If Liu Zhiji 劉知幾 (661-721) was the first scholar who established theories of history writing in his book Shitong 史通, Zheng Qiao wrote his masterly chapter Jiaochou lüe to explain the principles of compiling a bibliography. Zheng Qiao defends his independent system of categorization with the cogent argument that clear categories were important for the survival of texts. This could, for instance, be seen from the large amount of religious Daoist and of divination texts, two genres for which distinct categories had been established in catalogues, while of divinatory texts operating with hexagrams (like the Classic Yijing 易經) or legalist texts, only a small number was still available, because bibliographers did not much care about subclasses in these genres.
Zheng even compares the management of a text corpus with commanding a large army (lei [predicative use] shu you chi jun ye 類書猶持軍也): clear categorization rules help to master largest masses. A considerable part of his introduction into the discipline of cataloguing is dedicated to the question of lost texts. He mentions examples of the past, where specialized catalogues were compiled that listed lost texts, like Queshumu 闕書目 "Catalogue of missing books" from the Wei 曹魏 (220-265) and Soufang tushu mu 搜訪圖書目 "Catalogue of wanted books and maps/illustrations" from the Tang period 唐 (618-907).
Zheng Qiao did not only advocate the registering of texts that were lost—because there was the possibility that they came out one day—but also of contemporary texts. There would be plenty of techniques to reconstruct or retrieve lost material. He lists a large number of texts that were quoted directly or indirectly in other books, and could thus easily be reconstructed.
Examples for this were Li Zhou's 李舟 (fl. 759) dictionary Qieyun 切韻 that was basically just a phonetic arrangement of the character dictionary Shuowen jiezi 說文解字 (ordered according to radicals), or the astronomical texts Kaiyuan zhanjing 開元占經 and Xiangying yanlu 象應驗錄 that were implicitly quoted in the books Gujin tongzhan jian 古今通占鑑 and Qianxiang xinshu 乾象新書, or the pharmacopoeia Zhenglei bencao 證類本草 which quoted the integral texts of the books Bencao shiyi 本草拾遺, Shanfan bencao 刪繁本草, Yaodui 藥對, Nanhai yaopu 南海藥譜, Yaolin 藥林, Yaolun 藥論 and Yaoji 藥忌. Books (or texts) of this kind were, he maintains, "lost but by name, not in reality" (shu you ming wang, shi bu wang 書有名亡實不亡).
Concerning jurisdictional texts, older ones could be reconstructed on the basis of newer canons, just because the latter were enlarged and "modernized" versions of the former. Thus, the ancient law texts of the early Han period 漢 (206 BCE-220 CE) like Xiao He's 蕭何 (d. 193 BCE) Lüling 律令 and Zhang Cang's 張蒼 (d. 152 BCE) Zhangcheng 章程 - if these are to be seen as titles of texts-, were included in jurisdictional books of later ages (like Hanchao boyi or Han mingchen zouyi 漢名臣奏議). When looking at the physical size of books, it becomes clear that older texts are included in newer ones. The collected writings of Tao Yuanming 陶淵明 (c. 365-427), for instance, Tao Qian ji 陶潛集, had a size of 5 juan in Liang-period 梁 (502-557) catalogues, 9 in the Suishu 隋書 catalogue Jingji zhi 經籍志, and 20 in the Tangshu 唐書 bibliographies.
An important way for the retrieval of lost texts was to dispatch officials to inquire of private book collectors or to invite private owners to present their texts to the authorities. Zheng Qiao furthermore stresses the importance of employing experienced bibliographers and to keep them in office as long as possible. With their expertise long-serving librarians would be an immense help for the location of lost or unknown texts. They would first consult experts and search specialized catalogues. For books on the calendar, for example, they would first ask the Director of the Imperial Observatory (lingtailang 靈臺郎), for books on music the *Grand Director of Music (taichang yuegong 太常樂工). Dictionaries and primers were quite likely to be found in Buddhist catalogues, and books on hexagrams in those of diviners or Daoists.
Geography was another important criterion. Books on Wang Shenzhi 王審知 (862-925), for instance, were quite likely to be found in libraries in Fujian, because he had been the founder of the "empire" of Min 閩 (909-945, in today's Fujian), one of the Ten States 十國 (902~979). Texts might also be found in the possession of descendants or family members of the author. Official texts like such on rites and ceremonies, sacrifices, or such on verdicts could be found among the libraries and archives of authorities. In some cases, a diligent "book hunter" would ask his way from person to person, and trace back the history of a text from owner to owner.
The apparent loss of individual texts of whole groups of texts could sometimes be traced back to incomplete bibliographies. Zheng complains the missing of meteorological texts (fengyun qihou 風雲氣候) in the Tangshu catalogues and that of texts on the calendar (riyue zhi shu 日月之書) in the early Song-period 宋 (960-1279) bibliography Chongwen zongmu 崇文總目.
Another source of errors was wrong categorization. The military book Weiliaozi 尉繚子, for instance, was defined as a syncretist text (zajia 雜家) in the Hanshu 漢書 bibliography Yiwen zhi 藝文志 and is therefore not found among the military masters (bingjia 兵家). This error was repeated by later catalogues, and only the Chongwen zongmu puts this text in its proper place. The Hanshu bibliography also classified the Simafa 司馬法 "Methods of the Minister of War" as a ritual book, and the Taigong bingfa 太公兵法 "The Grand Duke's military methods" (i.e. Liutao 六韜) as a Daoist text.
Some further examples show that bibliographers often did not take time to inform themselves of the content of books. For this reason, the compilers of the Chongwen zongmu believed the critical dictionary Kanmiu zhengsu 刊謬正俗 (Kuangmiu zhengsu 匡謬正俗) to be a book on the Lunyu 論語 "Confucian Analects" because the first lemma was "Lunyu". The already mentioned books Hanchao boyi and Nantai zoushi, in fact texts on jurisdictional matters, were by the compilers of the Tangshu bibliographies believed to be texts about ancient administrative matters (gushi 故事), and some books on outer alchemy (waidan 外丹, Zheng says luhuo zhi shu 爐火之書 "books on the furnace"), Zhouyi cantong qi 周易參同契 and Zhouyi wuxiang lei 周易五相類, believed to be texts on the Yijing (also called Zhouyi 周易), just because the name of the latter was part of the titles.
In fifteen paragraphs, Zheng Qiao points at a number of inconsistencies and errors in earlier bibliographies. His conclusion is that the Suishu bibliography was quite reliable, while the Chongwen zongmu was less trustworthy because it had been compiled by many persons who did not successfully cooperate. It was important, he explains, that real experts were entrusted with the compilation of the parts of a bibliography they were professionals for. Liu Xiang 劉向 (79-8 or 77-6 BCE) and his son Liu Xin 劉歆 (d. 23 CE), the authors of the earliest imperial bibliography, Qilüe 七略, and its surviving part, the Hanshu catalogue Yiwen zhi, had only been general librarians, and not experts in many fields, with the consequence that only military texts and divination texts were clearly divided into subcategories and special fields. These parts had been compiled by the experts Ren Hong 任宏, Yin Xian 尹咸 and Li Zhuguo 李柱國. At least, the texts of other categories were arranged according to certain patters, yet without making this structure directly clear to the reader or user.
Zheng recommends a stringent and precise categorization on several levels and encourages bibliographers to look into the facts, for instance, that the corpus of texts on funeral clothing (sangfu 喪服) being a specialized discipline of its own had to be separated from commentaries on more general and comprehensive books on rituals and ceremonies like the Yili 儀禮. Probably even more important is Zheng's recognition of religious Daoist texts (daoshu 道書) as a category of its own, in contrast to the Daoist philosophers (daojia 道家), and of jurisdictional texts (xingfa 刑法) as definitely different from the writings of old legalist philosophers (fajia 法家). He also attacks some older bibliographers (without mentioning names) who integrated Buddhist and Daoist texts in one category and so neglected the fierce controversies of the two religions/philosophies.
Another important issue discussed by Zheng Qiao is that of descriptive catalogues. He is of the opinion that not all books need to be discussed in detail, but only those where remarks contribute to a better understanding of history and content. He furthermore criticizes the incoherent "syntax" of many bibliographies: The basic criterion for a catalogue should be the book, and not the author, because catalogues were arranged according to the literary genre. It was thus not correct to enumerate in one line two or three books written by the same author, as often seen. One text might deal with a different topic than another, and while one may be a specialized book by this author, another might be a collection of poems or letters.
Linghu Chu's 令狐楚 (c. 800) collected writings (Linghu Chu ji 令狐楚集) were therefore to be included in the collections of individual writings (bieji 別集), and his memorials to the throne (Linghu Chu zoubiao 令狐楚奏表) in the appropriate section (zouji 奏集). Yang Xiong 揚雄 (53 BCE-18 CE) was basically a
An often-found syntactic error was that the author was mentioned first, and then the book, the two elements combined by the word zuo 作 "wrote" (like Guan Chen zuo »Guan Lu zhuan« 管辰作管輅傳 "Biography of Guan Lu, written by Guan Chen", in modern setting 管辰作《管輅傳》). The word zuo is often left out, so that the reader might misread the title as "Biography of Guan Chen and Guan Lu" (Guan Chen Guan Lu zhuan 管辰管輅傳, in modern setting 《管辰、管輅傳》).
Although Zheng Qiao does not explicitly say how the syntax should be, it is clear that the book title should be written first, then probably the length of the text (in juan), and the author last, in the shape »Guan Lu zhuan«, yi juan, Guan Chen zhuan 《管輅傳》一卷，管辰撰. This structure would avoid any misunderstandings, even without the modern title (《》) and punctuation marks.