Introduction: During the last couple of years the library community has developed a number of comprehensive metadata standardization projects inspired by the idea of linked data, such as the BIBFRAME model. Linked data is a set of best practice principles of publishing and exposing data on the Web utilizing a graph based data model powered with semantics and cross-domain relationships. In the light of traditional metadata practices of libraries the best practices of linked data imply a restructuring process from a collection of semi-structured bibliographic records to a semantic graph of unambiguously defined entities. A successful interlinking of entities in this graph to entities in external data sets requires a minimum level of semantic interoperability. Method: The examination is carried out through a review of the relevant research within the field and of the essential documents that describe the key concepts. Analysis: A high level examination of the concepts of the semantic Web and linked data is provided with a particular focus on the challenges they entail for libraries and their meta-data practices in the perspective of the extensive restructuring process that has already started. Conclusion: We demonstrate that a set of heterogeneity conflicts, threatening the level of semantic interoperability, can be associated with various phases of this restructuring process from analysis and modelling to conversion and external interlinking. It also claims that these conflicts and their potential solutions are mutually dependent across the phases. [This paper was published as part of: Proceedings of the Eighth International Conference on Conceptions of Library and Information Science, Copenhagen, Denmark, 19-22 August, 2013.]