original paper is this: (from my perspective) PREMIS is not a schema to be implemented in the back-end of any digital preservation system. Should it be still be deemed a relevant technology, it might be studied in your requirements analysis, and you would make sure that your own system is not lossless in any way as to effect PREMIS \u201cconformance\u201d, but you would not match your \u201cschema\u201d to PREMIS, you would ensure that you can output it, \u201cpresent it\u201d that is, it would become one representation of data that can be generated from your system out of many. One view, or as I clearly point out, an aggregation, in the case we have chosen an event-based architecture.\u00a0<\/span><\/p>\n

This is not at odds with the (so-called) corrections that have been provided to me in the Code4Lib journal article from the PREMIS EC.<\/span><\/p>\n

That being said, a further thesis is that PREMIS events are often a lossy, stateful representation of data in a digital preservation system. PREMIS represents one-dimensional state (or slices of state) over a period of time. In the modern engineering world, we have at our disposal methods of capturing, greedily, all events in the life of a digital object and doing that will create a richer view of the life of that object, and, as a representation of that data, a richer PREMIS view of an object and its events over time if so desired<\/span>.<\/span><\/p>\n

The authors of the EC response labor heavily on their perception of a misunderstanding on my part about PREMIS and they can choose to do that but what may look like a misunderstanding of PREMIS is not a misunderstanding of technology:<\/span><\/p>\n

Conformance, in general, is defined as:<\/span><\/p>\n

> how well something, such as a product, service or a system, meets a specified standard<\/span><\/p>\n

And the PREMIS EC have decided to attach levels to conformance (also graduated levels, and degrees) to \u201cquantify(ing) the degree to which PREMIS has been implemented\u201d, three of which are anchored in implementation, apparently, three distinct implementations.<\/span><\/p>\n

    \n
  1. Mapping, indirect or otherwise,<\/span><\/li>\n
  2. Export,<\/span><\/li>\n
  3. Direct implementation,<\/span><\/li>\n<\/ol>\n

    I write:<\/span><\/p>\n

    PREMIS conformance should be separate from representation. If we acknowledge PREMIS is at least one important representation of preservation metadata, i.e. for its ability to act as an interface to those looking to interpret preservation metadata, then whether it exists logically on disk, or is generated through an event sourced projection, is irrelevant. How a representation complies with the PREMIS data model remains of greater importance, but this is measured from the same eventual view, whatever intermediate abstraction it sits within.<\/span><\/p><\/blockquote>\n

    The PREMIS EC can choose to have three graduated levels of implementation to quantify degree of implementation. They can also make it clear level three (internal representation) is not necessarily the final goal, but it might benefit you; but If you\u2019re not the PREMIS EC, don\u2019t go near it, there\u2019s no need.\u00a0<\/span><\/p>\n

    I posit that conformance is only how well you can map to PREMIS or access something PREMIS-like that satisfies its data model<\/span>. Your goal is to look at PREMIS as one interface you can potentially satisfy (you still need to describe objects uniquely; you need to describe agents engaging with them; rights need to sit somewhere)<\/em>, and once you can satisfy that interface you can access it in many different ways, and conformance should be measured against that, <\/span>if<\/span><\/i> PREMIS conformance is deemed valuable. <\/span><\/p>\n

    Put simply, conformance does not require levels. Levels may simply be the wrong word, these are just guides you might follow to demonstrate conformance (or ways that someone might audit a system to determine conformance).<\/span><\/p>\n

    The EC clipped this from one of the points they responded to:<\/span><\/p>\n

    Is level three (internal implementation) reasonable in today\u2019s software development world, is it reasonable in today\u2019s environmental climate?<\/span><\/p><\/blockquote>\n

    Do we sacrifice the potential to store and access other different, richer, more-complex, (or less-complex), representations about other cross-sections of our data at the expense of putting PREMIS at the core of our digital preservation system? \u2013 No. We can make it an output of many, and use its schema and data dictionary to output it, but we don\u2019t build around it, we essentially report around it.\u00a0<\/span><\/p>\n

    They argue:\u00a0<\/span><\/p>\n

    there are also benefits in choosing to take an internationally defined and agreed data model and use that as the basis of your system.\u00a0<\/span><\/p><\/blockquote>\n

    Well, if it\u2019s internationally defined and agreed, let\u2019s just do that! \ud83e\udd37<\/span><\/p>\n

    The benefits of not implementing an external data model are broadly around increased control and flexibility, however the trade-off to consider is the likely loss of easy interoperability and exchange with other systems.<\/span><\/p><\/blockquote>\n

    If you re-frame PREMIS as an interchange-format and you can prove that as useful, you absolutely have my buy-in and I will have designed you a system that doesn\u2019t preclude a PREMIS-like output, i.e. a way of aggregating more detailed information in your system and outputting PREMIS as a representation (a format) for others to understand.<\/span><\/p>\n

    The resurgence of OAIS?<\/span><\/h1>\n

    From the EC:\u00a0<\/span><\/p>\n

    There are two responses to this, the first is to note that access has always been considered a part of Digital Preservation, to the point that one of the functional areas of the OAIS model is Access.<\/span><\/p><\/blockquote>\n

    Who had OAIS on their World Digital Preservation Day (WDPD) Bingo Card?\u00a0<\/span><\/p>\n

    But also, no. This is a misleading read and deserves more context.<\/span><\/p>\n

    Access when it is considered part of digital preservation is when access is used as a measure of success of digital preservation (or indicator of the potential obsolescence of an object) – it is an intrinsic property of digital preservation.\u00a0<\/span><\/p>\n

    But the access function in OAIS is not that<\/a><\/span>. And even if you\u2019re crafty, and build an access component to a system that provides a feedback loop to digital preservation functions, it\u2019s not that part of OAIS.<\/span><\/p>\n

    \"Figure<\/a><\/p>\n

    Now, PREMIS does have some nice features that support access BUT we\u2019re talking \u201cevents\u201d, and information that supports digital preservation and even though there may be a way to encode events that provide a feedback loop to measure the success of preservation, e.g. {\u201cevent\u201d: \u201caccess\u201d, \u201cdetail\u201d: \u201ctried to open PSD in GIMP\u201d, \u201coutcome\u201d: \u201cFAIL\u201d}, true access goes well beyond the scope of my article and the spirit in which it was written.<\/span><\/p>\n

    We need to evolve<\/span><\/h1>\n

    The EC presents a somewhat dogmatic and institutionalised response. As a flaneur in the field, as someone who has worked implementing PREMIS in one of the most PREMIS heavy digital preservation systems out there, and involved too in efforts to minimise PREMIS verbosity, including my own event-like approaches I revisit Sharpe\u2019s paper in 2022\/2023. I do this asking, why don\u2019t we talk about it more? Why do I see projects today still see XML as the end goal of PREMIS?<\/span><\/p>\n