The issue of legacy data is real enough, since most manuscript collections will have some form of catalogue (a printed book or card catalogue) and only very rarely is one required to describe a group of manuscripts ex nihilo. The existing catalogue may be of such great authority that one feels that nothing in it may be changed, and one may even want to reproduce as closely as possible the physical appearance of the original; more commonly, one may simply lack the time, expertise, or funding necessary to recast and augment existing data, although one may still want to provide sufficient markup to facilitate some basic searching capabilities. On the other hand, the existing catalogue may be old and out of date, and one may wish to update the information it provides on the basis of more recent scholarship; one may also want to describe a manuscript, a group of manuscripts, or an entire collection in much greater detail than has been done before, and one may therefore feel in no way bound to reproduce or even acknowledge existing data. Then there are those who are not engaged in cataloguing at all, but who rather are interested in encoding codicological and/or philological data (whether from an existing catalogue or not) for other scholarly purposes. Ideally, a tagset for manuscript description should be able to accommodate the full range of eventualities. MASTER was reasonably good at catering to the latter type of user but was too inflexible to be of much use to those who were either not able or not of a mind to modify existing data, but were still interested in providing wider access to it.

The solution proposed by TEI-MMSS was to loosen up the DTD, essentially allowing the sub-elements of msDescription to occur in any order and to repeat, and also to occur within running prose. While this did undeniably ensure greater flexibility, the opportunities for abuse and illogicalities it opened up were worrying: things can, after all, become so flexible that they lose any kind of structural integrity.

Recognising both the failure of MASTER to deal in a satisfactory way with things like the issue of legacy data and the inadequacy of the solutions proposed by TEI-MMSS, the TEI Council in 2002 appointed a special task force whose job it was to review the current state of TEI-based recommendations for the detailed description of manuscript materials [...], to identify and define a common subset of those recommendations adequate to the needs of the TEI community [and] to document that set of recommendations in such a way as to facilitate their inclusion in TEI P5, once approved by the TEI Council.

The present writer served as chair of the task force; other members were Merrilee Proffitt and David Birnbaum, at the time both members of the Council, as well as the two TEI editors (ex officio). The Council's original charge and other documents pertaining to the work of the task force can be found on the TEI website ().

If one wanted to put this catalogue entry into machine readable form, but was not interested in or was unable to add any further markup, one could begin by either scanning or keying in the text, providing the necessary information for the mandatory msIdentifier element, and then simply wrapping the three paragraphs of prose in p elements, as in the following example:

A simple conversion of this kind would take no more than a few minutes if done by hand, and could be largely automated. Should one wish to approximate more closely the appearance of the printed text, the rend attribute could be used on the three p elements, with appropriate values; alternatively, the first paragraph could be tagged as a head and the third as a note (and moved inside the previous, and now only, p). The result would, with a suitable style sheet, be displayable in a browser; it would not, however, be very useful for search purposes (other than searches for shelfmark and running number).

In order to provide slightly richer markup, one could wrap the paragraphs in the appropriate special-purpose first-child-level elements of msDescription and add some of the phrase-level elements available when the manuscript description module is in use. Doing so necessitates some slight reorganisation of the data vis-à-vis the original printed source, but no actual rewriting of the text. Now, however, one would be able to search specifically for title, material, and date and place of origin.

One could also restructure the entire entry, using the full range of elements and sub-elements available within msDescription.

Note that here again it is largely a question of cutting and pasting sections of text from the original; there has been no rewriting of the text as such, although one might well at this stage wish to do so, extracting the data and updating and supplementing it as required. And clearly this is the most sensible thing to do: extract the data rather than worrying about the exact wording—which is precisely what the Bodleian have done in their electronic catalogue:

What one may not do is the following, viz. take the pre-existing text as it comes and mark it up with the relevant parts of msDescription:

Marking up legacy data in this way, which is what was proposed by TEI-MMSS (the example above parses against their DTD), is more time consuming than the cut and paste method, but requires an equal degree of familiarity with manuscript description conventions on the part of the person doing the conversion, who must be able to identify which parts of the text pertain to content, physical description, history, and so on. More importantly, flattening out the hierarchical structure of msDescription in this way leads to serious illogicalities, such as in the first section, where no fewer than six paragraphs are opened and closed within what is arguably a single sentence.

As was said before, the relative inflexibility of MASTER made it less than ideal for those working with large amounts of legacy data and a tight budget and/or timetable, that is to say probably the majority of electronic cataloguing projects today, while the complete lack of constraints in the system proposed by TEI-MMSS made it unsuitable for those wishing to have more rigorously structured data. The solution put forward in P5-MS, it is hoped, allows the needs of the full range of potential users to be accommodated.

In MASTER and TEI-MMSS there were type and status attributes on msDescription (and msPart). The intention with the former was to distinguish at the very top (as it were) between manuscripts proper and archival material (charters etc.). In P5-MS, however, the form attribute on the new grouping element objectDesc (discussed further below) provides the same information. The status attribute, on the other hand, was defined as specifying the compositional status of a manuscript or manuscript part and had in the MASTER scheme the possible values uni|compo|frag|def|unknown. TEI-MMSS proposed that these should be split in two, as a manuscript can be both composite and defective, with a status attribute the possible values of which were frag|def|unk, and a composite attribute with possible values y|n|u. Over the years it had been pointed out several times that, for one thing, status should probably be called msStatus or something similar, since it refers to the status of the manuscript and not of the description, and for another that the distinction between composite and non-composite manuscripts is in any case inferable from the presence or absence of msPart elements; similarly, a manuscript which is defective will (or should) have defective="true" on msContents or msItem. Thus the only possible value of status which cannot be inferred from other places in the document is frag; this was always potentially problematic anyway, as the distinction between a fragmentary and a defective manuscript is somewhat arbitrary (a rule of thumb was that if the cataloguer thought it likely that more than 50% of the leaves were missing, the manuscript was a fragment, otherwise it was merely defective). It was decided therefore to drop both these attributes in P5-MS.

TEI-MMSS also proposed a dateAttrib attribute on the elements msDescription and msPart, with possible values dated|datable|unknown, the idea being to distinguish, again at the topmost level, between manuscripts which are dated internally, those which can be dated on the basis of other evidence, and those for which an approximate date has be assigned by a scholar on the basis, for example, of palaeographical or orthographical features. Recognising the central importance of these distinctions to manuscript scholarship, the task force initially decided to incorporate this attribute, even proposing an analogous placeAttrib attribute, with possible values localized|localizable|unknown. Here again, however, these attributes merely repeat at a higher level information for which there already is a place elsewhere in the document, namely the evidence attribute, available on origDate and origPlace, which has the possible values internal, meaning that the manuscript is formally dated or localised by the scribe, external, meaning that the manuscript is datable or localisable via inferred knowledge from some aspect of the book itself, and conjecture, meaning that in the absence of internal or external evidence an attribution of place or date for the manuscript has been made by the cataloguer or scholar on the basis of his or her expertise. A value of internal for evidence on origDate thus means dateAttrib="dated", a value of external means dateAttrib="datable", and conjecture means dateAttrib="unknown", and the same is true of the values of evidence on origPlace. For this reason it was decided also to drop these attributes, leaving only those which are globally available.

This does, of course, mean that one can only distinguish diplomas from codices, composite from non-composite manuscripts, or dated and datable manuscripts from undatable ones if one chooses the more structured option, which will not apply to those hoping (or forced) to make do with an absolute minimum of tagging. It was felt, however, that there should only be one set of mechanisms available for making such distinctions, and that those wishing to do so should make use of those mechanisms.

In both MASTER and TEI-MMSS the sub-elements available within msIdentifier were country, region, settlement, institution, repository, collection, idno, and altName, which was intended for a former shelfmark or some name other than the shelfmark by which a manuscript is known. In MASTER it was decided that three of these, settlement, repository, and idno, should be required, since they provide what is, by common consent, the minimum amount of information necessary to identify a manuscript uniquely: place, repository, shelfmark. TEI-MMSS argued that all the sub-elements of msIdentifier should be optional and repeatable, and that msIdentifier itself should be allowed to repeat, as there are cases, for example, where there are two legal owners of the same manuscript, where a collection legally exists in more than one city, where a repository only has one manuscript, or only one of any significance, which has no call number as such but is instead known by one or more names, or where the manuscript was formerly owned by someone else, or had another shelfmark. TEI-MMSS also proposed that msIdentifier should be used when referring to another manuscript within a description, and pointed out that in such cases the sub-elements might appear in varied order and interspersed with plain prose. A type attribute was proposed, with the values primary|former|cited|msPart, to keep these various uses separate. The task force felt quite strongly that there should only be one msIdentifier per msDescription, i.e. that pertaining to the manuscript being described, and that most of the scenarios envisaged by TEI-MMSS could be dealt with either by existing means or through the introduction of an altIdentifier element, in effect a nested msIdentifier, which could be used for former shelfmarks or other alternative forms of identification, such as the running number of a printed catalogue, as in the example from the Bodleian Library given above. altName, renamed msName, is still available for nicknames. One of the cases mentioned by TEI-MMSS, that of a manuscript known only by a name but with no shelfmark as such, required a more radical solution, as did another case, not mentioned by TEI-MMSS but brought to the attention of the task force, namely that of scattered manuscripts, that is to say manuscripts which have been split up but which are nevertheless treated by the scholarly community as single units. One well-known example of this is the Old Church Slavonic manuscript known as , parts of which are found in three separate repositories, in Ljubljana, Warsaw, and St. Petersburg. This could be dealt with using msName followed by a series of altIdentifier elements, with an appropriate value on the type attribute to indicate the nature of the relationship between them.

The msHeading element was used in MASTER/TEI-MMSS in order to provide a short summary description of a manuscript, sometimes called a tombstone, such as might be displayed or printed as a heading to a catalogue description; it was made clear by both groups that msHeading was not intended to stand in place of a proper description, and that the elements internal to it (author etc.) were not there for search purposes. Unable to see any significant difference between msHeading and the standard TEI head element, the task force decided to replace the former with the latter. The content model of head allows for phrase-level elements only, which in P5-MS include title, origDate, origPlace, and note, leaving author, resptSmt, and textLang unavailable from the old content model (although head may also contain bibl, which contains author, resptSmt). But, as was said, since this element is intended only to provide a heading for a manuscript description, rather than the description itself, structured information on contents, date and place of origin, language, and so on should be given under the appropriate elements. Here, for example, is an example of an msHeading according to MASTER/TEI-MMSS:

The corresponding head element in P5-MS is:

Note that in the P5-conformant example only title has been used, chiefly for rendering purposes.

There are now two forms of the msItem element, one still called msItem, the content model of which is essentially the same as that of the old MASTER/TEI-MMSS element, and a new msItemStruct, which can be used for a more rigorously structured description; the contents of the two are identical and the only difference between them is that in the former the order of the elements is free whereas in the latter is constrained. Note that the finalRubric element, originally present in MASTER but subsequently removed, has been reinstated in both (although in msItem it would in theory also be possible to use rubric type="final"). A new element, filiation, has been introduced (borrowed from the Repertorium project) to provide a place for information concerning the manuscript's relationship to other surviving witnesses of the same text, its protographs, antigraphs, and apographs.

The greatest changes made to the original model proposed by MASTER and TEI-MMSS are in physDesc, where a number of grouping elements has been introduced in order to make the structure more logical, and several existing elements have been renamed for the sake of consistency. The first of these new grouping elements is objectDesc, which relates specifically to the text-bearing object. It contains two further elements, supportDesc and layoutDesc (both optional); supportDesc contains the elements relating to the physical object, or vehicle, on which the text is inscribed, support, extent, foliation, collation, and condition (all optional, but only in that order), while layoutDesc contains one or more layout elements, detailing the way(s) in which the text is organised on the page (or other surface). The structure of these sub-elements is essentially the same as it was in MASTER/TEI-MMSS, but at each level there is also the possibility of using paragraphs (p). Note than instead of the element form there is now a form attribute on objectDesc and similarly a material attribute on supportDesc, as it was felt that these were most likely to be of interest as search criteria (although the information may, of course, also be repeated in the prose).

The following, are two encodings of the same physical description, or at least that part dealing with the support and layout. First, the old MASTER/TEI-MMSS-conformant record:

Secondly, the new P5-MS:

Following objectDesc comes handDesc (formerly msWriting), containing one or more handNote (formerly handDesc) elements, and after that musicNotation, containing one or more paragraphs, decoDesc (formerly decoration), which contains one or more decoNote elements, additions, containing one or more paragraphs, bindingDesc, containing one or more binding elements, a new sealDesc element, containing one or more seal elements, and finally accMat, which formerly came under additional, containing one or more paragraphs. Here too, the structure of the sub-elements is essentially the same as it was in MASTER/TEI-MMSS, apart from decoNote, which previously had a large number of very specific attributes, viz. size, technique, style, and quality, all without fixed sets of values, and figurative and illustrative, the possible values of which were yes, no or unknown/non-applicable. It was decided to drop all of these, retaining only type and subtype (in addition to those globally available). Note that paragraphs are available as an alternative within all elements where there are special-purpose sub-elements, while it is hoped that more structured alternatives will be developed in the future for those elements for which there are none at present, binding for example.

An attempt at grouping the other first-child level elements of physDesc, that is, on the one hand, handDesc, musicNotation, decoDesc, and additions (what might be referred to as the meaningful inky bits as opposed to the matrix on which they are inscribed), and, on the other, bindingDesc, sealDesc, and accMat (things which happen to the manuscript after it has come into being and are less integrally a part of it), was abandoned, chiefly owing to a lack of suitable nomenclature, although it is hoped this may be taken up again at a later date.