Publication Group

unnumbered indicates why a publication group has no ISBN, where the reason is known:

• predates — issued before ISBNs were in use
• club — a book club numbered it its own way, with a CN or with nothing
• publisher — the publisher did not number it, as the Folio Society does not

Publisher Editions

A contains given here overrides the one on the work, for an edition whose contents differ from the collection’s usual ones.

Impression Groups

A set names a boxed or slipcased set of several volumes, which is not itself a book and so has no publication group. Each volume in the set carries the same set value, and the set is described separately: its publisher, year, how many volumes it holds, and whether its volumes were also sold separately, issued only as part of the set, or some of each. Whether the volumes were issued only as part of the set is recorded by the set's composition, and says nothing about where the ISBNs sit: a volume carries one if it was assigned one, and the set carries one if it was numbered as a product.

A set may be recorded before any of that is known, since sets turn up long before they can be worked out. Only its name is required, and its volumes need not be in the database at all. What a set may not do is hold some of its volumes and not say how many it has, so a partly entered set is reported rather than passing as complete.

Questions

questions records what we do not yet know about the books: things that would have to be looked up, asked about, or checked against a copy in hand. It may be given on a work, a publication group, a publisher edition or an impression group, and sits on whatever it is a question about, so that it is in front of whoever next edits that work. Answering one is a single edit: delete the question and correct the data.

It is not for uncertainty in a value that has been recorded. That is what first_known_impression, first_possible_impression, last_known_impression and last_possible_impression are for.

Contains

A collection gathers other writings, and a Contains item names one of them. It has exactly one of:

• work (string, the ref of a work this database holds)
• title (string, for a writing the database does not hold as a work of its own)

It sits on the Work, and again on a Publisher Edition whose contents differ: Tree and Leaf gained Mythopoeia in 1992, and Tales from the Perilous Realm gained Roverandom in 2008.

Cover Design

relation is one of illustration, based-on-design, based-on-artwork, slipcase-illustration, illustrated or from-source. All but the last name at least one artist; from-source gives a source instead, for a cover taken from a film or a television series.

Impressions

Only the number is required. A printer or a retail price can be read off a copy whose year cannot, and most printing records give one without the other.

printer sits here rather than on the impression group because it changes within a run: one edition under one ISBN moves from Clays in Suffolk to L.E.G.O. in Vicenza between one impression and the next. printed_in is given only where a firm printed somewhere other than its usual country, as GPS Group does in Bosnia and RR Donnelley in China.

boards is the colour of a hardback's covers and dust_jacket the finish of its jacket, and they are typed differently on purpose. A finish is a closed set — matte, glossy, satin — and the validator holds it to those. A colour is not: the source says "Saturated Grass green", "Dark 'military' Green", "Reddish Brown in colour (a little lighter than the 1st printing)", and two printings of one ISBN differ only in which green they are. A controlled list would erase exactly the fact worth recording, so boards stays as description, with the source's own words kept in a comment beside it.

Both sit on the impression because both change within a run: the boards of the 2013 Farmer Giles are a different brown at the third printing from the first.

A price is a mapping, not a line of text, because a book is priced in several currencies at once and the point of recording it is to be able to ask what it cost:

price:
  GBP: 6.99
  CAD: 9.99

Typefaces

It sits on the impression because a reset changes the type under one ISBN and one edition label, and is often the only sign in the book that the pages beneath have moved.

The names a source gives conflate three things: PostScript Monotype Plantin is a face, a foundry and the system it was cut for; Plantin Light is a face and a weight; Sabon is a face alone. Each such name is an entry, and the entry decomposes it — family for the face, which is what "everything set in Plantin" asks about, with foundry and technology where the name gives them. A family named by no impression, as garamond is, exists to be pointed at.

Printers

A country is where the firm ordinarily prints, and is left out for one that works in several; the impression then says where with printed_in.

Design Series

A design is a claim about the physical book, so format and traits say what a member has to be, and the validator holds every Impression Group in the series to it, against the Publisher Edition the group sits under. A series bound one way cannot hold a member bound the other; a series of de luxe editions cannot hold a printing whose edition does not say it is one.

traits is a floor rather than an exact match. The 2004 Lord of the Rings is anniversary and illustrated as well as de luxe, and is no less a Quarter-Bound De Luxe for carrying more than the series demands.

Both are optional, and a series declares only what is true of every member. Some can declare neither: the John Howe Gold Runic Border ran across both bindings, and a trait records what distinguishes an edition rather than what it looks like, so a book with no unillustrated counterpart to be told apart from does not carry illustrated however illustrated it is.

Publishers

A label is the publisher as the data spells it, exactly. A Publisher Edition and an Impression Group each name a publisher by that string rather than by key, and the validator checks that every one of them is in this file, so the vocabulary is authoritative without any record having to change.

A label may name two firms, as a title page does: Grafton / HarperCollins, Early English Text Society / Oxford University Press. Such an entry gives names, the keys it names in the order printed, and a relation: imprint, where the second is the house the first belongs to; joint, where two firms published together; or for, where the first printed the book for the second to sell, which is the shape of a book club or a direct seller and reads HarperCollins for The Book People. A compound carries no lineage of its own; that belongs on the entries it names.

Lineage is recorded two ways. imprint_of is structural and true at a moment: Voyager is part of HarperCollins. became is historical: George Allen & Unwin became Unwin Hyman, which became HarperCollins. Neither is transitive in the data, so a view that wants the whole line walks it. Succession is many to one: both Unwin Hyman and Collins became HarperCollins, so the line is a convergence rather than a chain.

Firms that never published Tolkien are here where they are part of that line — Swan Sonnenschein, Bell & Hyman, Harper & Brothers. Nothing marks them as such, because the data already says it: a publisher no printing names is an earlier firm. What the validator does require is that such an entry take part in a lineage, so that an entry nothing names and nothing relates to is reported rather than left to accumulate.

Sets

composition is one of gathered, exclusive or mixed, and records only whether a volume could be bought on its own. Where the ISBNs sit is a separate matter: a volume carries one if it was assigned one, and the set carries one if it was numbered as a product. Both vary independently — the seven-volume Lord of the Rings sets number every volume although none was sold apart from the set.

Artworks

A tg_page is a Tolkien Gateway page name, as on a Work, not a URL: the address is built when it is rendered, so a change in Gateway's URLs is a change in one template.