1 of 62

Resistance is Fertile

Building a Manualfesto for Lib Tech

Christina Harlow @cm_harlow code4lib 2017�bit.ly/c4l17harlow

2 of 62

bit.ly/c4l17harlow

3 of 62

Hi, I’m Christina.

I make data work.

Or try to.

My name is Christina, and I work with metadata in some way that I like to call metadata operations. I am lucky enough to do this work at a well-funded, large academic library, and to be now considered a professional worker, if still considered entry-level.

I often feel like I’m making up what it is I do as I go along, and I’ll be the first to tell you I’m not a technological guru, developer, or programmer. I’m just a girl who is handy with the metaphorical data duct tape.

Most of what I’ll say today is coming from my work context - namely, I work with metadata; I work openly; I develop iteratively; I experiment without fear; I try to focus on what we can do now instead of what we should be able to do in 20 years; and I spent a lot of time before being visible in communities such as code4lib as a library support staff person (and this last bit of context has left a big chip on my shoulder). All that said, I’d like to share some points on my work ethic and approach that is both meaningful and honest for me right now, and I hope it lands for you as well, in your own context.

----

Code snippet is taken from one of my very first library data Python projects, the Geonames OpenRefine Reconciliation API: https://github.com/cmh2166/geonames-reconcile This is not included because it is a particularly great example of code - actually, it isn’t. But it is a great representation of how and where I started in this library technology space - hacking together something that would fill a need too low for application development, too high for continued manual work via Excel.

4 of 62

Hey, library, let’s talk about diversity of views & opinions.

5 of 62

“There is no evidence that time heals all wounds, or even most wounds; instead, it freezes unnecessary enmity and makes it harder to overcome. Time allows perpetrators to forget the pain they have imposed… Therefore, this is no argument for silence.”

Sarah Schulman, Conflict is Not Abuse (2015)

Photo by Dorret / Creative Commons License

Maybe not. Perhaps our inward-looking selections this year signaled some correlation to a similar impulse I’ve had with regards to my work recently. Namely, we are in a highly polarized and dangerous political climate. White supremacy groups are more active - or perhaps, simply more visibly active in the public eye; bias incidents are up in our communities; black, brown, and trans people are being murdered; Muslims are stopped from returning to their homes; journalism and truth is being blocked.

This is not a simple state of affairs caused by 1 day’s changes, but issues that have grown over time, and whatever your political orientation, we can probably agree that we are in an intense period of cultural and political change.

Our communities are shaken, and I don’t know about you, but I’m feeling more acutely the need than ever to situate myself in some meaningful context for my professional practice. What about my daily work do I need to change because it in part reflected the values that got us here, even if not intentionally? This is not a time for status quo, this is not a time for our silence on these questions.

----

Background photo was taken on January 15, 2015 by Dorret. It shows marchers in a Black Lives Matter march in NYC.

The quote comes from Conflict Is Not Abuse: Overstating Harm, Community Responsibility, and the Duty of Repair (2016) by Sarah Schulman. In particular, Part Two: “The Impulse to Escalate”, Chapter 5, “On Escalation”. http://www.arsenalpulp.com/bookinfo.php?index=449

6 of 62

“As long as we’re not willing to put action behind our words, this profession isn’t a safe space. I can’t imagine a future for libraries that can talk the talk, but don’t walk the walk.”

Cecily Walker, LITA Forum 2016 Keynote

Photo by Gregory Varnum / Creative Commons License

In responding to this question, I earnestly think we don’t know what we believe until we see what we do. My position paper is the open documentation of my work.

But first, mad kudos and all of my respect to those theorists, planners, builders, thinkers, thought leaders, implementers that have pushed these discussions forward and thus have made this space available now to someone like myself who wants to Get. Shit. Done. Thoughtfully.

So, how do we make our ethics and politics apparent in our daily actions? Especially if we don’t work in some explicitly political or ethical space? How do we “walk the walk”?

----

The quote is from the LITA Forum 2016 Keynote, Dismantling Whiteness and Oppression in LIS through Community-Led Collaboration by Cecily Walker. http://wiki.litaforum.org/index.php?title=Presentations_and_Uploads#Keynotes

The background image is from a January 2017 DTW emergency protest against the Muslim Travel ban: https://commons.wikimedia.org/wiki/File:January_2017_DTW_emergency_protest_against_Muslim_ban_-_04.jpg The photo is by Gregory Varnum: https://commons.wikimedia.org/wiki/User:Varnent

7 of 62

Linked Data is not the solution.

Killing MARC is not the solution.

Ruby/Scala/… is not the solution.

JSON/XML/… is not the solution.

ETL or Agile is not the solution.

Tools are not solutions.

Tech, alone, is not “the solution”.

I started to answer this question by looking at the other side of the coin - what is my daily work now, and what technologies are being touted as solutions or game changers in other senses? Data work is my bailiwick, and that’s definitely shifting.

What our work looks like, outside of any explicit political or cultural concerns, is changing quickly - well, quickly for libraries, which means not very quickly at all, but it doesn’t mean we’re not extremely anxious about it. Sometimes we confuse the need for preservation and persistence of the information we manage with the need for preservation and persistence of our workflows and skillsets. Regardless, our shifting landscape is not, in my opinion, because of any technology - this is not a “LINKED DATA WILL REVOLUTIONIZE ALL THE ISH” talk.

This is not to say that the ideas on the slide do not have merit - obviously they do. But they alone are not a solution - not to the issue of getting our ideals more closely aligned with our outputs, nor a solution to how we shift our interactions with our work.

Instead, I’d present that our data technologies and our seeming incompetence on the whole to implement them in any wholesale, meaningful, or ecosystem-aware fashion, highlights shifts happening not in how we define data, but in how define professional relationships, our skillsets, our power hierarchies, our interactions with ourselves and with systems.

8 of 62

For me, data and metadata work in this space has served as that point of intersection between what abstractions we apply (and the biases inherent in that), what technologies we build (and the biases inherent in that), and how we represent resources we’ve decided important enough to capture (and the biases inherent in that).

How do we prioritize actions based on theories and politics? How do we say:

YES, THIS DATA IS AN ABSTRACTION THAT REPRESENTS A WORLDVIEW I CAN CRITIQUE BUT NOT FULLY SEE.

YES WE WILL PUBLISH THIS FLAWED DATA ANYWAY BECAUSE WE CAN’T FINISH THIS CRITIQUE OF ABSTRACTIONS & BIAS IN OUR OWN, CLOSED WORK SPACE.

WE CANNOT WAIT FOR PERFECT BEFORE RELEASING.

BECAUSE WE CAN’T ALONE DECIDE PERFECT WITH REGARDS TO WHAT THIS DATA REPRESENTS, OR HOW IT IS MANAGED.

I think these questions relate directly my experience of code4lib as well. At a very fundamental level, what is coding about? I’d posit, it is a way to convey meaning built off the presumptions captured in some sort of abstraction. We are about ways to communicate those abstractions, and work such that this communication can be more efficient and so that machines can process our ideas. Code4Lib, for me and like metadata work, exists in that liminal data space where ourselves and our viewpoints are actively translated into abstractions that machines can process.

The “making it work” spirit is another important aspect of code4lib for me. We want to get stuff done, and iterate, and experiment, and tinker, and build, and maintain - not just theorize, plan, or idealize.

And code4lib wants to support that, all of the above, in a way that anyone can take it and work with it. Run wild, experiment, fork, play, build.

9 of 62

Photo by Feral78 / Creative Commons License

This leads me to my work approach. Which is my life approach. Which indicates more about the time and place I grew up in than anything else, probably.

Riot grrrl and queercore are huge influences to my work approach to library data technology. I see every action I do, from spreadsheets to Git repositories to staff meetings to workshops to documentation, as political - not in terms of Republican or Democrat, but as a chance to be anti-fascist, to be inclusive, to be open, to be working towards an evolving, shared vision of a community. Each datum is a chance for meaningful action, in my view, in everyday work.

Before we unpack that, if you’re not familiar with Riot Grrrl or Queercore, very briefly:

“Riot grrrl is an underground feminist punk movement that originated in the early 1990s.... It is a subcultural movement that combines feminist consciousness and punk style and politics.” [https://en.wikipedia.org/wiki/Riot_grrrl]

“Queercore (or homocore) is a cultural and social movement that began in the mid-1980s as an offshoot of punk. It is distinguished by its discontent with society in general, and specifically society's disapproval of the gay, bisexual, lesbian and transgender communities.” [https://en.wikipedia.org/wiki/Queercore]

You don’t need to care about either of these to get the rest of this talk, but I hope others can see the connection I’m attempting to make here. This connection between taking discontent based off of a number of reasons and making into a work ethics and an aesthetic, is really powerful for me.

----

Background image: Brighton Grrrl style. Taken by Feral78 (https://www.flickr.com/photos/emmettgrrrl/) : https://www.flickr.com/photos/emmettgrrrl/6969925930/in/photolist-bBUG9J-uGM4Ho-tN2LtE-e6UkLr-uKjWNR-uJybhf-uK2JGe-usCdDf-tNchR5-uJKuWY-usrrgQ-usFvrg-tN2udJ-tNhfdP-usArwJ-uKpgeR-usBK4y-ust9R3-tNiM44-uKb9p8-tNbbCq-uJFyQh-usK1dZ-uK3n8R-uKfy8i-uswR9S-uGHqj1-tNbyQw-uK4cCB-uKce4X-uKay3n-usBJQZ-uK7Jsg-uJDt9m-tNfQo6-uKfswP-uJzYVJ-usygGq-tN7AcY-tNjYUg-usvURy-usxYE9-usHE7Z-uGK4T7-uK7xSZ-uGMAzd-usGLfH-tN4MVd-uGJ5YW-uKjCkZ/

10 of 62

Riot Grrrl Manifesto

LibTech Manualfesto

In that sense, I use the Riot Grrrl Manifesto - a series of points about the movement’s approach to society - to guide a sort of library data technology “manualfesto”.

Manualfesto seems to be an ideal match to the code4lib spirit and how we bridge politics and work. It supports the title of this talk - which I stole from authors Ruth Ozeki (“Resistance is fertile”) and Lisa Jervis (“Manualfesto”). As part of the “library data technology manualfesto”, I’m going to discuss a few areas, roughly.

In particular, I want to link these points to specific work tools or ideas we can take in our daily work and questions we can ask ourselves to make this connection more explicit, and thus, open for improvement.

----

Riot Grrrl Manifesto image: Taken from https://b-hayden1114-dc.blogspot.com/search?q=riot+grrrl+manifesto ; Unfortunately, uncertain of license.

You can also read the full text of the Riot Grrrl Manifesto here: http://onewarart.org/riot_grrrl_manifesto.htm

More about Riot Grrrl, if you’re interested: D'Angelica, Christ. Beyond Bikini Kill: A History of Riot Grrrl, from Grrrls to Ladies, 2009. Or Marcus, Sara. Girls to the Front: The True Story of the Riot Grrrl Revolution (2011).

11 of 62

Seeing our work tied to our politics & ethics

Photo by Cory Doctorow / Creative Commons License

12 of 62

“BECAUSE viewing our work as being connected to our girlfriends-politics -real lives is essential if we are gonna figure out how we are doing impacts, reflects, perpetuates, or DISRUPTS the status quo.”

--Riot Grrrl Manifesto

13 of 62

libraryfreedomproject.org

14 of 62

How do we InfoSec? Who do we rely on?

HTTPS

HTTPS Everywhere

Signal encrypted texting and calling

1Password

LastPass

Tor Browser

VPNs

EFF’s Surveillance Self-Defense

DuckDuckGo

15 of 62

How do we handle privacy & ethics with regards to all data? Is this transparent process with a diversity of voices?

Image by Unknown / Taken from https://www.pinterest.com/pin/401805597977239254/

Such as, how do we handle our priorities, privacy, and ethics with regards to all data - not just patron data, which we seem to have a clear and shared understanding of the need to protect in library communities.

In deciding what policies we make, not just because of legal questions, but because of professional ethical questions, do we have a transparent process for those policies being made? Do we have a diversity of voices in those processes? I can say for my own work, I think for data outside of patron logs and analytics, we really don’t have a clear understanding of where information security or ethics of privacy come into play - thinking particularly about the cases such as data and metadata tracked in our digital repositories and exhibition applications.

----

Not All Riot Grrrls Are White image: Taken from https://www.pinterest.com/pin/401805597977239254/ ; Unfortunately, uncertain of license.

16 of 62

Image is taken from the presentation of Lukas Koster at ELAG 2015

As alluded to in the previous slides, not all of our data for which we need to concern ourselves with privacy and ethics ARE IN OUR CIRCULATION OR PATRON LOGS. It is increasingly spread among many, many more platforms.

This leads me to another concrete action we should be better at in libraries for a number of reasons - capturing and understanding our full data (including metadata) ecosystems. We too often tie data processes, policies, expectations, tooling and work to specific applications or subdomains - the “silos”. However, in recent work, I wonder less about how to make all the silos into one big silo, and more about what it means to have the silos talk to each other more readily, efficiently and fully than we tend to support now. The example presented on this slide is taken from the analysis work of Lukas Koster at the University of Amsterdam. It tracks ‘dataflows’, and goes beyond application diagramming for a specific domain - say, an ILS - or a particular project - say, some digital collections - and it follow data, entered, managed, how it is stored, where it ends up.

This supports a number of topics I think we should prioritize, but in this context it supports our better ability to say here is what data is where.

----

Image is taken from the work of Lukas Koster available at https://www.slideshare.net/lukask/datamazed-with-notes

Learn more about Lukas’ work at http://commonplace.net/2014/11/library-data-flows/

17 of 62

DataRefuge

datarefuge.org

Moving from data security to data preservation, the Data Refuge project and the related “Data Rescue events” are a clear place where our work meets our politics and goals. DataRefuge tries to capture and protect federal data on climate and environmental research. We will hear about this more today, but quickly, what is interesting about this work is realizing that our data is often in a precarious state not just due to shifting politics or research organization support, but inadequate management - bad formats, bad access, bad capture, bad preservation. To explore this topic further, I have some questions in my open speakers’ notes slide I invite you to respond to.

----

More about Data Refuge: https://www.datarefuge.org/

More about Data Refuge and the Data Rescue events: http://www.ppehlab.org/

Question for the audience:

So, what we mean by capture/rescue? We are constantly shifting in what counts as preservation and even as ‘data’ libraries manage. What do we preserve? What do we assign that meaning to? Importance to? Should this be expanded?

Data Refuge looks explicitly to manage preservation and access to federal and environmental data, but this approach to a scoped area of data in need of protection - including not just policies and ethics, but protocols, data harvest tools, and stores - is perhaps a model to be expanded for capture of other datasets, or even our metadata.

Additionally, we often do not manage or store our data in ways that the representations are eas[ier] to handle (using non-proprietary formats, data stores with adequate query or retrieval functionalities, having context and models coupled with data) for ease of preservation and comprehension later on.

There have been a number of efforts by our colleagues who are research data librarians to train folks in these skills, but I often wonder at our own ability in this front as library technologists who often have their vision of data limited to the particular application or job area they themselves have. In libraries, we like to pride ourselves on models and data standards, but I am not sure I’ve ever found a library dataset that fully implemented the purported data model or standard - if we can even tell what model or standard is being used without external discussions. This is especially true for anything “complex”.

18 of 62

How do we capture provenance, versioning, originating context?

Can we have “facts” without this? Shared understandings?

Work by Trevor Muñoz: https://nbviewer.jupyter.org/gist/trevormunoz/8358810

We in libraries are just starting to scratch out what our services and goals should look like around data management work - whether it is “research data”, servers full of binaries, or large aggregates of metadata.

And we in libraries hardly have many examples of solutions that can work at scale with this data and remain open, transparent, or community driven (more on this in a minute), albeit the work of places like Internet Archive on supporting that large infrastructure is inspiring. I am also inspired by scholars like Trevor Muñoz, whose Python Notebook for data manipulation is on the slide, and translates some of the recommendations we are making for researchers - enabling ability to share not just data, but context, technical and possibly other - to what we could use in our own day to day work on library metadata. Can we do better than a provenance note and a last updated date? OF COURSE WE CAN.

----

This image is a snippet from: Using Pandas to Curate Data from the New York Public Library's What's On the Menu? Project, Trevor Muñoz [10 January 2014]: https://nbviewer.jupyter.org/gist/trevormunoz/8358810

I’d also highly recommend following Trevor’s work: http://trevormunoz.com/notebook/

19 of 62

docnow.io

20 of 62

“Co-creation is much more than just consulting with the community on metadata, or tweaking a project that’s already done… We bring some knowledge of digitization, workflows and metadata, but I think we could let go of control a lot more and truly co-create with our communities”

--Tara Robertson LITA Keynote, LITA Forum 2016 Keynote

21 of 62

Work by Documenting the Now, taken from: https://github.com/DocNow/catalog/blob/master/_data/datasets.yml

22 of 62

“BECAUSE doing/reading/seeing/hearing cool things that validate and challenge us can help us gain the strength and sense of community that we need in order to figure out how bullshit like racism, able-bodieism, ageism, speciesism, classism, thinism, sexism, anti-semitism and heterosexism figures in our own lives.”

--Riot Grrrl Manifesto

23 of 62

DIY & Sharing the “Means of Production”

Photo by Greg Neate. / Creative Commons License

24 of 62

“BECAUSE we must take over the means of production in order to create our own moanings.”

--Riot Grrrl Manifesto

25 of 62

“The means of production for the archives of humanity are up for grabs, and within our reach is the possibility of new production methods that resist the recreation of existing patterns of exclusion and marginalization.”

--B. Sadler, C. Bourg. “Feminism and the Future of Library Discovery”

26 of 62

librecat.org/Catmandu

27 of 62

Code Snippet from https://gist.github.com/phochste/c87c81c79d8b8a6a2179;

While I’m not a Perl person at all - I cannot get beyond thinking the programming language looks like a bunch of emojis got beaten up - I appreciate the receptiveness of this development community to others’ ideas and questions, and my ability as an outside user to dig into codebases and try to alter some of the tooling when and if it doesn’t fit my exact need - which, with data, IT OFTEN DOESN’T.

In libraries, we too often run into data - especially metadata tools - that we cannot dig into. It solves one problem for us, according to how that one developer or programmer understands the need. But the tool now obscures how we understand interactions of data and even interactions of standards.

Catmandu works with data between systems, and leaves the ecosystem decisions up to you.

----

Code sample taken from https://gist.github.com/phochste/c87c81c79d8b8a6a2179, and here is a blog post explaining the original context: https://librecatproject.wordpress.com/2015/06/03/matching-authors-against-viaf-identities/

The drawing is by Patrick Hochstenbach and is taken from https://librecatproject.wordpress.com/2016/02/25/catmandu-1-00/. Unfortunately, uncertain of license.

28 of 62

How do you connect development to specifications to levels of implementation decisions?

Code snippet is an RML, or RDF Mapping Language Example, by the presenter. More on RML: http://rml.io/

Additionally, I am interested in how we connect but keep separate specifications from development or programming. I know that DSLs (or Domain Specific Languages) and transformation languages often end up more complex than the programming to convert data itself, and I recognize that they do not fit for all cases.

Yet, I remain interested and intrigued in DSLs because they do allow us to collaborate with a wider range of folks on our library data projects, which I think is vital. Often, the translation of data standards and specifications to programming functions and expectations is not transparent. This limits the adaptation of our colleagues’ non-programming knowledge. The data model or mapping is now married to the codebase. This can limit our tooling options, and it obscures the relationship between data expectations.

----

Code snippet is an RML, or RDF Mapping Language Example, by the presenter. More on RML: http://rml.io/

29 of 62

Screenshot is a Nifi example by the presenter for handling ETDs (Electronic Theses and Dissertations). More on Nifi: https://nifi.apache.org/

Looking beyond CLI tools like Catmandu and DSLs, something I’m particularly interested in that sits at the intersection of transparency of workflows, working with data outside of specific applications/domains, and enables __scalable__ data work for an increasingly large ecosystem and datasets, is Apache Nifi.

While I’m not certain yet if Nifi is an answer or a sledgehammer hitting a thumbtack, it is something I’ve become increasingly interested in while trying to work with data across all of our library repositories at my current place of work. I want a way to build dataflows and make them transparent - and machine repeatable, concurrent, with logged provenance and versioned data. I can also version the dataflows - which can be shared as XML files. It expands what we saw with the data ecosystem analysis work before.

Apache Nifi is now an Open Source project, but originated in the NSA or National Security Agency for them to manage metadata for … well, tracking us. Which means we know it’s metadata management capabilities at scale and in a distributed fashion have been tested, ha. But there is also the tension between using a tool built for surveillance we already discussed as a problem (or possible oppressive), and the needs this tool can fill in our adaption of it - primarily, extremely fast, concurrent ETL (or extract-transform-load) work. Can we adequately reclaim these tools?

----

Screenshot is a Nifi example by the presenter for handling ETDs (Electronic Theses and Dissertations). More on Nifi: https://nifi.apache.org/

30 of 62

“And when we talk about these kinds of racist experiences and pointers that happen in technical systems, we also hear in the public discourse these things talked about, again, as anomalies, as glitches, rather than helping us to understand and unveil the ways that programmers are people who write, and code is a language. And all languages are value-laden, including binary code languages.”

--Safiya Noble, Biased Data

Let’s move now to another technology that intersects with biases and needs that have been brought up multiple times in our communities and meetings - search and ranking data algorithms. This quote from Safiya Noble discusses the bias of code, particularly in the context of the issue of search results of interfaces like Google and how they reflect racist, sexist and other biases in their results ranking. If you do not know of Safiya Noble’s work, I’d highly recommend listening to her Keynote at DLF Forum 2015.

In our work with data and trying to surface and counter bias (I would argue we cannot remove it entirely), both the algorithms of our discovery or access interfaces AND our data models and standards need to be reviewed. This algorithm and data model conscious development sets clear our politics, our data, and our approach, and really shows where we diverge from simply trying to be “like Google”.

----

Quote taken from Safiya Noble at Biased Data [2015]: http://opentranscripts.org/transcript/safiya-noble-biased-data/

Hear Safiya Noble’s DLF Forum 2015 Keynote here: https://open.library.ubc.ca/cIRcle/collections/55474/items/1.0220808 or read more of her work here: https://safiyaunoble.com/

31 of 62

How do you document your decisions in creating “discovery”?

Background image: SharedShelf Metadata Fields to Solr Mapping, a simple example of starting to document where and how our models and our data discovery intersects (and could be a place to start reviewing for bias and ways to balance this).

Fighting biases in algorithms and data models again highlights the need for transparent and open processes for our data and development work. It again highlights that we need to have a diversity of voices can not only review the work, but also alert us to issues. We need to have a process in which we can resolve these in our data models, in our data capture, in our algorithm development, and in our tooling - and in all of the policies and workflows around those steps.

On that note, how do you then document your decisions in creating discovery?

----

Background image: SharedShelf Metadata Fields to Solr Mapping, a simple example of starting to document where and how our models and our data discovery intersects (and could be a place to start reviewing for bias and ways to balance this).

32 of 62

How do you document the decisions in deciding on your resource models?

Who is involved in balancing modeling, bias, politics & work needs?

Background image: CIDOC-CRM Conceptual Reference Model page for P93 took out of existence (was taken out of existence by) in Version 6.2: http://www.cidoc-crm.org/Property/P93-took-out-of-existence/Version-6.2

Or, how do you account for bias in your models? Who is involved in managing that, and where is it documented (again on the documentation)?

I’d reiterate here two things here:

We cannot wait for perfect in our models in any sort of manner, if only because we alone cannot necessarily see or uncover our own biases in data modeling and metadata work.
We need to have methods for iterative updates - from the modeling to the implementations. This means models versioning, data versioning, code versioning, update mechanisms, clear logging of provenance, batch update methods, and processes for a diversity of voices to be involved in this work - among other work areas.

----

Background image: CIDOC-CRM Conceptual Reference Model page for P93 took out of existence (was taken out of existence by) in Version 6.2: http://www.cidoc-crm.org/Property/P93-took-out-of-existence/Version-6.2

33 of 62

librarycarpentry.github.io

Now I’m going to shift from inclusion of multiple and diverse voices in our data modeling and tools to how we help get more people working in that space. I sincerely don’t believe everyone needs to learn to script or program in order to take part in the work discussions I’ve lined out previously, and in some ways, I’d love to have metadata standards workshops for programmers to reverse this trend a bit.

Yet, there is something to be said for empowering people with the pieces - the small bits of code referred to by Hector, perhaps - they need to make their own tooling, not just use a GUI.

Library Carpentry has been introduced to you during the conference. I love that all of the curriculum materials are available online and open for PRs - add your thoughts! Comments! Examples! Ideas!

----

Image taken from the Library Carpentry homepage.

Learn more about Library Carpentry: https://librarycarpentry.github.io/

34 of 62

What is the last workshop you ran at your place of work? Do you explicitly invite all “classes” of workers?

Photo by Dorret / Creative Commons License

35 of 62

Have you helped coordinate a regional or local Code4Lib? Why not?

Or, taking this a bit broader, have you considered running local or regional events to expand that collaboration to your neighbors? We *are* allowed to work together outside of conferences and ordained committees (this is a joke). And Code4Lib is us - take it and run. I’ve loved being involved in local and regional code4libs - they have been a chance to interact with my neighboring library technologists who cannot make it here to the national or haven’t felt comfortable jumping into Code4Lib yet.

----

Background image is from: ©Cornell University Marketing Group, 301 College Ave, Ithaca, NY 14853; 607-255-7675; photo@cornell.edu; http://marketing.univcomm.cornell.edu.

This image was originally used for the Code4Lib New York State Midsummer Meeting website, a regional that is an example of what this slide is mentioning: https://code4libnys.github.io/2016/

36 of 62

Openness & Transparency

37 of 62

“It seems to me that we’ve spent enough time on closing down our information systems while attempting to broaden our influence through the production and delivery of innovative library programs and services. When we support the development and use of closed systems we’re failing our communities.”

--Cecily Walker, LITA Forum 2016 Keynote

38 of 62

Open Data

Image taken from “How can I get data out of Wikidata”.

Wikidata, for all it’s faults (it isn’t the best of Linked Data examples), is a great example of an Open Data project or framework - in my opinion. It helps set us up with the questions:

Where possible, is your data open?

Did you document the discussions that decided on the level of openness?

Do you revisit those decisions often?

Are there fully updated and described dumps of that data available?

What about an API that follows an open and easily find-able specification?

Does the open data leverage some sort of existing standard, but expand it?

Note, this goes beyond just making some data available on GitHub. We need to think more broadly about what open means - in terms of access (do people know about your API or data dump? Is it updated often and representative of the dataset?) and managed (do you version?).

----

Image taken from “How can I get data out of Wikidata”: https://www.wikidata.org/wiki/Wikidata:Data_access#How_can_I_get_data_out_of_Wikidata.3F

39 of 62

We talk about our responsibility to the collections, but what about our responsibility to communities.

-- Tara Robertson LITA Keynote, LITA Forum 2016 Keynote

40 of 62

Open Source Development

Then, there is open source development, something in particular our code4lib community is really invested in. I would like to propose some questions that indicate my questions for open development:

Where possible, is your codebase open? Where and why is this not true? What people are you leaving out of your work by doing this?

Is the documentation for use and development up to date? What about documentation where able of your decision process for things like development priorities?

Does your application make resources open but also transparent? Can we tell (if we want to) how your search ranking works? Why that was decided upon? Can we pull the source data and be able to figure out the modeling?

When interacting with open tools or data, did you review your presumptions? Did you make a request that clearly documents your POV? Your use case? See where it fits in an existing priority of development that also balances available resources to develop? Did you attempt a PR with that development request?

41 of 62

writethedocs.org

None of this speaks to OSS development processes specifically, which this audience honestly would know better than I do. But it does lead me to something we all know we should do better at: Documentation, documentation, documentation.

Documentation not only helps us, but it enables others to take part. Docs create inclusivity, and as such, should be a much higher priority for work than they tend to be - whether we are documenting our application codebases or documenting our decision processes or documenting our models.

On the topic of documenting data models in particular, there is some focus on simply documenting what standard(s) we are using for metadata - and then presumption that our implementation entirely matches that standard. It is an example of where we both lose the concrete place in which our abstractions occur, but also often hinders implementation and updates because we so rarely take the standard as is in our instance data.

While WriteTheDocs has a developer focus, I’ve had great experiences with expanding their open curriculum to broader library work - including documenting data models and our decision processes.

----

More information on Write the Docs: http://www.writethedocs.org/

42 of 62

“BECAUSE we are interested in creating non-hierarchical ways of being AND making music, friends, and scenes based on communication + understanding, instead of competition + good/bad categorizations.”

--Riot Grrrl Manifesto

43 of 62

WE REQUIRE TRANSPARENCY

FOR OPEN ORGANIZING & COLLABORATION

Often communities get built around open data or open source projects, but how should we approach community organizing work in our professional and political space as well?

I like to think about this through another series of questions - such as:

Is the community open and transparent about who makes decisions and how?

Can anyone freely call together new meetings, convene sessions, ask questions, be heard?

Does it fall to 1 or 2 “king makers” to make the “community” work?

This leads me to say to you: Transparency always. Not as an afterthought. Not as an “if we have time”.

Transparency is 1 step (but not the end goal) to work that says, not only am I open to collaboration, I expect collaboration. Then others can participate, but also, we can all learn. Critique. Analyze. See what we couldn’t see on our own. And expand.

----

Background photo from https://pixabay.com/en/hands-working-together-teamwork-952510/, CC license.

44 of 62

Re-making our Power Hierarchies & Structures, Explicit & Implicit

Transitioning from openness, what does this connection between our politics and our daily work mean for the power structures we interact with? Particularly in library data technology, we are at the intersection of a number of power structures, both obvious and hidden.

There is that of skills - librarians “or” developers.

There are our HR (human resource) hierarchies - paraprofessional versus professional.

There are hierarchy of resources and communities served - public versus private institutions, those with large endowments - such as my own place of work - and those working on smaller budgets, etc..

First, though, I want to explore a particular power structure or situation that I encounter a fair amount.

----

Background image: Draft Record Liner Lyrics, "Rebel Girl," circa 1993. The Kathleen Hanna Papers. Taken from http://www.papermag.com/johanna-fateman-on-the-riot-grrrl-movement-images-from-the-riot-grrrl--1426932930.html Unfortunately, license unknown.

45 of 62

“Relying on centralized authority management or metadata creation for everything, and the corporatization of library infrastructure, actively resists that decentralizing force, further limiting our own effectiveness in the construction of radical democracy.”

--Mark Matienzo, 2015 LITA Forum Keynote

Centralization.

How centralization of our data and standards evolved over time in libraries does make some sense. It’s efficient. It coordinates our processes across a complex ecosystem. It handles a number of viewpoints and needs.

I appreciate and will read the full quote that I take a snippet of for this slide. This is from Mark Matienzo’s keynote To Hell with Good Intentions, because we are setting up the issue of our metadata and resource centralization within a very, very complicated landscape:

“Centralizing the process of naming in any context overly concentrates that power. I am not saying that all vendors are inherently bad, or that institutions responsible for metadata standards and authority files are bad. Yes, there’s a lot of work to be done to make linked data itself easier to publish, consume, and reuse for all kinds of institutions and communities, but I worry about our deference to this centralization. Despite my concerns about the lack of access to effective user-facing tools for linked data, I still believe its power is in its ability to leverage that decentralization. Relying on centralized authority management or metadata creation for everything, and the corporatization of library infrastructure, actively resists that decentralizing force, further limiting our own effectiveness in the construction of radical democracy.”

----

Quote taken from the LITA Forum 2015 Keynote, To Hell with Good Intentions by Mark Matienzo: http://matienzo.org/2016/to-hell-with-good-intentions/

46 of 62

“Finally, create opportunities for the community to participate meaningfully. Not just listening but actively engaging…. Listening to and acting on feedback is important, but think about ways for others to get involved. Can you provide APIs to let developers and users actually do something directly?”

--Rob Sanderson, Linked Pasts Keynote 2016

47 of 62

uh oh...

These quotes and my interpretation of them immediately hits on two main players in the centralization power structure, players that regularly appear in my daily work.

I want to emphasize here, explicitly, this is not about the people there. This is about the expectations we have placed on these institutions and how we, the community using library data technology, need to shift our understanding in tandem with the institutions as a whole. The people at these places are amazing.

That said…

----

Image on the left: The Library of Congress in 1898. Image is in public domain. Taken from https://en.wikipedia.org/wiki/Library_of_Congress#/media/File:The_Library_of_Congress,_Washington-LCCN2008678216.jpg

Image on the right: The Kilgour Building, the headquarters of the library non-profit OCLC. Image is CC BY-SA 3.0, taken by Matkatamiba. Taken from https://en.wikipedia.org/wiki/OCLC#/media/File:Kilgour_Building,_OCLC,_Dublin,_OH.jpg

48 of 62

Because this is not acceptable...

Image taken from the id.loc.gov Robots.txt file.

This is the id.loc.gov robots.txt file. We’ve seen this file in code4lib presentations before. In case you don’t know it already, it requests no more than 1 HTTP request per 3 seconds. For a supposedly Linked Open Data service hub.

This doesn’t support production work in any sense of the imagination. This doesn’t support community involvement, as id.loc.gov development is often shared with the public after it is “complete”. It doesn’t support data interactions beyond basic testing.

Sure, there are the RDF data dumps, but those incomplete, out of date, and the modeling has not adequately changed since they first appeared.

I know there is SRU and Z39.50 for the MARC (albeit, going that route, we run into issues of identifier management). I’m not saying that all of LC’s efforts in networks and standards are baseless - they serve some value in trying to work with a wide-ranging and opinionated professional domain.

But our understandings of what data collaboration means - whether in modeling development or in data access - needs to evolve. This is not acceptable.

----

The image is taken from the id.loc.gov robots.txt file available at: http://id.loc.gov/robots.txt

49 of 62

Because this is not community collaboration...

Image taken from the BIBFRAE listserv website.

Going a bit further, this is an image of a portion of the BIBFRAME listserv interface.

I don’t think the listserv itself is a bad technology - though it does tend to bring out the worst in us often. What I worry about is the lack of genuine community input process for the modeling, mapping, or work around BIBFRAME.

Where are the working documents? Where is the evidence that our community has a place to engage, not simply ask a question after the fact? Is it only through archaic power structures that this happens? And if so, who are the gatekeepers for those power structures that govern involvement?

I want to be clear on one point - I do think we need some RDF model for handling bibliographic materials going forward. I don’t think BIBFRAME 2 is without merit - it has a place. I say this because developers and code4lib members tend to jump immediately to being anti-BIBFRAME, without doing any sort of adequate review.

What is missing with BIBFRAME is any ability for community engagement in working on a standard that will affect a huge number of the people in our professional domain. I know community engagement at this scale is not easy. But that, honestly, is what needs to happen.

----

Image taken from the BIBFRAME listerv site: https://listserv.loc.gov/cgi-bin/wa?A0=bibframe

50 of 62

& unclear & opaque consortia are not an acceptable alternative...

Image taken from the OCLC Metadata Management FY17 United States Price List 1

Now, are there other solutions to the problems around BIBFRAME and the Library of Congress’ exposure of bibliographic data, standards and models? Not really.

So this slide is the unclear price list for OCLC membership levels. I attempted to decode our OCLC bill at my previous position. I was unable to determine what we paid to this non-profit consortium for what services, as well as unclear what collaboration opportunities existed beyond purchasing a service.

We know that OCLC Research are doing brilliant work. I wouldn’t want to stop that for anything, nor the OCLC APIs, which are often the only scalable way to access a version of bibliographic and authority data.

But how does that map back to what has become the paid access point for many institutions to OUR OWN data - especially those institutions who aren’t as well funded as others? What expectations or modeling decisions by OCLC alone are built into the data exposed via their APIs? What presumptions of the models behind things like VIAF and FAST are not explicit or open to community development? How do we manage access to datasets like OCLC Works, which are research, certainly, but have immediate value now?

And how much are we paying for what?

----

Image taken from the OCLC Metadata Management FY17 United States Price List 1: https://www.oclc.org/wcsstore/OCLC/files/pricelist_us_L4Y34Hsp634O7ywZX83Pk49WD8.pdf

51 of 62

& we need to get that having a standards committee is not necessarily community collaboration.

The PCC (Program for Cooperative Cataloging) may be less applicable to the majority of the folks here at code4lib, but this purports to take on some of the role of managing standards like MARC or eventually BIBFRAME. Yet, as a community and specifically for those that interact with PCC, we have to get past the idea that a committee made up of practitioners from well-resourced and self-selected institutions are who, “naturally”, should be leading our standards, our abstractions, our data modeling, our management of what is essentially a communal, professional data ecosystem.

We need to also see the issues of lack of technology for support of distributed or even centralized datasets in this space, further separating the work of these standards and data centralized organizations and our daily work with our tools and technologies.

----

Image is a screenshot of the Program for Cooperative Cataloging website. Taken from https://www.loc.gov/aba/pcc/

52 of 62

“It is learning how to take our differences and make them strengths. For the master's tools will never dismantle the master's house. They may allow us temporarily to beat him at his own game, but they will never enable us to bring about genuine change. And this fact is only threatening to those women who still define the master's house as their only source of support.”

--Audre Lorde

In particular, the issues I’ve tried to raise with regards to specific institutions that our community has supported and built over time, makes me think of this Audre Lorde quote, which could honestly also apply to the very white and middle class riot grrrl or code4lib.

This quote is taken from a lecture meant to specifically react to the lack of women of color, poor women, and queer women in a feminist academic meeting in NYC. It argues against falling prey to the false consensus that limits our options. We should instead act out with our own power, not continue to work in another’s paradigm.

This is not to say that we shouldn’t build something, or use what we have now to effect change or just do our day to day work, but we must see the connection between our self-selected representation in those structures that decide our data work - those abstractions, those biases, those access, tools, power - and the larger communities we are meant to be a part of, not doing the work for.

This is also not to say that OCLC or LC do not have a place - THEY DO - but the power structures around the sharing of our data and models must shift.

----

Quote is taken from The Master's Tools Will Never Dismantle the Master's House by Audre Lorde. http://collectiveliberation.org/wp-content/uploads/2013/01/Lorde_The_Masters_Tools.pdf

53 of 62

This is not an error of theirs, but an error of ours.

54 of 62

iiif.io

55 of 62

Yet we in libraries hardly have many examples of solutions that can work at scale with this data and remain open, transparent, or community driven. Distributed Networks are something other communities are thinking about, however. IPFS is an example of this - and while still in construction, this seems an ideal technology for the questions of managing distributed data.

Additionally, we need to think about performance, especially when our single datastore / centralized systems already don’t scale well, necessarily. Is this a place for big data technologies like Apache Kafka, Spark, Hadoop, other? Because none of us alone are big data, but all of together in this peer to peer distributed ideal will be approaching big data.

----

Image taken from the IPFS homepage: https://ipfs.io/

56 of 62

“Are we prioritizing the ontologist and data manager, or the developer that has to work with the result? In my experience, we tend to meet our own requirements first and hope that the developer can make do with what they get.”

--Rob Sanderson, Linked Pasts Keynote 2016

There is another type of power structure we need to consider, however - that of the theorist and that of the implementer. We, myself included in this library technology ‘we’, get stuck on theorizing + discussing the ideal ad nauseum, then building the incomplete. Or worse, specifying the incomplete for someone else to build. Then we get annoyed at the incomplete, intimidated by the idea of iterated work or continual maintenance, and we look for a quick idea or technology to get us the rest of the way in that work. This idea could be that MARC must die. Or that Linked Open Data will save all the stuff. Or that our OSS communities are what is needed to save our work as a whole.

This is where our proposed technology ‘solutions’ are obscuring the much more pervasive fundamental shift in our professional politics and ethos - in my humble opinion.

When should we be communicating better with each other?

----

Quote is taken from the Linked Pasts 2016 Keynote, Community Challenges for LOD by Rob Sanderson: https://gist.github.com/azaroth42/a7f9dd25ccf9e6679dd7f67bfed4e050

57 of 62

COLLABORATORS, NOT CLIENTS.

COMMITTERS & MAINTAINERS, NOT DEPLOYERS.

COLLEAGUES, NOT STAFF.

AGENTS, NOT THE AUDIENCE.

58 of 62

Territorialism is mad efficient…

But it will be the death of us.

59 of 62

BECAUSE I believe with my wholeheartmindbody that girls constitute a revolutionary soul force that can, and will change the world for real.

--Riot Grrrl Manifesto

60 of 62

BECAUSE I believe with my wholeheartmindbody that code4lib constitutes a revolutionary soul force that can, and will change the world for real.

I believe in this community so strongly, it disappoints me when we don’t achieve more, or when we get stuck thinking about what we should be in some ideal future, not what we are now, or when we split ourselves among ‘coders’ and ‘not coders’.

We don’t get paid enough. We work too much. Yet, we are here. I imagine, this is because you, like me, want to be engaged in your work. Let’s make this happen, through thinking about how our politics and our ethics intersect with our daily work and technology. By collaborating, always being open and transparent, by bringing in diverse voices for discussion of our policies and procedures, by building new inclusive power structures, by building awesome new data infrastructures and models!

And whether or not you care about this idea of the library data technology manualfesto, I hope some of these points lands for you too. Riot grrrl or no, metadataist or developer, librarian or administration, you do you, and lets aim for engagement so we can head for....

61 of 62

62 of 62

Resistance is Fertile

Building a Manualfesto for Lib Tech

Christina Harlow @cm_harlow code4lib 2017�bit.ly/c4l17harlow