I will say that the main check on my enthusiasm was that my method seems
to abandon the (anti-)parallelism between sic and corr.  If corr marks
the process of correcting a physical document (and contains text in the
state of being added or deleted physically) and emend marks the process
of emending a conceptual document (and contains text in the state of
being added or deleted conceptually), what does sic do?  And what is the
equivalent of sic for emend?  Obviously one solution is to assume that
corr not only describes a process but one's attitude towards the process
(in which case "corr" would mean "correction I accept" and "sic" would
mean "correction" I reject), but a) this information actually seems more
like it ought to be on a cert. attribute or something, b) if there is a
distinction between "corr" and "emend," there ought also to be one
between "corrSic" and "emendSic".

My only reasons for not trying to go a level higher and see an "alt"
element are procedural and structural:
1) from a procedural point of view: an apparatus is not a process, it is
a state (i.e. it represents a collection of readings all of which are
intended to be seen at the same time); corrections and emendations are
processes by which a text changes from an input to an output (even if we
decide to show both in our final scholarly product).  Because of this, I
would encode a variorum as an apparatus, rather than a series of
emendations (i.e. I agree with the TEI presentation of the Chaucer
examples in the app section, where they handle readings from different
modern editions as alternate readings).
2) from a structural point of view, corrections/emendations and
apparatus occupy very different parts of a document:
corrections/emendations are part of the textual flow; apparatus are
closest to notes--and certainly out of the textual flow.

Julia Flanders wrote:

>I must say, at first blush this seems fairly ingenious to me. The
>Women Writers Project has encountered situations where it would have
>been very useful to be able to include tags inside an attribute
>value, and corr= (in our case) is certainly the most frequent.
>
>I wonder, though, whether this question could be addressed at a
>higher level of abstraction. You say that <app> didn't feel right
>because it referred to a specific kind of document structure, and
>because you were already using it somewhere else. But what you've
>created is similar (in that it coordinates several readings within a
>larger structure), and the fact that you can imagine other similar
>structures with other names (e.g. <emend>) suggests that there's
>something larger at work here. I wonder whether what's really needed
>is something like
>
><alts type="corr | emend | readings | witnesses | etc.">
>    <alt type="orig | add | del | lemma | reading | witness |
>etc.">the first reading of the group</alt>
>    <alt type="orig | add | del | lemma | reading | witness |
>etc.">the second reading of the group</alt>
>    <!-- as many more as you need -->
></alts>
>
>The semantics I'm imagining for those attribute values are something like:
>
>corr=a correction to the original by a scribe
>emend=an emendation to the original by an editor
>readings=alternative hypothetical readings in cases where the
>original is unclear
>witnesses=alternative witnesses where more than one exists
>
>orig= the original reading from the source
>add=something added by a scribe
>del=something deleted by a scribe
>lemma=a lemma
>reading=one possible reading (of several, in cases where the original
>is unclear)
>witness=the reading from one witness (of several, in cases where more
>than one exists)
>
>It should also be possible to nest these, e.g.
>
><alts type="witnesses">
>    <alt type="witness">the first witness of the group</alt>
>        <alts type="readings">
>                <alt type="reading">one possible reading of that witness</alt>
>                <alt type="reading">another possible reading of that
>witness</alt>
>        </alts>
>    <alt type="witness">the second witness of the group</alt>
>                <alt type="reading">one possible reading of that witness</alt>
>                <alt type="reading">another possible reading of that
>witness</alt>
></alts>
>
>The usual methods of indicating priority among readings and other
>useful information could be used.
>
>Looks like a feature structure, doesn't it...?!
>
>Certainly one wouldn't want to use this in every case where one now
>uses <sic> or <corr>, but for some kinds of projects and for some
>kinds of readings, I think it would be helpful to have a
>generalizable structure which could be used for all kinds of
>alternatives (including ones now unforeseen), rather than
>anticipating a few kinds of structures and giving them specific names
>which rule them out for rarer cases.
>
>best wishes, Julia
>
>
>
>>Dear colleagues,
>>    I have a number of questions about a prejudice of mine: I don't like
>>using attributes to record what might otherwise be considered character
>>data, e.g.
>>
>>(1) <corr sic="be&ouml;">f&ouml;o</corr>
>>
>>Since corr and sic are mirror images of each other, the Guidelines leave
>>the choice of which text goes on the attribute up to the encoder, but
>>suggest that the attribute be reserved for the form that does not
>>contain any child tags. I don't like this because it seems to me to make
>>the sic information harder than necessary to process, and because the
>>content of the "sic" attribute seems to me to be less obviously metadata
>>than something like resp="dpod", and because I find that the "sic" and
>>"corr" forms often contain equal amounts of children.
>>
>>The TEI guidelines do suggest another encoding is to pair <sic> and
>><corr> elements (18.1.5), but note that the connection between the two
>>is then implicit and positional rather than explicit unless enclosed in
>>an element such as "app". In my own encoding, therefore, I have recently
>>changed approach I believe more radically: I use the following syntax:
>>
>>(2) <corr><del>be&ouml;</del><add>f&ouml;o</add></corr>
>>
>>The assumption here is that "correction" is really the name of process:
>>i.e. a recognition that somebody decided to fix something; deletion or
>>addition, however are the names of specific states: text either is
>>marked for deletion or it is not.  This is not a perfectly developed
>>rationale, but hopefully it is to be understood.  I realise the TEI
>>suggests using <app> as a method of encoding this: since, however,
>>apparatus is a standard document structure and, in my case, is used
>>elsewhere in my edition in the sense of "alternate readings collated
>>
>>
>>from different witnesses", I really dislike that solution.
>
>
>>I have three questions, then:
>>a) is my dislike of the attribute in example (1) reasonable?
>>b) is the encoding in (2) reasonable or seen as completely unnecessary?
>>c) if the encoding in (2) is non unreasonable or unnecessary, what do
>>people think of my next stage, by which I use <corr> for changes to the
>>physical text of a monument (i.e. scribal changes in an actual witness)
>>and a new but nearly completely parallel element <emend> for changes to
>>the text-as-concept (i.e. by a subsequent editor)?
>>
>>-dan
>>
>>--
>>Daniel Paul O'Donnell, PhD
>>Associate Professor of English
>>University of Lethbridge
>>Lethbridge AB T1K 3M4
>>Tel. (403) 329-2377
>>Fax. (403) 382-7191
>>E-mail <[log in to unmask]>
>>Home Page <http://people.uleth.ca/~daniel.odonnell/>
>>
>>

--
Daniel Paul O'Donnell, PhD
Associate Professor of English
University of Lethbridge
Lethbridge AB T1K 3M4
Tel. (403) 329-2377
Fax. (403) 382-7191
E-mail <[log in to unmask]>
Home Page <http://people.uleth.ca/~daniel.odonnell/>