Using a CMap with a non-embedded font

Question

I would like to (for example) use MS Arial without embedding it. I can do this just fine if I use the Arial internal CIDs with /Encoding /Identity-H, but that is not what I want to do - I want to use /Encoding 123 0 R to refer to a CMap so that I can use Unicode code points in my text like this: (^@H^@e^@l^@l^@o^@ ^@W^@o^@r^@l^@d^@!)Tj<br /><br />I have the CMap working perfectly if I embed the font. However, if I do not embed the font, I get an error saying that Acrobat "Cannot find or create" the font. As noted above, the external font works OK if I use /Identity-H instead of my CMap. Is it simply not possible to use a CMap with an external font, or am I missing some vital step?<br /><br />Here's a simplified excerpt from my PDF document:<br /><br /><code><br />100 0 obj<br />&lt;&lt;<br />/Type /Font<br />/Subtype /Type0<br />/BaseFont /Arial,Bold<br />/DescendantFonts [ 101 0 R ]<br />/Encoding 200 0 R<br />%/Encoding /Identity-H<br />/ToUnicode /Identity-H<br />&gt;&gt;<br />endobj<br /><br />101 0 obj<br />&lt;&lt;<br />/Type /Font<br />/Subtype /CIDFontType2<br />/BaseFont /Arial,Bold<br />/Encoding 11 0 R<br />/FontDescriptor 102 0 R<br />/CIDSystemInfo << /Registry (Adobe) /Ordering (Identity) /Supplement 0 >><br />/DW 1050<br />&gt;&gt;<br />endobj<br /><br />102 0 obj<br />&lt;&lt;<br />/Type /FontDescriptor<br />/Ascent 1069<br />/CapHeight 0<br />/Descent -271<br />/Flags 32<br />/FontBBox [ -1000 -308 2231 1077 ]<br />/FontName /Arial,Bold<br />/ItalicAngle 0<br />/StemV 0<br />/AvgWidth 442<br />/MaxWidth 1000<br />/MissingWidth 1000<br />&gt;&gt;<br />endobj<br /><br />200 0 obj<br />&lt;&lt;<br />/Type /CMap<br />/CMapName /UniAC-UCS2-8-H<br />/CIDSystemInfo<br />&lt;&lt;<br />/Registry (Adobe)<br />/Ordering (Identity)<br />/Supplement 0<br />&gt;&gt;<br />/WMode 0<br />/Length 0<br />&gt;&gt;<br />stream<br />%!PS Adobe 3. 0 Resource CMap<br />%% DocumentNeededResources: ProcSet (CIDInit)<br />%%IncludeResource: ProcSet (CIDInit)<br />%BeginResource: CMap (UniAC-UCS2-8-H)<br />%Title: ( UniAC-UCS2-8-H Adobe Identity 0)<br />%% EndComments<br />/CIDInit /ProcSet findresource begin<br />12 dict begin<br />begincmap<br />/CIDSystemInfo<br />3 dict dup begin<br />/Registry (Adobe) def<br />/Ordering (Identity) def<br />/Supplement 0 def<br />&gt;&gt;<br />end def<br />/CMapName /UniAC-UCS2-8-H def<br />/CMapVersion 10. 001 def<br />/CMapType 1 def<br />/UIDOffset 950 def<br />/XUID [ 1 10 25343] def<br />/WMode 0 def<br />1 begincodespacerange<br /><0000> <FFFF><br />endcodespacerange<br />1 beginnotdefrange<br /><0000> <001F> 0<br />endnotdefrange<br />1 begincidrange<br /><0020> <007e> 3<br />endcidrange<br />endcmap<br />CMapName currentdict /CMap defineresource pop<br />end<br />end<br />%%EndResource<br />%%EOF<br />endstream<br />endobj

Answer 1

You asked "why can't you use a variant encoding" - so I'd like to explain.
The reason that Acrobat fails is that for an unembedded CIDFont, if the CIDSystemInfo is not one of the predefined CID collection (Adobe-Japan1, Adobe-CNS1, Adobe-GB1, Adobe-Japan1), we cannot reliably provide font substitution. We have to consider the situation where the PDF is open on a system without the (same) referenced font. For the PDF to be useful, the
exact same font must be present on all systems, otherwise we lose the Portability in PDF.
There are many versions of "Arial,Bold" with different number of glyphs and different glyph IDs assignment. The embedded CMap has the following mapping:
<0020> <007e> 3
In order to point to the same GIDs, all "Arial,Bold" must have glyph index 3 for U+0020, GID 36 for U+0041, and so on. There is no such guarantee in TrueType spec.
Hope that explains the situation.
Leonard