Digital libraries and institutional repositories

Library and Information Science (B.LISc) · Library technology · free study notes for exams, placements and interviews

In short

A digital library is an organised, managed collection of digital objects with services for search, access and preservation; an institutional repository (IR) is a special case - an open digital archive of the intellectual output of one institution (theses, papers, reports, question papers, courseware). The subject is examined in B.LISc / M.LISc papers on Information Technology Applications, Digital Libraries and Knowledge Organisation, and in UGC-NET Library and Information Science Paper II. In practice, Indian employers (IITs, IIMs, IISc, central and state universities, CSIR and ICAR labs, DRDO, ISRO, NIC, publishers and corporate knowledge centres) hire LIS graduates specifically to install and run DSpace or Koha-plus-DSpace stacks, create metadata and upload theses to Shodhganga. Interview rounds concentrate on DSpace workflow, Dublin Core elements, OAI-PMH harvesting and the UGC mandate on electronic submission of theses.

Explained

Definitions, evolution and types

Digital library: a collection of digital objects (text, image, audio, video, data) with metadata, organised and curated, searchable, and served over a network with defined access and preservation policies. Digitised + born-digital content both count.
Distinguish: digitised library (scanned analogue holdings) · digital library (managed collection + services) · virtual library (pointer to resources held elsewhere) · hybrid library (print and digital managed together - the real state of most Indian college libraries).
Institutional repository (IR): defined by Clifford Lynch (2003) as a set of services an institution offers to its community for the management and dissemination of digital materials created by that community. Essentials: institutionally defined, scholarly, cumulative and perpetual, open and interoperable.
Types of repository: institutional (Shodhganga at INFLIBNET, IISc ePrints, IIT repositories) · subject/disciplinary (arXiv, PubMed Central, RePEc) · national/aggregator (NDLI at IIT Kharagpur) · data repositories (research data) · thesis repositories (Shodhganga, Shodhgangotri for synopses) · consortium repositories.
Five functions of a digital library: selection/acquisition, organisation (metadata), storage and preservation, access and retrieval, user services.
Indian landmarks: IISc ePrints (2002, first Indian IR, EPrints software) · Shodhganga (2010-11, INFLIBNET, Indian ETDs) · NDLI (National Digital Library of India, MoE, single-window access to lakhs of learning resources) · e-ShodhSindhu (merged UGC-INFONET / INDEST / N-LIST consortia for e-journal subscription) · National Mission on Libraries. Digital Library of India (DLI) at IISc/CMU was the early mass-digitisation project.

Open access, mandates and licensing

Open access (OA): free online availability with permission to reuse. Declared by the Budapest Open Access Initiative (2002), Bethesda (2003) and Berlin Declaration (2003).
Green OA = self-archiving the author's manuscript in a repository (what an IR does). Gold OA = publishing in an OA journal, often with an article processing charge. Diamond/Platinum = OA journal with no APC. Hybrid = subscription journal with paid-OA articles. Bronze = free to read, no reuse licence.
Versions: preprint (before peer review) · postprint / author accepted manuscript (AAM) - what most publishers permit in an IR · publisher PDF (version of record) - usually not permitted. Embargo = delay before the deposit becomes public. SHERPA RoMEO lists publisher self-archiving policies.
Indian policy: the UGC (Minimum Standards and Procedures for Award of Ph.D. Degree) Regulations, 2022 require universities to submit an electronic copy of every awarded thesis to INFLIBNET for hosting in Shodhganga, accessible to all. DBT-DST Open Access Policy (2014) requires research funded by them to be deposited in an institutional or central repository. ICAR, CSIR and ISRO run their own repositories. The One Nation One Subscription (ONOS) scheme, operational from 1 January 2025, provides centrally funded access to journals for government institutions - note it is a subscription-access scheme, not an OA mandate.
Licensing: Creative Commons - CC BY · CC BY-SA · CC BY-NC · CC BY-ND and combinations; CC0 places work in the public domain. Avoid NC/ND if the aim is maximum reuse.
Copyright in India: the Copyright Act, 1957 (as amended, notably in 2012). Section 52 fair-dealing exceptions permit certain educational and library reproduction; the 2012 amendment added s.52(1)(zb) for accessible-format conversion for persons with disabilities. Author's economic rights last the author's lifetime plus 60 years. A repository needs a deposit licence (a non-exclusive right to archive and distribute) signed by the author - it does not need copyright transfer.

Metadata, standards and interoperability

Metadata = structured data about a resource. Kinds: descriptive (title, author, subject), structural (how pages/chapters fit together), administrative (rights, provenance, technical, preservation).
Dublin Core (DC): the default repository schema; ISO 15836 / NISO Z39.85. The 15 unqualified elements: title, creator, subject, description, publisher, contributor, date, type, format, identifier, source, language, relation, coverage, rights. Qualified DC adds refinements (dc.date.issued, dc.contributor.advisor) and encoding schemes. All elements are optional and repeatable.
Other schemas: MARC 21 / MARCXML (library catalogues) · MODS (richer, MARC-compatible XML) · METS (packaging structure + metadata) · PREMIS (preservation metadata) · EAD (archives) · VRA Core (visual) · ONIX (publishing) · LOM (learning objects) · DDI (social-science data).
Interoperability: OAI-PMH (Open Archives Initiative Protocol for Metadata Harvesting) - HTTP-based, XML responses, six verbs: Identify, ListMetadataFormats, ListSets, ListIdentifiers, ListRecords, GetRecord. Unqualified oai_dc is the mandatory format. Distinguish data provider (the repository, exposes metadata) from service provider (the harvester - aggregators such as CORE, BASE, OpenAIRE, NDLI).
Newer layers: ResourceSync (synchronisation, successor approach to harvesting) · SWORD (deposit protocol) · Signposting; Z39.50 and SRU/SRW are search-and-retrieve protocols, not harvesting.
Persistent identifiers: Handle System (DSpace default, e.g. 10603/12345 for Shodhganga) · DOI (a Handle with a registration agency - Crossref, DataCite) · URN, ARK, PURL; ORCID for authors, ISNI for organisations, ROR (Research Organization Registry) for institutions. Purpose: resolve to current location even when the URL changes.
Registries: OpenDOAR (repositories), ROAR, ROARMAP (mandates), DOAJ (OA journals), DOAB (OA books), FAIR principles (Findable, Accessible, Interoperable, Reusable) for research data.

Repository software and the DSpace workflow

DSpace - most used in India; Java, PostgreSQL (Oracle support was dropped from DSpace 8), Solr index, Handle server, Tomcat; DSpace 7 and later use an Angular front end talking to a Java REST API back end (the dspace-server webapp); DSpace-CRIS is a separate 4Science extension, not the DSpace back end. Content model: Community → Sub-community → Collection → Item → Bundle → Bitstream. Item = metadata + bundles (ORIGINAL, THUMBNAIL, TEXT, LICENSE); bitstream = the actual file.
EPrints - Perl/MySQL, used by IISc ePrints; strong for article self-archiving. Greenstone - New Zealand Digital Library/UNESCO, C++/Perl, excellent for CD-ROM and offline collections and multilingual content, uses its own Greenstone Archive Format. Fedora (repository architecture, often with Islandora or Samvera/Hyku front ends) · Omeka (exhibits) · Invenio (CERN, powers Zenodo) · Dataverse (research data) · CONTENTdm, Digital Commons (commercial/hosted).
DSpace submission workflow: (1) depositor selects a collection; (2) describes the item - fills the DC metadata form; (3) uploads bitstreams; (4) verifies; (5) grants the distribution licence; (6) then step 1 Accept/Reject reviewer, step 2 Accept/Reject/Edit metadata, step 3 Edit metadata only; (7) on approval the item is installed, gets a Handle and is indexed. Collections may enable any subset of the three review steps.
Other DSpace features: batch import (Simple Archive Format: a directory per item with dublin_core.xml and contents), CSV metadata import/export, embargo on bitstreams, item withdrawal (tombstone, Handle retained) versus deletion (expunge), groups and resource policies (READ, WRITE, ADD, REMOVE, ADMIN), e-person accounts, Solr-based statistics, OAI-PMH endpoint at /oai/request.
Digitisation process: selection → copyright clearance → preparation/collation → scanning (flatbed, planetary/overhead, book scanner; 300-600 dpi for text, 400-600 dpi plus colour targets for images) → OCR (Tesseract, ABBYY; Indic-script OCR accuracy is markedly lower than for Latin script) → quality control → metadata creation → derivative creation (archival master TIFF, access JPEG/PDF-A) → upload and checksum → backup.
Formats: archival masters - TIFF (uncompressed), PDF/A (ISO 19005), WAV, FFV1/MKV, XML/UTF-8, CSV; access copies - JPEG, JPEG 2000, PDF, MP3, MP4. Prefer open, well-documented, widely used, lossless, unencrypted formats.

Preservation, architecture and evaluation

OAIS reference model (ISO 14721): entities Ingest, Archival Storage, Data Management, Administration, Preservation Planning, Access; packages SIP (submission), AIP (archival), DIP (dissemination). TDR audit standard: ISO 16363; CoreTrustSeal certification.
Preservation strategies: refreshing (copy to new media) · migration (convert to current formats) · emulation (run old software environment) · normalisation (convert on ingest to a standard format) · technology preservation (keep old hardware) · bit-level preservation with checksums (MD5, SHA-256) and fixity checks · LOCKSS/CLOCKSS distributed copies. Keep 3 copies, 2 media, 1 off-site (the 3-2-1 rule).
Threats: media decay, format obsolescence, software/hardware obsolescence, loss of metadata and context, organisational failure and funding loss, link rot (hence persistent identifiers).
Architecture layers: content/storage → metadata and index (Solr/Elasticsearch) → application/repository logic → interface (OPAC-like discovery, responsive UI) → interoperability (OAI-PMH, REST API, SWORD) → identity and access (LDAP/Shibboleth single sign-on).
Evaluation and impact: usage statistics (COUNTER for e-resources; Solr statistics in DSpace), deposit rate per faculty, altmetrics, citation advantage studies, Webometrics Ranking Web of Repositories, server uptime, metadata completeness and error rate.
Advantages: 24×7 remote multi-user access, no physical wear, full-text search, space saving, increased visibility and citation of institutional research, preservation of rare and local material, support for distance learners, compliance with UGC/funder mandates, low marginal cost per additional user.
Limitations: high initial and recurring cost (scanning, storage, staff), copyright barriers, format obsolescence and migration burden, bandwidth and power dependence, low faculty self-archiving without a mandate, plagiarism and misuse fears, digital divide and language/Indic-script issues, need for continuous skilled staff, preservation is an unending commitment not a project.

Setting up and running an IR - practical sequence

Policy first: scope of content, who may deposit, accepted formats, metadata standard, embargo rules, withdrawal policy, preservation commitment, licence - approved by the institution's authority.
Advocacy: mandate from the Vice-Chancellor/Director (ROARMAP-type), faculty orientation, mediated deposit by library staff since author self-deposit is low.
Technology: Linux server or cloud VM; DSpace with Java, Tomcat, PostgreSQL, Solr, Handle prefix from CNRI or via INFLIBNET; HTTPS, backups, UPS.
Build: define Communities (departments) and Collections (document types), configure the submission form (submission-forms.xml), controlled vocabularies, item templates, authority control for author names.
Populate: retrospective digitisation plus current output; batch import of back files; register the OAI endpoint with OpenDOAR and have NDLI/BASE harvest it.
Operate: metadata quality checks, link and fixity checks, statistics reporting, annual policy review, plagiarism screening before thesis deposit, accessibility (WCAG) and multilingual interface.
Indian legal/practice notes: content hosting is governed by the Copyright Act, 1957 and the Information Technology Act, 2000 (intermediary due diligence); for personal data in repositories the Digital Personal Data Protection Act, 2023 applies, but under the DPDP Rules, 2025 the Data Fiduciary duties (notice, consent, breach intimation, erasure) commence only from 13 May 2027 - until then IT Act s.43A and the SPDI Rules, 2011 govern sensitive personal data. RPwD Act, 2016 requires accessible formats for persons with disabilities (Sugamya Pustakalaya is the Indian accessible-book repository).

Key points

Digital library - managed collection of digital objects plus services for access and preservation, not merely a pile of scanned files.
Institutional repository - open archive of one institution's own intellectual output; Lynch's definition is the one examiners want.
Green vs gold OA - green is self-archiving in a repository; gold is publishing in an OA journal, often with an APC.
Shodhganga - INFLIBNET's ETD repository; Indian universities must deposit awarded Ph.D. theses under UGC Ph.D. regulations. Shodhgangotri holds synopses/research proposals.
NDLI - National Digital Library of India, IIT Kharagpur, MoE; an aggregator, not a publisher of content.
e-ShodhSindhu - consortium for subscribed e-resources; ONOS (from 1 January 2025) centralises journal subscriptions for government institutions.
Dublin Core - 15 optional, repeatable elements; ISO 15836; oai_dc is the mandatory OAI-PMH format.
OAI-PMH - six verbs; repository = data provider, harvester = service provider.
DSpace hierarchy - Community → Collection → Item → Bundle → Bitstream; Handle assigned on install.
Workflow steps - describe, upload, verify, licence, then up to three review steps before the item goes live.
Handle/DOI - persistent identifiers that resolve despite URL change; Shodhganga Handles begin 10603.
OAIS (ISO 14721) - SIP → AIP → DIP, six functional entities; ISO 16363 audits trustworthiness.
Preservation strategies - refreshing, migration, emulation, normalisation, bit-level fixity; follow 3-2-1 copies.
Archival master vs access copy - TIFF/PDF-A master, JPEG/PDF derivative; 300-600 dpi for text scanning.
Software map - DSpace (Java/PostgreSQL), EPrints (Perl/MySQL), Greenstone (offline/multilingual), Fedora, Invenio, Dataverse, Omeka.
Deposit licence, not copyright transfer - the author keeps copyright and grants a non-exclusive right to archive.
Mandate beats persuasion - deposit rates stay low without an institutional mandate and mediated deposit by library staff.
Copyright Act, 1957 - s.52 fair dealing and library exceptions; author's life + 60 years for economic rights.

Formulas and examples

Storage per scanned page (uncompressed TIFF, bytes) = (width_in × dpi) × (height_in × dpi) × bit-depth / 8
Total storage = pages × size per page × (1 + derivative overhead) × copies
Digitisation cost = pages × (scanning rate per page + OCR rate per page) + metadata cost per item × items
Throughput = pages per hour × productive hours per day × working days
Metadata completeness (%) = (filled mandatory elements / (mandatory elements × items)) × 100
Deposit (compliance) rate (%) = items deposited / items published × 100
Cost per download (₹) = total annual repository cost / annual full-text downloads
Relative recall = relevant retrieved by one system / total relevant retrieved by all systems; Precision = relevant retrieved / total retrieved

Example 1 - storage for one scanned page. An A4 page (8.27 in × 11.69 in) scanned at 300 dpi, 8-bit greyscale.

Pixels: 8.27 × 300 ≈ 2,481 across; 11.69 × 300 ≈ 3,507 down.
Pixels per page = 2,481 × 3,507 ≈ 8,700,867 ≈ 8.70 megapixels.
Bytes at 8 bits (1 byte) per pixel ≈ 8,700,867 B ≈ 8.3 MB (÷1,048,576).
At 24-bit colour it is 3 bytes/pixel ≈ 24.9 MB; at 1-bit bitonal ≈ 1.04 MB. Answer: ≈8.3 MB greyscale, ≈24.9 MB colour.

Example 2 - project storage and cost for a thesis collection. 500 theses × 250 pages = 125,000 pages, greyscale 300 dpi masters at 8.3 MB plus access PDFs at 0.4 MB per page.

Masters: 125,000 × 8.3 MB = 1,037,500 MB ≈ 1,013 GB ≈ 0.99 TB.
Access PDFs: 125,000 × 0.4 MB = 50,000 MB ≈ 48.8 GB.
Total one copy ≈ 1,062 GB ≈ 1.04 TB; with 3-2-1 (three copies) ≈ 3.11 TB.
Cost at ₹1.20 per page scanning + ₹0.80 per page OCR = ₹2.00 × 125,000 = ₹2,50,000; metadata at ₹60 per thesis = ₹30,000. Total ≈ ₹2,80,000 (rates are indicative - vendor quotes vary widely).

Example 3 - throughput and schedule. Two operators, each 45 pages/hour effective, 6 productive hours/day, 22 working days/month.

Per operator per month = 45 × 6 × 22 = 5,940 pages; two operators = 11,880 pages/month.
125,000 ÷ 11,880 ≈ 10.5 months. To finish in 6 months: 125,000 ÷ (6 × 22 × 6 × 45) = 125,000 ÷ 35,640 → 3.5, so 4 operators.

Example 4 - deposit rate and cost per download. A university published 420 papers in a year; 273 were deposited in the IR. Annual repository cost ₹6,40,000 (staff ₹4,20,000, server ₹1,20,000, other ₹1,00,000); full-text downloads 80,000.

Deposit rate = 273/420 × 100 = 65%.
Cost per download = 6,40,000 / 80,000 = ₹8.00.
If a mandate raises deposits to 390 (92.9%) and downloads to 1,20,000 at the same cost, cost per download falls to 6,40,000/1,20,000 ≈ ₹5.33.

Example 5 - metadata completeness. A collection of 1,200 items with 6 mandatory DC elements (title, creator, date.issued, type, language, identifier.uri). Audit finds 180 items missing dc.language.iso and 45 missing dc.date.issued.

Required element-instances = 1,200 × 6 = 7,200; missing = 180 + 45 = 225; filled = 6,975.
Completeness = 6,975 / 7,200 × 100 = 96.875% ≈ 96.9%.

Feature	DSpace	EPrints	Greenstone
Written in	Java	Perl	C++ and Perl
Database	PostgreSQL	MySQL/MariaDB	Own index (GDBM/SQLite)
Best for	Institutional repositories, ETDs	Article self-archiving, CRIS-like use	Offline/CD-ROM, multilingual collections
Metadata	Qualified Dublin Core, configurable	Dublin Core based, rich fields	DC, extensible
Handles/PID	Handle built in	Internal IDs, DOI plugins	None by default
Indian example	Shodhganga, most IIT/IIM repositories	IISc ePrints	UNESCO-supported local collections

Aspect	Digital library	Institutional repository
Content source	Acquired or digitised from many sources	Produced by the host institution
Primary aim	Access to collections	Visibility and preservation of own output
Access	May be licensed/restricted	Normally open access
Example	NDLI, Europeana	Shodhganga, IISc ePrints

Where it is used

University and institute libraries: running DSpace IRs and uploading theses to Shodhganga - IITs, IIMs, IISc, central/state universities, NITs; posts titled Library Assistant, Professional Assistant, Semi-Professional Assistant, Library Information Officer.
National bodies and consortia: INFLIBNET Centre (Shodhganga, e-ShodhSindhu, VIDWAN, ShodhShuddhi plagiarism check), NDLI at IIT Kharagpur, DELNET, National Mission on Libraries, Raja Rammohun Roy Library Foundation.
Research labs and ministries: CSIR (CSIR Central and lab repositories), ICAR (KRISHI repository), ISRO, DRDO, NIC and PIB archives, Parliament and state assembly digital archives.
Cultural and heritage digitisation: National Archives of India, IGNCA, Sahapedia, Digital Library of India legacy collections, state manuscript missions, Sugamya Pustakalaya for accessible formats.
Corporate, publishing and vendor roles: metadata and digitisation services (Informatics, Lucidea/Softlink-type vendors, scanning service providers), knowledge managers in IT and pharma, e-content and LMS teams, repository/OJS hosting support engineers.
Skills sought: DSpace/Koha installation on Linux, Dublin Core and MARC mapping, OAI-PMH configuration, batch import, OCR and quality control, Handle/DOI registration, usage reporting.

Interview questions

Q: How is an institutional repository different from a digital library? A: An IR holds the intellectual output of one institution and is normally openly accessible and mandated; a digital library may collect content from many sources, including licensed material, and exists mainly to provide access.
Q: Name the DSpace content hierarchy. A: Community → Sub-community → Collection → Item → Bundle → Bitstream, where the item carries the metadata and the bitstream is the actual file.
Q: What is OAI-PMH and who are its two actors? A: A simple HTTP/XML protocol for harvesting metadata, with six verbs; the repository is the data provider and the harvester, such as NDLI or BASE, is the service provider.
Q: How many Dublin Core elements are there and are any compulsory? A: Fifteen in the unqualified set, and all are optional and repeatable - though a repository usually makes title, creator, date and type mandatory locally.
Q: Why use a Handle or DOI instead of a URL? A: A persistent identifier is registered once and resolved through a resolver service, so citations keep working even if the server, domain or directory structure changes.
Q: Which version of a journal article can usually be deposited? A: Normally the author accepted manuscript (postprint), subject to the publisher's embargo; the publisher's PDF is usually not permitted - check SHERPA RoMEO or the agreement.
Q: Explain green and gold open access. A: Green is self-archiving an acceptable version in a repository at no cost to the author; gold is publishing in an open-access journal, which may charge an article processing charge.
Q: What is the OAIS model and its three information packages? A: ISO 14721, the reference model for an open archival information system; content moves in as a SIP, is stored as an AIP and is delivered as a DIP.
Q: Difference between migration and emulation? A: Migration converts the file to a current format and risks loss of fidelity; emulation reproduces the original software or hardware environment so the original file can still be rendered.
Q: What does an author actually sign when depositing? A: A non-exclusive distribution licence allowing the repository to store, reformat and distribute the work; copyright stays with the author or the publisher as applicable.
Q: Faculty are not depositing. What would you do? A: Get a formal institutional mandate, offer mediated deposit where library staff upload on the author's behalf, show download and citation statistics to authors, and link deposit to annual appraisal and funder compliance.
Q: Why keep an uncompressed TIFF when the PDF looks fine? A: The TIFF is the archival master retained for future migration and re-derivation; the PDF is a lossy access copy whose quality cannot be recovered later.
Q: What is the difference between withdrawing and deleting an item in DSpace? A: Withdrawal hides the item but keeps a tombstone and the Handle so the citation resolves; deletion (expunge) removes it entirely and breaks existing links.

Common mistakes

Treating a folder of scanned PDFs on a shared drive as a digital library - without metadata, persistent identifiers, access policy and preservation it is only storage.
Writing that all 15 Dublin Core elements are mandatory, or confusing Dublin Core with MARC 21 fields.
Saying OAI-PMH transfers full-text files; it harvests only metadata, and the harvester links back to the original object.
Confusing green with gold open access, or assuming all open access involves paying an APC.
Uploading the publisher's version of record without checking the publisher's self-archiving policy and embargo.
Treating digitisation as the whole of preservation - without checksums, fixity checks, format migration and off-site copies, the files are still at risk.
Mixing up Shodhganga (full theses), Shodhgangotri (synopses), e-ShodhSindhu (subscribed e-resources) and NDLI (aggregated learning resources) in exam answers.
Arithmetic slips in storage sums - using 1,000 instead of 1,024 inconsistently, or forgetting that colour scanning triples a greyscale figure.