In encoding an edition of a Sanskrit text in verse, I need to introduce 
two kinds of reference mechanisms to folio- and page-numbers of Sanskrit 
manuscripts and Tibetan translations. Both references have the structure 
"14a2", with "14a" representing the folio- and "2" representing the 
line-number.

1) References to witnesses which contain the complete Sanskrit verse 
text, and only the verse text, or a Tibetan translation thereof: these 
can easily be inserted using <pb/> and <lb/> notation, e.g. <pb n="14a" 
ed="PV-1"/>, <lb n="4" ed="PVt-D"/>.
2) References to witnesses of works which only incorporate individual 
verses, or incorporate all verses as part of a prose commentary: these 
cannot be entered using a "break"-notation since:
(a) what matters is not some form of break within the text, but the 
location where a verse begins and ends in the witness,
(b) there may not even be a break to report. For instance, a Sanskrit 
manuscript which contains a commentary may contain one verse within one 
and the same line, in which case a "break"-notation is inapplicable.
(c) such witnesses contain the text typically in non-contiguous form. 
For instance, verse 367 may cover 14a2-3 in witness PV-1, and verse 368 
may be found at 15b1.

What approaches are meaningful for markup of references of type 2, which 
might more appropriately be described as the incorporation of material 
from this text into other works? Semantically, this seems like a reverse 
of <q> or <quote>, i.e. not material which is attributed to an external 
agency or marked as ostensibly quoted, but material from the encoded 
work which other works attribute to an external agency  by, for 
instance, adducing it verse by verse before commenting upon it.

Note also that these other witnesses are not available in digitised form 
and hence cannot be referenced through reference mechanisms to other files.

I thought of using nested <span> attributes, but am not sure where to 
fit in the identifier of the witness, and at any rate, it does not seem 
to be consistent with the semantics of <span> in P5 as marking the span 
targeted by some sort of analysis or interpretation. One could argue 
that location information is some sort of analysis, but then the 
analysis given and the span covered by it would be identical. 
Intuitively, this seems odd, but I can't at the moment think of an 
appropriate element in TEI P5 for such cases.

To clarify what is needed, here's an example, using <span>:

<lg n="367">
<span from="13a1" to="13a2" type="PV-1">
<span from="152b2" to="152b2" type="R-t-D">
<l>
<seg type="foot" n="a">
grāhākākārasaṅkhyātā
</seg>
<seg type="foot" n="b">
paricchedātmatātmani
</seg>
<seg type="foot" n="c">
sā yogyateti ca proktaṃ
</seg>
<seg type="foot" n="d">
pramāṇaṃ svātmavedanam
</seg>
</l></span>
</span>
</lg>

Thanks in advance, and best regards,

Birgit Kellner