Although none of them has survived, catalogues like the lDan kar ma continued to be made, and it was only a matter of time before one of them came to be regarded as definitive, that is, moved from being descriptive—a simple inventory of the holdings of a particular monastery or palace library—to being prescriptive. We can say, in fact, that the formation of the Tibetan canon, or at the very least its shape, can be traced back to the work of cataloguers grappling with the task of imposing some kind of order on the sheer mass of Buddhist literature available to them. When that endeavor was combined with the editorial response provoked by the huge number of copies of individual texts in circulation, each carrying its own peculiar readings, the canon as we know it today was born. It is, however, also likely that the Tibetans were inspired by the Chinese example to attempt a definitive edition of their sacred texts. At any rate we know that at the beginning of the fourteenth century a decisive step was taken at the bKa' gdams pa monastery of sNar thang in gTsang near gZhis ka rtse. An account of this is found in the Deb ther sngon po ("The Blue Annals"), written by gZhon nu dpal (1392-1481) in 1476-1478, less than two hundred years after the event. In his sketch of the sNar thang scholar bCom ldan rigs (or rig) pa'i ral gri, gZhon nu dpal tells us (DTNP: 410-412) that his accomplishments were such that:

[page 75] ...he had many pupils who were fine scholars, and it is said that two thirds of the canon specialists (piṭakadhara, sde snod 'dzin pa) gathered at sNar thang. The great scholar 'Jam pa'i dbyangs was also one of his pupils, but because he once dressed up as a demon and menaced his teacher in the sacred courtyard (?),8 he was severely reprimanded and no longer allowed to stay with him. Having as a result taken up residence at Sa skya, he received an invitation from the Mongols and became the court chaplain of Buyantu Khan,9 where he composed a ṭīkā on the Pramāṇavārttika with a summary appended. No matter how many times he sent gifts to bCom ldan through the imperial messengers, the latter displayed no pleasure at all. Finally he sent him a small chest full of ink, with which he was very pleased. bCom ldan also composed sixteen volumes of treatises. The great scholar known as dBus pa Blo gsal was also a pupil of bCom ral and the Reverend 'Jam dbyangs. bCom ral verified the number of sections, the colophons and so on of the sacred word (bka') of the Sugata and also classified the treatises (bstan bcos) and then wrote the bsTan pa rgyas pa, a treatise which puts them together in their various categories.10 Later, the Reverend 'Jam dbyangs sent copious quantities of materials. In accordance with his request to dBus pa Blo gsal and others that they make copies of all the sacred word and the treatises in translation (bka' dang bstan bcos 'gyur ro cog) and keep them at sNar thang Monastery, dBus pa Blo gsal Byang chub ye shes, the translator bSod nams 'od zer and rGyang ro Byang chub went to great pains to find original exemplars (phyi mo) of the sacred word in translation (bka' 'gyur) and of the treatises in translation (bstan 'gyur)11 and make good copies of them, after which they were kept in the monastery known as 'Jam lha khang. From these, many copies spread to other places: in Upper Tibet they spread to such places as Grom pa Sa skya and Khab Gung thang, while in Lower Tibet too three copies went also to 'Tshal Gung thang, and three copies to sTag lung and its environs.12 Bringing the bsTan 'gyur from sNar thang, Bu ston Rin po che13 removed the duplicates, since the sNar thang one, being the very first, was a collection of whatever exemplars were to be had,14 arranged in proper order what had not been in any order, and added over a thousand new religious texts, after which it was kept in the monastery of Zha lu. Taking that as his exemplar the teacher Nam mkha' rgyal mtshan15 made a copy at gZhis kha Rin spungs, which was kept in the Dharma college of rTses thang.16 This supplied the exemplar for those kept at Gong dkar and gDan sa Thel.17 All the innumerable copies produced thereafter—the separate copies which Khams pas made and took to Khams, the copies which[page 76] were made using these as exemplars in Khams itself, the copy made by the Chos rje mThong ba don ldan,18 the copy made in dBus by the Du dben sha ba,19 the copy made from precious substances at 'Tshur phur by the Chos rje Rang byung ba,20 the copy made at Byams pa gling by Yar rgyab dPon chen dGe bsnyen pa,21 the copy in 180 volumes made by gZi Kun spangs pa,22 right down to when sTag rtse ba,23 built a fine monastery and made a copy which includes many exemplars obtained later, in addition to the former bKa' 'gyur and bsTan 'gyur—these also came into existence thanks to the Reverend 'Jam pa'i dbyangs, the pupil of bCom ldan rigs pa'i ral gri, and these two in the final analysis owed it all to the grace of rNgog lo tsā ba, who owed it to the grace of the scholars of Kashmir, and ultimately to the grace of the Buddhas.24

This account is worth quoting in full for a number of reasons, not least because of the light it throws on the motivation for the compilation of the sNar thang "edition."25 As gZhon nu dpal tells the story, this particular collection was made only in response to the request, and with the substantial material assistance of 'Jam pa'i dbyangs, whose contribution was therefore pivotal.26 There is thus a strong suggestion of Chinese influence, since working at the Yuan court 'Jam pa'i dbyangs would no doubt have been influenced by his Mongol patrons' sense of the importance of previous editions of the Chinese Buddhist canon produced under imperial sponsorship, and by their desire to add lustre to this tradition.27 We know too that sNar thang, like Sa skya, had very close connections with the Mongol rulers of China.28 Thus the initial compilation of the Tibetan canon may be seen as a distant echo of that well-known process by which the Chinese culturally subverted foreigners who had conquered them by force of arms, and its political implications merit attention. But what is equally interesting about gZhon nu dpal's account, on a more personal and human level, is the implied additional motivation for 'Jam pa'i dbyangs's initiative. Practical jokes often backfire on their perpetrators, but this hair-raising schoolboy prank had spectacular consequences. bCom ral must have given his hapless student such a severe dressing-down that the poor man smarted from it for the rest of his life, engaging in pathetically extravagant attempts to win back his teacher's favor. In this way a brief moment of boyish fun can be seen as the starting point for centuries of sober scholarly activity.29

[page 77] gZhon nu dpal also paints a vivid picture of the veritable explosion of bKa' 'gyur and bsTan 'gyur copies from sNar thang in the fourteenth and fifteenth centuries, as Tibet was swept by what we might call a "bKa' 'gyur craze." But he tells us little about the corresponding flow of copies towards that center which preceded the compilation of the "edition." Fortunately the details of that are preserved in the section colophons to the Tshal pa bKa' 'gyur (see below) which have been carried over into the Li thang and other editions.30 These are documents of capital importance. From them we learn that the Sūtra section of the Old sNar thang was based on over a dozen different sūtra collections (mdo mangs) from the libraries of Sa skya, gTsang Chu mig ring mo, Shog chung, sPun gsum, Zha lu, and other monasteries, together of course with those held at sNar thang itself. The Tantra section was based on at least five exemplars from Sa skya, Thar pa gling, and sPun gsum, and was arranged according to catalogues compiled by Grags pa rgyal mtshan (1147-1216), 'Phags pa (1235-1280), Rigs pa'i ral gri and others. The Vinaya was based on a manuscript edition compiled by mChims ston Nam mkha' grags pa, abbot of sNar thang from 1250 to 1289, compared against the Vinaya texts of Rung klung shod grog Monastery and others. Nam mkha' grags pa's text had itself been based on the edition made at La stod 'Ol rgod Monastery by Dharma seng ge using copies obtained from bSam yas mChims phu and other monasteries in dBus and gTsang with the help of the teacher and Vinaya specialist (vinayadhara, 'dul ba 'dzin pa) Zhing mo che ba Byang chub seng ge during the time of the Vinaya specialist of rGya, dBang phyug tshul khrims 'bar (1047-1131). We see then from these colophons that the sNar thang "edition" was the result of the gathering in of texts from various monastic libraries in gTsang and surrounding areas,31 and at the same time the culmination of several centuries of collecting and cataloguing activity at a number of centers, including Sa skya.

On some points, however, the testimony of these sources is frustratingly vague. In particular, we do not know whether the scholars of sNar thang took the original manuscripts of all these collections back to sNar thang, or returned home with complete copies of them, or, working from one of their catalogues, copied only those individual works not already in their possession. The DTNP gives the impression that bCom ral and his disciples had first worked on the translations of sūtras and śāstras held at sNar thang,[page 78] and had written several catalogues, before the collection process began, so it is quite possible that they collected selectively and to order. With two or more teams working concurrently, such a procedure is bound to have produced multiple copies of some texts. The DTNP enumerates three significant features of the copy of the sNar thang bsTan 'gyur which Bu ston worked on: it was incomplete, it was not in order (at least not to Bu ston's satisfaction), and it contained duplicates. What was true of the bsTan 'gyur is equally likely to have been true of the bKa' 'gyur; it is quite possible that it too contained multiple copies of texts, either different translations of the same text,32 or different recensions of the same translation. This means that both the sNar thang bKa' 'gyur and bsTan 'gyur may simply have been better arranged collections of high-quality copies, rather than editions in our sense of the word, and that therefore they still required editorial attention.

It is my belief that the initial collection of copies which took place at sNar thang was soon followed by a second phase in the production of the bKa' 'gyur and bsTan 'gyur collections that we know today, and that this phase was carried through in at least two different places.33 One of these places was Tshal (or 'Tshal) Gung thang Monastery in dBus, where a new edition of the bKa' 'gyur was produced during the years 1347-1351 under the sponsorship of the local ruler, Tshal pa Kun dga' rdo rje, also known as dGe ba'i blo gros (1309-1364). Since the original section colophons of this edition have survived we know a great deal about it. We know, for example, that the texts of the sNar thang "edition," of which three copies were employed, were substantially revised (using the Mahāvyutpatti and other such works to standardize the wording), and that their order was also rearranged, with a number of titles being deleted from the bKa' 'gyur because they were deemed to belong to the bsTan 'gyur.34 A three-volume set of tantric texts translated during the early period (rNying rgyud) was also added. The result is known as the Tshal pa edition. The second center of editorial activity was Zha lu in gTsang. We cannot yet be sure that Bu ston carried out a complete revision of the bKa' 'gyur (as well as the bsTan 'gyur) at Zha lu, but there are indications that he did edit both collections, even though gZhon nu dpal mentions only his bsTan 'gyur edition.35 However, we have firm evidence that Bu ston worked on substantial portions of the bKa' 'gyur, and that this editorial work was continued by his successors[page 79] at Zha lu and rGyal rtse (see Harrison, 1994). This aspect of the history of the bKa' 'gyur is rather problematic, but there are good reasons for believing that at some time in the first half of the fourteenth century a Zha lu bKa' 'gyur also came into existence, and that this edition may have been closer to the Old sNar thang than its Tshal pa counterpart, at least in terms of organization. I shall call this edition the *Zha lu ma, using an asterisk to mark its hypothetical status.36 Both the Tshal pa and Zha lu editions may well have been based on the same raw materials, but especially in the matter of the deletion of duplicates, different decisions could easily have been arrived at, which would account for much that was to follow.

From this point on our discussion concerns the bKa' 'gyur rather than the bsTan 'gyur, although we should note that the evolution of a basically bipartite canon seems to be a peculiarly Tibetan innovation.37 (This scheme was also adopted by the Bon pos, whose own canon, divided into bKa' 'gyur and brTen 'gyur, appears to have been systematized in the late fourteenth and early fifteenth centuries (see Kvaerne: 38-39) in imitation of the Buddhist model.) The bKa' 'gyur section of the Tibetan Buddhist canon has in its turn three major divisions: 'Dul ba (Vinaya), mDo (Sūtra) and rGyud (Tantra), thus making it a kind of tripiṭaka in itself, arranged according to the three "vehicles" or three different levels of religious avocation (sdom gsum): 'Dul ba for "Hīnayāna" (i.e., Mainstream Buddhism), mDo for Mahāyāna, and rGyud for Vajrayāna. To put it like this, however, oversimplifies the picture, because although the 'Dul ba section is comparatively clear-cut, the other two are not. Thus the mDo section, broadly conceived, is broken down into Sher phyin (Prajñāpāramitā texts), Phal chen (the Avataṃsaka Sūtra), dKon brtsegs (Ratnakūṭa texts), Myang 'das (Mahāparinirvāṇa Sūtra) and mDo sna tshogs or mDo mang(s) (miscellaneous sūtras) sections, while the rGyud texts are divided, following the classification scheme promoted by Bu ston and others, into four main classes, supplemented in some editions by the rNying rgyud ("Old Tantras") and gZungs 'dus ("Dhāraṇī collection") sections.38 These sections and subsections do not appear in the same order in all editions, partly because of different schemes for classifying the sequence of the Buddha's teachings (see, e.g., Skorupski: xiv-xvii). The same holds true for the order of the individual texts within the sections, especially in the rGyud, where[page 80] the placing of particular tantric cycles often indicates sectarian preferences.39 The study of the complicated issues involved here is one way of determining the affiliations of the editions. However, repeated re-arrangements of the bKa' 'gyur make it difficult for us to determine the original order of the Tshal pa and *Zha lu ma editions on the basis of their descendants.