I think there is nothing wrong with your suggested <noteSubj>. On the
other hand, if you want any kind of interoperability with generic TEI
tools (of which, I know, there are very few at the moment), you would
do much better to stick with either in-line <note> elements or out of
line <note> elements that point to markup. Furthermore using <note>s
makes your encoding clear to any reader even mildly familiar with the
TEI Guidelines.

> First, the descriptions of <note> in P4 pretty clearly identify
> it as something to be used to mark existing notes

As Lou has pointed out, this isn't really the intent, and I think
you've thoroughly demonstrated that a new example of a modern
editor's <note> is called for in P5.

> ... but for editorial reasons we are keen to keep the content of
> our document limited to the content of our source manuscript.

Although I adhered to this doctrine for many years, and do find it
quite comforting, I'm afraid it's soon to be a thing of the past. In
(unextended) P5 it will be all but impossible to apply consistently.
In which case, you may want to let go of it now, while you can still
proudly say you weren't dumped, it was *your* decision to leave the
doctrine. (I'm imagining the final scenes to _Annie_Hall_ as I write
this :-)

> Since <note> requires the note text to be included as content (ms
> text <note>here is the opinion of the electronic editor</note> more
> ms text), it violates this rule.

Only for in-line <note>s. Out-of-line <note>s, especially if they are
in the not-yet-named bag of holding that is a container for odds and
ends (see feature request 1022699 at,
probably don't violate this "rule".

> So, if the editor wants to comment on the first two letters in a
> word, the editor must place <c> markup around those letters only to
> act as references, not because the letters themselves warrant
> individual markup.

While this is true in P4 (although one might well argue for <anchor>s
or a <seg> over <c>s), it isn't entirely true in P5. In P5 a <note>
can use XPointer extensions to point to characters other than via an
entire element. However, it is quite possible that software to process
such extensions will be, shall we say, limited.

But more importantly, I don't see what the problem is inserting e.g.
<c> tags around the letters of interest at all. I think there may be
some advantage to being consistent, and using, <anchor>, <seg>, or
<ab> with type="noteTarget" (or whatever) for every occurrence of stuff
that needs to be marked up for no other reason than to be pointed at.
But I'm personally quite happy inserting such markup. IIRC the
Guidelines are quite clear that this is one suggested use of these
elements exist.

> Obviously, this can also be impractical when pointing a note to a
> large segment of text that contains a lot of markup.

This doesn't make a lot of sense to me. If the segment is a single
element (say <div type="chapter">), it doesn't matter how large it is
or how much markup there is inside it, one points to it with a single
IDREF. All done. So I'm guessing you mean an arbitrary segment of text
that, e.g., crosses element boundaries. In which case, it may be the
case that the more markup there is, the less likely you'd have to
insert "point at me" markup. E.g., modified from the Nameless

<sp who="Seyton">
  <l part="Y" id="sha-mac505016">
    <w id="xw01" wt="at" pos="at">The</w>
    <w id="xw02" wt="n" m="sg" pos="n">queen</w>
    <w id="xw03" le="i" wt="pnp" m="" pos="dt">my</w>
    <w id="xw04" wt="n" m="sg" pos="n">lord</w>
    <w id="xw05" le="be" wt="vp" m="" pos="vp">is</w>
    <w id="xw06" wt="aj" pos="aj">dead</w>
    <lb n="Mac.2338"/>
<sp who="Macbeth">
  <l part="Y" id="sha-mac505017">
    <w id="xw07" wt="pnp" m="" pos="pnp">She</w>
    <w id="xw08" le="shall" wt="vm" m="pt" pos="vm">should</w>
    <w id="xw09" wt="vp" m="inf" pos="vp">have</w>
    <w id="xw10" le="die" wt="v" m="ppt" pos="v">died</w>
    <w id="xw11" wt="av" pos="av">hereafter</w>
    <lb n="Mac.2339"/>
  <l part="N" id="sha-mac505018">
    <w id="xw12" wt="pcl" m="ex" pos="pcl">There</w>
    <w id="xw13" le="will" wt="vm" m="pt" pos="vm">would</w>
    <w id="xw14" wt="vp" m="inf" pos="vp">have</w>
    <w id="xw15" le="be" wt="vp" m="ppt" pos="vp">been</w>
    <w id="xw16" wt="at" pos="at">a</w>
    <w id="xw17" wt="n" m="sg" pos="n">time</w>
    <w id="xw18" wt="prp" pos="prp">for</w>
    <w id="xw19" wt="dt" pos="dt">such</w>
    <w id="xw20" wt="at" pos="at">a</w>
    <w id="xw21" wt="n" m="sg" pos="n">word</w>
    <lb n="Mac.2340"/>

If one wanted to annotate the passage from "lord" to "word", one
obvious method is to use an out-of-line <note> thusly:

  <note type="editorial" resp="DP" target="xw04" targetEnd="xw21">

(Although, to be fair, if one really only wanted to point out the
subset of that range that are words in the metrical lines spoken by
the characters, it would be better to use

  <note type="editorial" resp="DP" target="xw04 xw05 xw06 xw07 xw08 xw09
     xw10 xw11 xw12 xw13 xw14 xw15 xw16 xw17 xw18 xw19 xw20 xw21">

but that's not all that important right now, and the fact that the
Guidelines, as they stand, give you the choice seems to me to be a
testament to them.)

But if it were not so heavily encoded, as follows:

  <sp who="Seyton">
    <l part="Y">The queen, my lord, is dead.</l>
  <sp who="Macbeth">
    <l part="Y">She should have died hereafter;</l>
    <l part="N">There would have been a time for such a word.</l>

you would have to insert new markup (in P4 land) in order to point to
it from the <note>:

  <sp who="Seyton">
    <l part="Y">The queen,
       my <anchor type="noteTarget" id="soiS1"/>lord, is dead.</l>
  <sp who="Macbeth">
    <l part="Y">She should have died hereafter;</l>
    <l part="N">There would have been a time for such a
       word<anchor type="noteTarget" id="soiE1"/>.</l>

  <note type="editorial" resp="DP" target="soiS1" targetEnd="soiE1">

As I said, I don't really see how that markup presents a problem.