Of page-breaks and TEI
 
I have been having some small struggles with encoding page
breaks in various 18th and 19th century printed books (they
allowed me off manuscripts for this one).  Take the case of pages
346-7 of the reprint of Hannah More's "Village Politics" in her 1818
"Collected Works".   At the base of p. 346 against the right margin
we have the catch-word "Tom."; at the top of the next page we have
the page number 347 against the right margin and a running
header "Village Politics", centred and in small caps; at the base of
this page we have the signature "Q6", in a slightly smaller type
face than the main text and centred and the catchword "merly".
 
So far as I can see, P1 suggests capturing all this information
within the "page.break" tag.  Adapting the example on p. 125 of P1,
you might code this page break between pages 346 and 347 in
Hannah More:
<page.break n=347 ed=1818 sig=Q6 catchword="merly">
There seem to be a few problems in this.  Firstly, you would not
know, just from looking at this coding, whether the 347 referred to
the preceding or following page.  In fact, P1 states clearly (on p.
125) that the page.break must be placed "at the start of every new
page", and so the 347 refers to the following page.  This is fine if
one knows the documentation, but it would seem better if the
page.break mechanism were so designed that one could tell from
its use alone just which page was which, and not rely on everyone
knowing and obeying the administrative fiat by which P1 declares
that page breaks do not come at the end of pages, nor between
them, but at their beginning (I believe some time in P1's dark past
there was an intense theological debate over just where page breaks
happened.  Oh to have been there).  Actually, it would make better
sense if P1 called this element "new.page" (which is what it is)
and not "page.break".
 
There is another problem with the dictum that page-breaks come at
the beginning of the page.  According to P1, this means that catch-
words and signatures (which appear at the bottom of the page) must
be captured in the page.break tag which is placed at the top of the
page.  Thus, the catch-word "merly", which appears at the bottom
of the page and connects this page to the next, is bundled with the
page.break tag at the top of the page.  This might suggest to the
unwary reader that this catchword actually connects this page to
the last, not this page to the next.  It certainly seems counter-
intuitive, to me at least.  The P1 page.break mechanism also gives
no indication as to where in the text the page number actually
appears: whether at the bottom or top of the page, right or left or
centred, surrounded by printer's ornaments, etc.
 
Yet another problem is the insistence, it seems, that the text of all
these (page numbers, signatures, catch-words) must be captured as
attributes of all-powerful page.break.  What of the case where one
wishes to include tags within the text of (say) a catch-word?  A
catch-word might contain an interesting abbreviation which one
might wish to register with <abbrev> or <expan>; it might be
italicized (as in the catch-word "Tom." at the bottom of p. 346); one
might wish to attach a note to an unusual signature; formatters will
find it more difficult to extract the text of the attributes and display it
correctly.
 
The greatest problem looks to me to be the running header,
"Village Politics".   There seems no place for this in the page.break
mechanism, or anywhere else in P1 for that matter.  Given that
running headers can be very important in all sorts of ways we
need them.  Dickens delighted in clever running headers, a sort of
marginal commentary on his text, and I seem to recall a nice
Randall McCleod (I think) article in which he drew elegant
conclusions about the printing of Shakespeare's Sonnets from the
changing spacing in the running headers from compositor to
compositor.  We need running footers too, for that matter.  Indeed,
it would be rather difficult to encode the page.breaks in P1 with P1'
s own mechanisms: P1 has BOTH running headers and footers!
 
So now for some mild suggestions.  Instead of a single page.break
tag, I suggest we have a small swag of tags, all grouped within a
page.break element.  Thus:
        A page break is defined as occurring not at the beginning or
end of a page but between two pages;
        A page break may contain two elements, <footer> and
<header>, in that order.  The first page.break of a text might contain
only a <header>, the last only a <footer>;
        As well as the running text of headers and footers, <footer>
and <header> may also contain the elements <catchword>,
<signature> and <page.number>;
        The page.break element may have an attribute which gives
the number (etc) of the page immediately following (as it does in
P1).  In computing terms this makes excellent sense, rather than
relying on applications extracting the number from within the
<header> or <footer>: this would let a one-pass system sweep
through and marry text and page numbers in a single run.   In
contexts where one is not interested in headers and footers but just
wants to record pagination one could then code page breaks
<page.break n="123"> , just as P1 proposes.  But where you want to
pack in lots of information about headers, text and layout features,
footers, etc, the system here proposed will let you do that.
 
So here are the page breaks between 346/347 and 347/348 of
Hannah More encoded as this suggests:
<page.break n=347>
        <footer><catch.word>Tom.</></footer>
        <header>Village Politics. <page.number>347</></header>
</page.break>
 
and
<page.break n=348>
        <footer>
                <signature>Q6</>
                <catch.word>merly</>
                </footer>
        <header>Village Politics. <page.number>348</></header>
</page.break>
 
Alternatively, if one is not interested in headers, footers, etc, you
could just have:
        <page.break n=347>
and
        <page.break n=348>
 
One could even encode the page break between pages 124 and 125
of P1 with this!
 
Peter Robinson
Oxford University Computing Service