[XeTeX] c.t.t. post re: \mag

[Jonathan Kew]

On 19 Feb 2008, at 2:12 pm, Andreas Matthias wrote:

> Jonathan Kew wrote:
>
> [ \mag ]
>
>> I have just checked in corrections to both xetex and xdvipdfmx, and I
>> believe this stuff is now working correctly.
>
> I just spotted that the left and top margins (the origin of TeX's
> coordinate system) are not scaled by \mag. This is in contrast to
> pdftex. I'm not sure which one is correct.

Hmmm. Neither am I.

Experimenting with plain TeX + dvips + ghostscript, it seems that  
\mag does *not* affect the top and left margins; the origin of a \mag- 
nified page remains at (1in, 1in) from the corner of the sheet.

With pdfTeX, on the other hand, the origin does move according to  
\mag, as you noted.

So there seems to be a discrepancy or incompatibility between the  
existing engines/paths already. Let the user beware!

Thinking it over, I don't recall a completely explicit statement in  
The TeXbook about the relationship between \mag and the page origin,  
but I think we can deduce that Knuth did not expect it to shift, if  
we consider his \magnification macro (in plain.tex). This changes the  
\mag parameter so as to scale fonts, skips, etc., and then explicitly  
resets \hsize and \vsize using "true" dimensions so that the text  
area remains 6.5" x 8.9", but it does NOT touch \hoffset and  
\voffset. Nevertheless, The TeXbook tells the reader: "There will be  
an inch of margin at both left and right, assuming that the paper is  
8 1/2 inches wide." (p.60) This is only true if the default origin is  
unaffected by \mag.

JK