Technical
Site

Related Links

Approved by the TSC on 16 February 2016.

Technical Strategy

This draft is in response to the recommendation made at the FHISO Board meeting on 2015-11-27 that

the TSC will draft and release a short high-level strategy paper to communicate the TSC's vision for FHISO.

This document gives an overview of how the TSC envisages technical work progressing, both in the short-term, and how that short-term work fits into the longer-term vision. It is not intended to be binding, and the TSC may publish future versions of this roadmap that differ from this one, perhaps substantially.

Background

According to by-law 3.3,

The scope of FHISO's work encompasses all aspects of technical standards and guidelines used to manage information in genealogy and family history.

This couild be paraphrased as "whatever genealogical matters our members are interested in". It is intended to give FHISO the flexibility to carry out work on any aspect of genealogy in which there is sufficient interest; it allows the production of multiple standards, and also the generation of less formal guidelines.

FHISO grew out of the BetterGEDCOM project, a project whose goal was

to develop a standard for genealogy data archiving and transfer that would be accepted internationally.

A major new standard of that nature, one more modern than GEDCOM and that removes many of GEDCOM's limitations, remains a major long-term objective. But the TSC believe it is unrealistic to produce an all-encompassing standard with the resources at FHISO's disposal, and were such a standard to be produced, it runs the risk of being ignored, much as other attempts at a new standard such as the GenTech Data Model have been.

Instead, the TSC plan to develop a series of standards each covering a narrow area. Over time, these will fit together to form FHISO's new data model, but at first they are likely to be used in existing data models, in particular GEDCOM and GEDCOM X, which the FHISO Board have identified as the two dominant non-proprietary data models at present.

First steps

The FHISO Board, on the TSC's recommendation, have decided that FHISO should initially

work on a focused standard for a citation elements vocabulary and, if resources permit, a vocabulary for describing source derivation.

These specific areas have been selected because they are relatively small and self-contained vocabularies that we believe can be developed relatively quickly and can slot into existing data models. Our contacts in the industry suggest that a citation elements vocabulary would fill an important niche in today's standards, while the source derivation vocabulary fills a small but significant gap identified by our exploratory work into sources & citations. These two vocabularies are not intended to be all of FHISO's work on sources and citations — just the first priorities.

Vocabularies

As our two first projects involve designing vocabularies, at the Board's request the TSC have drafted a policy on the preferred nature of vocabularies. A general policy on vocabularies is desirable so that all FHISO vocabularies share a common style, and that this is not simply based on the requirements of the first vocabulary to be designed. This policy is based on ideas discussed by the Sources & Citations Explorary Group as well as industry best practice.

The TSC welcome feedback on this draft policy, and suggest that such discussion is best directed to the tsc-public mailing list.

Citation Elements Vocabulary

For the purpose of this standard-development activity, a citation element is a piece of data about a source that might reasonably be included in a formatted citation. Examples might include the author and title of a book, page numbers, the name and address of an archive, the date an online source was accessed, and the reference number of a manuscript. Normally they are language-neutral, for example Settipani's well-known book is called Les Ancêtres de Charlemagne regardless of whether one is writing in English or French; but not always: an English work would rarely cite the Confucius Genealogy Compilation Committee as 《孔子世家谱续修工作协会》.

Citation elements are not intended to provide an exhaustive description of sources, and it is anticipated that applications may store information about sources beyond a simple collection of citation elements. Discursive notes, although sometimes included in a citation, are not considered to be citation elements for our purpose; nor is information about if and how one source derives from another.

In producing this standard vocabulary, we will need to seek out examples of citations in different styles and languages. Most of all, we need citations to a wide range of sources types: not just the usual censuses and parish registers, but also books, journals and manuscripts, paintings and photographs, monuments, statues and stained-glass windows, websites and television programmes. How are they cited? And what elements do these citations contain?

Identifying the citations elements required is just part of the problem. Terms needs careful definition to remove ambiguity over their use. The standard should identify what range of values an element is expected to have: perhaps the value will be a term from another vocabulary; or perhaps it will be a boolean, an integer, a date, or an unstructured piece of text. If the element is a date, what requirements are placed on the date format used? Designing a date format is not part of this project, and it is possible that this vocabulary will be used in multiple data models with distinct date types; but any essential requirements for the particular use should be noted.

The aim is that applications should be able to produce properly-formed citations in the vendor's preferred style using the data found in the citation elements. Ideally this should be possible using standard templating technology such as XSLT. It is not FHISO's job to produce such a formatter, but as the standard develops, we should be considering whether data in the citation elements can be converted algorithmically to forms used in real-world citations. For example, some citation styles for books put the author's surname first, followed by first names or initials; other styles put the name in its conventional order. How should this be supported?

Full systems for representing names and addresses are outwith the scope of this project. But it may be that a lightweight microformat may be appropriate, such as GEDCOM's use of / to delimit surnames. In other cases it may be useful to document a convention on a field's use, for example that addresses components are separated by commas and written from the smallest to the largest unit.

The conceptual requirement for algorithmic formatting of citations with standard templating technology places more tangible restrictions on the implementation. Citation elements may not contain sub-elements, though they may still have structure in the form of typed data or through a microformat; nor may they be pointers or references to other structures. Elements may be multi-valued where appropriate: that is, a source may have multiple citation elements of the same type. Meaning should only be associated with the order of values if it is tolerable for the information encoded by the order to be lost. For example it might be acceptable to use the order of several 'author' elements to denote the order in which the authors are conventionally listed. No meaning should be associated with the order of different types of elements.

Source Derivation Vocabulary

Our Sources & Citations Exploratory Group (S&CEG) spent a considerable time looking at layered citations — citations of sources which are derivatives of another source, for example a transcription made from a microfilm of a parish register. Such derivation information is beyond the initial scope of our citation element vocabulary, but we do plan to address it separately.

The S&CEG identified several forms of derivation with different degrees of faithfulness, ranging from high-quality photographic reproductions through to translated abstracts of sources, and secondary sources that paraphrase the original. A source derivation vocabulary will describe these various types of derivation.

There are likely to be situations when an application needs to flatten a sequence of derivations to a single one. In the example of a transcript of a microfilm of a register, house style might dictate that only source consulted and its ultimate source should be mentioned: that the transcript derives from the register. A rule will be needed that says how to describe the combination of several derivations; if the derivation types can be ordered by faithfulness, this may be as simple as selecting the least faithful step.

Follow-up citations work

When these two vocabularies are complete, there are several options for related follow-up work.

An obvious and fairly straightforward task is to define extensions to GEDCOM and GEDCOM X to allow these vocabularies to be used. GEDCOM entirely lacks a way of recording source derivation, while GEDCOM X allows a link but lacks any means of describing the derivation; both languages would benefit from the addition of such a facility based on our source derivation vocabulary. Neither GEDCOM nor GEDCOM X have good facilities for recording everything about sources that is found in real citations; in particular neither has adequate facility for recording page numbers, it being entirely absent from GEDCOM X, while the format of GEDCOM's PAGE tag's values is poorly standardised.

Another facet of FHISO's early citation work could be to look at which citation elements make sense for which types of source. Clearly a 'journal name' field makes little sense for a tombstone: should it be prohibited? Indeed, a vocabulary of source types might be a useful further development. GEDCOM's list of source types lacks many obvious types, such as journals, paintings, websites, databases and CD-ROMs, which GEDCOM X only offers a coarser-grained division into physical and digital artefacts.

The system of citation elements being developed does not cope with layered citations, but can be readily extended to include them. Another possible extension is to document properly the model for turning multi-layer citation elements into a formatted piece of text. This might writing include a basic, proof of concept templating engine.