Giving Depth to TEI-Based Descriptions of Manuscripts: The Golden Gospel of Ham

The work on the Golden Gospel of Dabra Libānos za-Ham in Eritrea, one of the main archives of the medieval history of Ethiopia and Eritrea, was initiated within the frame of the Ethiopian Manuscript Archives project, ANR EthioChristProcess, and Beta maṣāḥǝft. The work aims to describe the codicological disposition of the archival notes (i.e. arrangement of texts within the codex and its by-side folia), to provide their transcription, translation, and indexation. This article will discuss the encoding method for this work in progress, supported by TEI.


Introduction
'The era of heroes like A. Dillmann has gone; not because we are so different or lazy, but because of the general conditions of modern life-thus collective work on well organized data bases might well be the melody of the future', stated Manfred Kropp in 1994, 1 a pioneer of the use of digital methodologies in Ethiopian studies. In the same article Kropp calls for attention to archival notes transmitted in codices or bound to codices. 2 Since 1994 the corpus of known archival notes in Gǝʿǝz has been increasing, the latest in-depth study of the archival notes and its possible function in Ethiopian society was carried out by Anaïs Wion. 3 Further on in his article, Kropp, himself an editor of many archival notes (at least one of which was annotated in TEI), 4 draws attention to a publication by Carlo Conti Rossini of an extraordinary file of royal diplomas found in the 'golden gospel codex' of a richly endowed monastery in [Eritrea] (Dabra Libānos). 5 No translation and only very short and provisional comments are found in  This work was produced thanks to the Beta maṣāḥǝft project (Schriftkultur des christlichen Äthiopiens und Eritreas: eine multimediale Forschungsumgebung) funded within the framework of the Academies' Programme (coordinated by the Union of the German Academies of Sciences and Humanities) under survey of the Akademie der Wissenschaften in Hamburg; as well as thanks to ANR EthioChristProcess project and HornEast project, funded by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement no. 726206). This article reflects the state of the project on the moment of submission (4 March 2021). 1 Kropp 1994, 118, also quoted in Liuzzo 2019, xxii, providing the point of the state of the art since. 2 Kropp 1994, 120. 3 Wion 2019. For the XML encoding of archives, see Wion 2018. 4 Kropp 2005. 5 Conti Rossini 1901. An index of names in the text is provided by Bausi 2007. this publication. In 1975 the French archaeological mission in Ethiopia had the good fortune to find and film the original manuscript to Conti Rossini's work, but again, nothing in detail has been published until now. 6 It should be noted here that an immediate review by Boris Turaev on Conti Rossini's publication went apparently unnoticed by Western historians. Turaev recognized the importance of the archive, arguing with some of its historical conclusions, but, what is important, translated three notes, number 9, 20, 33, according to Conti Rossini. 7 Although Turaev's translations should be recognized as a first attempt to interpret these archival notes in extenso, he disregards all difficulties of the notes by simple omission. 8 Concerning the photographing of the original manuscript on which Conti Rossini worked, this was carried out by the French archaeological mission in Ethiopia: indeed, as Roger Schneider's archive shows, 9 some 20-30 photos of openings and single pages from the estimated 162 folia were taken in 1975. What Kropp ignores is that Alessandro Bausi could visit the monastery and take photos of the original manuscript in 1993 and 1994, amounting to some 32 photos of openings. 10 The photos taken by Schneider and later by Bausi reproduce almost all the archival notes Conti Rossini edited in 1901, which shows how well the traditional guardians of history, the monks of the monastery of Dabra Libānos, have 6 Kropp 1994, 130. 'Golden Gospel codex' according to the terminology employed in Ethiopian and Eritrean studies is used to define a manuscript of the four Gospels containing archival notes (often as an added content) of various historical interest. 7 Conti Rossini 1901, 193, 203-206, 216-218. 8 Turaev 1901 Roger Schneider's archive, i.e. his images and notes, was donated to the Walda Masqal Centre (Addis Abäba, Institute of Ethiopian Studies) by the scholar's family after his death. Marie-Laure Derat carried out the first systematization and description of Roger Schneider's archive (Derat 2011), facilitating the continuation of his various projects, such as the project here. 10 Bausi 1997. Images are available at https://betamasaheft.eu/manuscripts/DabraLibanos HamGG1/viewer. The documentation was acquired by the Missione Italiana in Eritrea (MIE), 1992-1994, funded by Consiglio Nazionale delle Ricerche and Alma Mater Studiorum Università di Bologna, and directed by Irma Taddia. While ignoring the exact conditions in which the French mission was conducted, the scope of the Italian mission was to document monastic places and their libraries. The permission obtained limited them to photos of single, isolated folia. Further to which, there were many practical aspects that impacted the mission's work: limited time, limited amount of film, and, as it was not possible to check the quality of each image in advance, it was crucial to photograph twice to guarantee good quality of at least one of the images. The contribution of this work is invaluable to our current knowledge even given these restrictions.
preserved their historical archive despite the colonization of Eritrea, the liberation war, and all other obstacles. This gives hope that one day someone appropriately trained will have the opportunity to make a holistic documentation of this unique manuscript and the archive it contains. At the moment both points have to be acknowledged: that due to the sets of images available nowadays, our chances of studying Dabra Libānos archive documentation has improved since Conti Rossini's publication, as the research by Marie-Laure Derat clearly shows, 11 and that our understanding of the archive remains partial (and will be always partial). Discussed in this contribution will be how TEI encoding helps organize the existing scattered knowledge and make the most of the available information, by aligning it and contextualizing it in the Beta maṣāḥǝft research environment. The advanced encoding methodology enables the production of visualization and analytical tools that can be implemented for other manuscript descriptions and to exploit the encoding for analysis.

Physical Description of the Archive and its Strata: Identifying Units
The Golden Gospel of Dabra Libānos za-Ham is a title assigned to an archive found in the monastery of Dabra Libānos in the village known as Ham, today's Eritrea.
As it was noted above, the core part of the archive is the four Gospels codex. However, there are other by-side folia, whose exact relationship to the four Gospels codex is crucial for our understanding of the archiving process. Before discussing the encoding, it is necessary to provide some details on the archive's physical structure and disposition of the leaves, as provided by the previous scholars, who were able to see the archive in situ. These leaves do not constitute a unit of any kind, most not even a quire, and their grouping has been justified basically merely through the typology of contents and the researcher's interests.
As stated above, this archive was introduced to the academic world in 1901 through Italian scholar Carlo Conti Rossini's title L'evangelo d'oro. 12 Conti Rossini was able to visit the place and see this archive, which resulted in the publication of the archival documents. In Conti Rossini's Introduction, the archive is described as follows: Il manoscritto è, come tutti i codici abissini, in pergamena. Della consueta copertina di legno, la tavoletta posteriore è esternamente foderata di metallo giallo tutto a rabeschi e ad ornamenti; la anteriore ha foderati dello stesso metallo soltanto i bordi, su cui è incisa la leggenda […] A questi evangeli sono stati premessi, posteriormente, altri numerosi fogli. Anzitutto v'è un quinternetto, 4 fogli di mm. 18 × 16, contenenti i due primi la fine di un Gadla Qirqos, […] e gli altri due aventi preghiere e atti di donazioni alla chiesa. In mezzo a questo quinternetto ne fu inserito un altro di 8 fogli, di scrittura del secolo XVII, a due colonne di 18 righe ciascuna, contenente altri atti di donazioni. Seguon due fogli, di varie mani, delle dimensioni del ms. degli evangeli, e di cui il primo certamente proviene da un codice molto più antico: hanno atti di donazione e sul verso del secondo foglio una grande croce cofta. Dopo altri cinque fogli, con la lettera d'Eusebio a Capriano e con sette qamar, la cui minuta e sottile scrittura è quasi interamente svanita, son 15 nuovi fogli, aventi al recto del primo la figura d'un tempio e in tutto il resto atti di donazioni, copiati da varie mani e in secoli diversi. Sono in tutto 34 fogli aggiunti a mano a mano al manoscritto donato da re Salomone. 14 'The manuscript is, as all Abyssinian codices, made of parchment. The backboard of the usual wooden binding is covered in yellow metal decorated and adorned. The front board has only the borders covered in metal and on it a legend is chiselled ወአሰ This is the starting point for any representation of the manuscript, in the absence of a catalogue description. Before one looks at how it can be complemented with later publications following an autopsy and accompanied by photographic documentation, one must look at what it reports and how this can be encoded into a TEI description based on the Beta maṣāḥǝft guidelines. 18 If the description by Conti Rossini is manually converted into a table and the different units are split according to their type and aligned to the main sequence available for the object, à La syntaxe du codex, 19 the following table can be constructed. 20 17 This general dating is now untenable: Schneider proposed to date the legend of the front cover to the twelfth-thirteenth century (Schneider 1989), while Siegbert Uhlig further suggested this dating to the writing of the Gospel (Uhlig 1988, 119-121). The authors intentionally leave this discussion open, implying that various units of production require their own dating, which will be discussed in upcoming publications. 18 http://betamasaheft.eu/Guidelines/, accessed on 30 August 2021. 19 'à La syntaxe du codex' is a reference to the methodology of stratigraphic analysis of a codex, of its syntax, that takes into account its observable discontinuities and allows to identify caesurae and, as a result, postulate unities in-between. This methodology is articulated in Andrist et al. 2013. 20 In Table 1, UniMat = Unit of the support material for writing; UniEcri = Unit of writing; UniDec = Unit of decorative elements; UniMep = Unit of layout; UniCont = unit of content. With a slash before an abbreviation we indicate beginning of the Unit and with a  Table 1 shows that Conti Rossini already applied a logic by which the important aspects of the description of a manuscript are its discontinuities and he grouped folia according to the caesurae he could observe. He then identified and edited in chronological order of date of pertinence the text, which he roughly classified as 'acts and donations'. Aside from some perturbation which will be seen and the omission of one leaf in the count (impossible to know which one), it is still correct and useful in many respects. This is encoded in the TEI by assigning an element <msPart> to each of the groups identified by Conti Rossini. It is also the oldest available description and, despite minor mistakes later observed, it attests to the state of preservation of the manuscript. Due to the possibility of nesting elements in a TEI description, in our description of the manuscript hosted in Beta maṣāḥǝft it was possible to represent the fact that the second quire is within the first, by nesting two <msPart>s. 21 It was also possible to represent the grouping by size of the leaves Conti Rossini had already reported and make a <msPart> for the entire group of leaves of that size and, slash after an abbreviation we indicate its end, in the same way as it is done in Andrist et al. 2013. 21 https://betamasaheft.eu/DabraLibanosHamGG1. within it, <msPart>s for each of the units identified already at the beginning of the last century.
No innovation or specific choice has been made for this manuscript, standard practice has been implemented to add a little depth and semantic to the available description.
In 1975 this archive was also viewed by Schneider, who took some 20-30 pictures, mostly of documents known to some extent already from Conti Rossini's edition. Schneider left many notes but published only a single one-page article discussing elements for dating the codex, 22 which will be referred to later. Concerning the physical description of the codex's structure, Schneider says the following: À l'intérieur il y a d'abord 12 folios d'un format plus petit (18 × 16 cm.) qui n'appartiennent pas au codex proprement dit. After Schneider, the monastery of Dabra Libānos was visited by Bausi in 1993 and1994. Having no access to Schneider's pictures, Bausi took his own set of images. While the set of images taken in 1993 mostly coincide with the folia Schneider photographed, differing in only a few folia, those taken in 1994 introduce a completely new documentation of 'a smaller' archive containing document copies. This documentation enables a better understanding of Conti Rossini's publication and helps address further questions concerning archiving processes and 'copying of archival notes'. Bausi writes the following: In sintesi, Conti Rossini distingueva il testo del Vangelo (128 ff.) da un gruppo di 34 fogli aggiunti in principio, rilegati in due diversi fascicoli, a loro volta analizzabili in gruppi di fogli omogenei. Il primo fascicolo, di dimensioni minori dei fogli dei Vangeli, di 12 ff., divisi in due gruppi di 4 e 8 ff., con gli 8 ff. inseriti dopo i primi 2 ff.; il secondo di 22 ff. e della stessa dimensione dei Vangeli, formato da 2 ff. più 5 ff. più 15 ff., per un totale complessivo di 34 ff. 24 'To summarize, Conti Rossini distinguished the text of the Gospel (128 fols) from a group of 34 leaves added at the beginning and bound in several fascicules, which in turn could be broken down into groups of homogeneous leaves. The first fascicule, smaller than the leaves of the Gospel, made of 12 fols, consists in four and eight leaves, with the eight leaves added after the first two; the second consists of 22 leaves and of the same dimension of the Gospels, consisting of 2 leaves plus 5 leaves plus 15 leaves adding up to a grand total of 34 leaves.' Next to the above-quoted description, Bausi noted 35 leaves and not 34 as previously reported. 25 Converting Bausi's description into a table à La syntaxe du codex, a table identical to that of Conti Rossini's appears with one extra leaf. Bausi also identified several texts that do not appear in Conti Rossini's edition and located as many as possible of those known to their exact relative location (signature of picture and number of folia, according to his own foliation). This work complemented Conti Rossini's efforts adding a precise location of the texts, that had previously been absent.
Bausi briefly repeats what Conti Rossini wrote regarding the archive arrangement introducing, however, the notion of a group of thirty-five folia, added at the beginning, bound in two different quires. The identification and location of the documents and their copies have been a major addition to our knowledge of the object enabling an encoding of this information, adding information of relative location to <msItem>s in each relevant <msPart>. Creating a Textual Unit record, with the associated Clavis Aethiopica identifier, 26 into Beta maṣāḥǝft for each edited document in Conti Rossini enabled an indication of those documents, which in several cases appeared in more than one location within the group of added leaves.
Schneider and Bausi took photos that were eventually made available to scholars, thus providing further grounds for the encoding. Firstly the attribution 24 Bausi 1997, 47. 25  of each set was noted in the main manuscript record, then a <facsimile> 27 was linked to each set of photos taken, which contained the sequence for that image set. It goes without saying that all three image sets, Schneider's of 1975 and Bausi's pair of 1993 and 1994, were only partially overlapping. To anchor description, images, and manuscript text to create a solid basis for a precise relative location of the texts, the Transkribus tool was used on selected sequences of the images and from there an encoded segmentation of the images was exported. 28 In so doing, it is now possible to retrieve fragments of the images and reuse them in their precise referencing. Though not carried out in this contribution, it is demonstrated in the web view of the XML edition.
This made it possible not only to obtain encoded and referenceable elements with an identifier for each region on each page photographed, but also to generate an automated transcription, using the model elaborated at the Hiob Ludolf Centre for Ethiopian and Eritrean Studies and make it available to any user of Transkribus. Now one could refer to each line precisely on each available photo 27 On TEI facsimiles: https://www.tei-c.org/release/doc/tei-p5-doc/de/html/PH.html#PHFAX. 28 Transkribus is an expert tool provided by the Read Coop, see https://readcoop.eu/transkribus.
It performs layout analysis and supports Handwritten Text Recognition (HTR). of the manuscript, and count on an incorrect (but better than nothing) transcription of the text at that line, linked with the exact portion of the image carrying it. This, however, did not solve the issue of having three different sets, only partially overlapping. A fourth <facsimile> was added copying over the best images for each folium in the sequence, aligning that to the transcription instead. The new facsimile is actually a digital reconstruction matching the existing available images and the transcribed text, performing the same function digitally as a paper model, which both authors constructed separately for this effort at different stages. Needless to say, a digital surrogate will never be as flexible and useful an heuristic tool as a material surrogate, especially one handcrafted.
In this <facsimile> we used <surfaceGrp>, to gather surfaces of folia rather than surfaces of images as dealt with by Transkribus, with groupings of leaves and their digital surrogates thus described. Transkribus exports a <pb> and related <facsimile> for each image of an opening. Here there is one <surfaceGrp> for each folium containing a zone that is the left side of the opening of an image (the recto of the folium) and the right side of the following (the verso of the folium). No photographic documentation is available of folia 1, 2r, and 35v, as well as the entirety of the Gospel (some images of the texts within it were made by Schneider).
The descriptions determined to some extent the selection of photos to be taken, together with the availability of time. Furthermore, the organization of the photos and their archiving and description was affected. The observation of the images and the descriptions enabled proposing the reconstruction of some of the transformation which happened in time. The photos determine a reconstructed hypothetic sequence which is encoded and upon which the contents have been reanchored alongside their transcription. This has been done by means of milestone elements linked to the images. These should clearly connect all aspects and information available.
As Bausi introduced a foliation taking into consideration two units defined by him as 'Fascicolo I' and 'Fascicolo II' and there are no systems of reference in Conti Rossini and Schneider, it is useful to provide an extraction from the TEI encoded alignment of the systems to reference the manuscript, which includes the new one. The updated encoding and alignment of information, and the unification of the system of reference in a single foliation sequence, makes it possible to also update the correspondences already provided by Bausi in the following table, where line references are also provided, as they became available due to the segmentation in Transkribus. The above table shows an additional improvement providing a good contribution while producing an encoded version of this manuscript's description, which is the classification as main content or addition. At this stage, the work allows a far more precise analysis of the units available in the portion of the codex under scrutiny. The final goal of this analysis is to reconstruct an archiving process.
Having now in our TEI encoded version identified parts, contents (distinguishing main contents and additions), each available image and system of reference, and each text region and line, it was possible to move on to encode decorations, hands, layout, as well as the ruling, where possible, assigning a @xml:id to each, a <locus> with relative location correct to the line when possible and @corresp to link elements to one another.
An <msPart> was declared for almost each folium of the so-called 'fascicolo II', to respond to the current state in which the loose leaves have become such independent units. These units are already the result of an analysis of the discontinuities, carried out before the encoding. The presence of so many convergent discontinuities to make for so many units provides further confirmation of the present structural description, 31 and serves as a basis by describing the current status (current UniCirc = circulation unit) and furnishing enough identification to carry out the description.
The table that can be extracted from the encoded description, taking into account all the previous documentation, is as follows (convergent discontinuities have been marked by a thicker line). 31 See Andrist et al. 2013, 8-9;Liuzzo 2021, §18. In other words, based on La syntaxe du codex, whenever we find two (or more) discontinuities that coincide, for example a textual unit (UniContMain) and a layout unit (UniMeP) or a material unit (UniMat) and a decoration unit (UniDec), etc., they are considered here as convergent discontinuities and can be used as a basis for postulating a Unit of Production. The identifiers in this table reproduce the ones assigned in the XML file thus ensuring consistency. The description in the script linked at the top of Table 5 explains also how more precise boundaries than folia may be extracted from the encoding.
The information employed for drawing Table 5 is limited in three aspects: (1) three sets of images do not cover the entire archive; (2) the present analysis has been done without direct contact with the archive; (3) the real historical formation of the archive will remain an enigma as much information remains unavailable.
When one has to rely on fragments of knowledge, the TEI is of a great help in structuring the knowledge and making it transparent. One can work at the level of fragments (photographed fragments), describe everything that may be described, align each piece of content with a precise fragment, thus forming a solid basis for various hypotheses, without fearing that new evidence will destroy the work. As a result, up to this point entirely verifiable and available information has been provided that is editable by anyone else with an interest in so doing.

Using the Units for a Reconstruction of the (Trans)formation of the Archive
We can exploit the extracted information and its organization to add further observations. These observations can then be encoded in the XML file as <re-lation> elements following a Web Ontology referring to the specific meaning this has in the Semantic Web based on La syntaxe du codex. 32 What follows is a lightly edited reproduction of the content of the XML file edited by the authors as text, further enriched by encoding. It is an attempt at a structural description which takes into account the analysis carried out doing the encoding. It is offered in this instance also to demonstrate how the encoding of a structural description may be done entirely in a machine-readable way and, of course, as an enrichment of running descriptions of the observations, hypotheses, and conjectures researchers have made. The proposed reconstruction is entirely hypothetical, of course. The identifiers of production and circulation units can be found in the XML file.
First one proceeds with the observation of two transformations that are clear to all scholars: the 'addition' of the small quire and of the other leaves 'of the same size as of the manuscript' to the material containing the Gospels. The marked discontinuity here is one of content rather than a discontinuity related to the material but, upon closer examination, it is not possible to group these into a single unit.
Due to the different size of its leaves, compared to those of the rest of the manuscript, the small quire, bound together at the beginning of the text block, is a material unit. The eight leaves within the small quire are of a different parchment and are thus a separate material unit. Nothing indicates that these four bifolia were once separated, thus there is no sign that this was a quire or a production unit separate from the rest of the small quire. It is merely part of it, or rather the two outer bifolia were, in the same act of production, used to surround the inner leaves, which share the same layout, ruling, and writing. The small quire constitutes one production unit bound into a single quire. All other leaves are of the same size as those containing the text of the Gospel, constitute a material unit with the leaves containing the Gospels, and are homogeneous in content among themselves. The small quire was added to the manuscript. This is an A1 transformation, that is, 'addition of material support and content', according to the typology introduced in La syntaxe du codex, which was also used in the discussion above. There is therefore a UniCirc before the transformation and one after. The manuscript as we know it corresponds in terms of material to the produced circulation unit. The manuscript prior to that addition is another circulation unit of which one may be certain. For the latest additions, presumably not seen by Conti Rossini, it may be assumed they were made to the circulation unit which already had the small quire. This is, however, hypothetical, as there may have been additional or other reasons for Conti Rossini not listing these texts in his edition. This abstract event which affected the manuscript is an A2 transformation, 'addition of material support and content'. Each of these additions is a production unit, involved into one multiple transformation which groups different acts which cannot be distinguished further at the moment. The units before and after the transformation are UniCirc, but, as it is not possible to order them one by one, it is only possible to group distinct additions and consider them as they took place in one single transformation, even though each single transformation with its own UniCirc can be distinguished. The issue is to work out to which UniCirc the additions were made, and this too can only be done for some additions, if at all. What is left now, having been subject to additions as a whole, is a unit that has been subject to an UA1 transformation, that is, a 'union followed by additions'.
A further quite obvious observation is that, although lightly tightened together with strings, the leaves from 13 to 35 are basically all detached from one another. The leaves of the same size as the Gospel placed after the small quire have been detached from one another and it is not possible to determine their actual initial order and placement. This occurred probably before the small quire was added. If this occurred by dint of a copy being made from the contents of some of these leaves (see below), then the subject of the transformation is the circulation unit before the transformation which added the small quire. The transformation is a simple division (D3). The detachment however did not produce different units of circulation, as the leaves continued to circulate together.
Another observation regarding the two bifolia containing the quire is the appearance of a brownish colour in the photos (Fig. 3).
The two leaves featuring the end of the Gadla Qirqos are distinct from the rest due to the ruling pattern and material (Fig. 3), which were probably taken (at some point in time but it is impossible to estimate when) from another, unknown, manuscript to produce the small quire. It is necessary to define this as a production unit separate from that of the small quire, which is now inside it. Furthermore, the small quire's sequence of content tells of a single production intention and, however independent, is a production unit. The unit from the other manuscript was then subject to a transformation. This transformation is a simple division (D2), followed by the addition to another unit. Therefore it has to be said that the unit with the other ruling pattern, now inside this one, was also part of the same transformation. This transformation is an addition of support and contents (A1). Here too it is possible to group the transformation into a single, complex one. This is probably a MA1 transformation-'mutilation of one codex in order to add to it another one', where a codex has been somehow deprived of two bifolia to make our quire, which contains the homogeneous one.
The additions on folia 11 and 12 use the existing ruling in some way, but provide no indication of having already been present or not. If a table had been organized by quire, which is not suitable here as the quires are not as relevant as the leaves, a convergent discontinuity would be revealed. In Table 5 two convergent discontinuities show themselves. The outcome of this transformation is the observable production unit which also results in a new circulation unit. Whether the texts (UniCont) on p1 fols 11v and 12v were already present when the leaves were used to produce the 'small quire' is not known. Conti Rossini dates them later than the Gadla Qirqos text, which however is of no help in this instance. They were added to that material in any case. This is an A2 transformation, an addition of content. Not much can be added with this, if there is no precision in terms of reference to units which actually existed at some point.
The following, instead, are permutations of type P1, changes in the order of the folia, which produced new UniCirc, but not new UniProd (= production unit). Upon cross checking convergent discontinuities in the table view, it can safely be said that the folium numbered by Schneider 17/18 (the same as Bausi II 15) is the leaf Carlo Conti Rossini had prior to the one with the cross in his description, that is, the 'ancient leaf', which cannot be seen in any of the available photo sets but upon which Conti Rossini reports. Following Conti Rossini's description of the two leaves after the small quire of his grouping, only the one with the cross on the verso can be observed. In fact, photo 001 by Schneider (Fig. 4) 33 shows that the sequence displays the end of the small quire and then the folium with the Coptic cross on the verso. Of all folia available, the one that carries the texts which Conti Rossini dates the earliest is the one now numbered 27, which is delimited by convergent discontinuities.
Therefore it may be said that there was at least one permutation of the order of the leaves, before they were photographed in 1975, and the one that was at the top of the pile of leaves of the same size of the manuscript was 'moved' to the position in the photo sets as folium 27. This is a permutation of folia. For the <msPart> containing the Eusebian letter and the canons, identified by decoration and content unity, the sequence at Carlo Conti Rossini's time was (with reference to the current pagination) the following: 27 (the leaf with old texts), 13 (cross on verso), 14, 15, 16, 17, 18, 19 (tempio on recto). The permutation results in a unit of circulation only.
The continuity between leaves 19 and 26 may be based on the convergence of discontinuities of content (CCR 7), material, decoration, layout (one column), and on the two-lines text at the top of 19r and 26r, which is apparently of the same hand.
It can be seen that another permutation occurred regarding the most recent photos. This is easily confirmed and dated after 1975, for, in the Schneider photos, the current folium 26 follows directly onto 19 and is paginated with 3 for the recto and 4 for the verso. This is another permutation of folia, resulting in a further unit of circulation only. It is assumed, however, that this came about due to Schneider's photographing practice. In all likelihood, he moved the folium himself indicating the displacement by means of a sheet of paper.
There is another observable permutation affecting the same-sized group of leaves as the one containing the Gospel. The leaves Schneider paginated as 7/8 and 9/10 had been moved to their current placement at 24 and 25 by the time Bausi took his photographs of the folia in 1993 (Bausi II 12 and II 13). It may have appeared obvious to someone that, in their position at that time, they interrupted the continuity of a content unit (CCR 20). This is another permutation of folia. The further permutation results in a further unit of circulation only. This unit, which contains what was last seen, corresponds to the described object.
It is now possible to define a reconstructed production unit which represents the state of the leaves surrounding the decorated leaves and possibly closer to what was observed by Conti Rossini. Other texts which are discontinuous with the rest may have been added to this unit, such as CCR 8.
A unit, corresponding roughly to a fascicule, circulated until Schneider took photos as 20,24,25,21,22,23. It is impossible to say more on this unit, whose unity at some point may be attested by a univocal deletion intervention visible on the photos. However, the layout of 24 and 25 stands out as a separate unit, characterized by the two-columns layout and the blank first column on the verso of both leaves. This allows one to define a further unit of the remaining leaves sharing the one-column layout and the dotted line separators.
What remains an important, albeit hard to evaluate, aspect is the relation between the texts present in the small quire and those in the leaves of the same size of the main content, namely the Gospels. In the following table, we have extracted these contents side by side. There is little one can say without the support of proper palaeographic dating and knowledge on the rest of the manuscript and its structure, on which one copy was copied from the other. It appears sensible to think instinctively that the damaged or physically deteriorating one would be replaced by the neaterlooking one. The addition of content and material involving CCR 7 tells of a separate production intent to that of the Eusebian Canons following the existence of this, therefore, ignoring the palaeography, the possibility of it being a copy from the so called 'small quire', originally written here, or being copied from another source, into the leaves of the same size of the Gospel, remains open. Yet, from the data available, the choice that was made (in terms of an act of copy from the bigger leaves to the smaller ones) does not appear to cover mostly 'endangered' documents. Even if that were the case, some ordering would have to be sought, but in that regard also not enough elements are present to reach a conclusion. If copying into the small quire from the larger leaves and skipping the independent unit made of 28 and 29, then copying CCR 20 before CCR 19 would appear to have been a kind of reverse working. However, the leaves may have been already separated and in no particular order. Nor is there any possible matching with the hypothetical grouping of leaves by their layout, but the same grouping as identified by the discontinuities occurs if we look at the sequence of leaves hypothetically copied. At some point, there may have been a sequence, attested by the hypothetical copy in the small quire, with 28, 29 followed by the group formed by 20, 21, and 22; then 19 and 26; then 23, on the recto of which CCR 24, CCR 25, or CCR 3 were not yet present, but were all added later, seemingly by the same hand that also included some additions in the small quire. Following this logic, the question arises as to why CCR 21, which is written in the same hand as CCR 19 and CCR 20 (or a hand very similiar to it), is not presented among the copied documents in the small quire? At this stage, one should consider the content and its interest for the monastic community of Ham. While CCR 18,19,20,22,and 33 are directly concerned with the rights and rules of the community, CCR 21 is dedicated to the church of Māryām za-Saʿagā with an interest 'so that they may celebrate the commemoration of Maṭāʿ'. 34 Although the authors of this contribution were not yet able to locate the church of Māryām za-Saʿagā, everything suggests that the CCR 21 benefactor is not that of CCR 18, 19, 20, 22, and 33. Thus it is concluded that the small quire is a collection of documents concerned with the monastic community of Ham itself.
Following the reverse process of copy encountered for CCR 20 and CCR 19, however, on the basis of the above-postulated unit of circulation, CCR 23 is to be expected here, but it is absent. Here another logical approach may stand. CCR 23, 'Land grant by Gabra Krǝstos', is attributed to Gabra Krǝstos, ʿāqqābe saʿāt of Maṭāʿ, 35 and thus directly related to the community. Here two observations are worthy. The first is that the title 'ʿāqqābe saʿāt' was deliberately erased (see fol. 33rb, l. 5), an action one may not discount as insignificant. Conti Rossini's publication supplied it on the basis of its occurrence at the end of the document, where it was not erased. The second observation is that the copied notes were attributed to royal figures, while that of CCR 23 is not. Where the sequen-34 Conti Rossini 1901, 206, l. 3. 35 ʿĀqqābe saʿāt, lit. 'keeper of the hour', title given to the abbot of Dabra Libānos of Ham, and, at the end of the thirteenth century, also to the abbot of Dabra Ḥayq (Derat 2018, 273). tial expectation is CCR 23 there is, instead, CCR 34, that is not a royal donation, nor is it attested elsewhere and may well have been added from another source or newly composed, perhaps in consideration of space left in the quaternion.

Conclusions
Following the work of organizing existing knowledge, we provided a collaboratively edited, and editable, standardized description of the leaves of the codex, the texts of which were first edited by Conti Rossini. Having included in that XML edition all available materials with some added value, it is believed that the community of interested scholars has, if not additional information, an optimized collection of available data on this important collection of texts, that may be expanded and referenced further in or out of the research environment from which this work has been based.