[metapost] Conformance to Document Structuring Conventions Specification ver 3.0

[Michail Vidiassov]

Dear All,

Running mpost test.mp, where test.mp is as follows:

beginfig(1);
prologues := 1;
label("test" infont "Times-Roman" scaled (bp/pt), (0,0) );
endfig;
end

results in the following output:

%!PS-Adobe-3.0 EPSF-3.0
%%BoundingBox: -7 -3 7 3 
%%Creator: MetaPost
%%CreationDate: 2006.02.11:1052
%%Pages: 1
%%DocumentFonts: Times-Roman
%%BeginProlog
  /Times-Roman /Times-Roman def
/fshow {exch findfont exch scalefont setfont show}bind def
%%EndProlog
%%Page: 1 1
-6.94502 -2.85501 moveto
(test) Times-Roman 10.00005 fshow
showpage
%%EOF

The first line of the output implies that output is intended to conform to
PostScript Language Document Structuring Conventions  ver 3.0,
as described in 5001.DSC_Spec.pdf from Adobe.
But conformance is not perfect. 
(It looks like DSC ver 1 with %%EOF from ver 2.1 and %%BeginProlog from 3.0,
see part "A.1 Changes Since Earlier Versions" in the DSC spec mentioned above.)

Errors:
1. There should be %%EndComments line after %%DocumentFonts.
  In part 4.1 of DSC ver 3.0 doc it is stated that:

> The header section consists of DSC comments only and describes the 
> environment that is necessary for the document to be output properly. 
> The end of the header section is denoted by the %%EndComments comment

2. Prolog section used incorrectly.
  In the same part 4.1 there is this text:

> The beginning of the procedures section is indicated by the %%BeginProlog 
> comment. This section is a series of procedure set (procset) definitions; 
> each procset is enclosed between a %%BeginResource: procset and 
> %%EndResource pair. Procsets are groups of definitions and routines 
> appropriate for different imaging requirements.
> The prolog has the following restrictions:
>  Executing the prolog should define procsets only. For example, these 
> procsets can consist of abbreviations, generic routines for drawing 
> graphics objects, and routines for managing text and images.

The proper place for other initialization commands is in the setup section.
  Part 4.2 of the specification tells:

> The document setup section is denoted by the %%Begin(End)Setup 
> comments. The document setup should consist of procedure calls for 
> ..., running initialization routines for procsets, ...

Thus the the definition of fshow is to be wrapped with
  %%BeginResource: procset / %%EndResource pair,
while "/Times-Roman /Times-Roman def" has to go to the setup section,
whose bounds are %%BeginSetup / %%EndSetup.

IMHO, what now is

%%BeginProlog
  /Times-Roman /Times-Roman def
/fshow {exch findfont exch scalefont setfont show}bind def
%%EndProlog

has to be changed in one of the following ways:

a) Prolog is used properly

%%DocumentProcSets: mpost
%%DocumentSuppliedProcSets: mpost
%%EndComments
%%BeginProlog
%%BeginResource: procset mpost
/fshow {exch findfont exch scalefont setfont show}bind def
%%EndResource
%%EndProlog
%%BeginSetup
  /Times-Roman /Times-Roman def
%%EndSetup

b) Prolog is not used at all,
%%BeginSetup / %%EndSetup pair is used instead of %%BeginProlog / %%EndProlog.

%%EndComments
%%BeginSetup
  /Times-Roman /Times-Roman def
/fshow {exch findfont exch scalefont setfont show}bind def
%%EndSetup

3. DocumentFonts is used according to DSC spec ver 1

DSC spec ver 3 doc told me:

> %%DocumentFonts: should be the union of the %%DocumentNeededFonts: 
> and %%DocumentSuppliedFonts: font lists. 
> Note This comment is provided for backward compatibility and may be 
> discontinued in later versions of the DSC. Use the more general comments 
> %%DocumentNeededResources: and %%DocumentSuppliedResources: 
> instead.

Thus %%DocumentNeededFonts comment is to be added, or
%%DocumentFonts is to be replaced with %%DocumentNeededResources.

4. Fonts are not reencoded. (Not a bug, but a missing feature.)
  Although psfonts.map is used to get PostScript names of fonts,
  infomation on font encoding from the same file is ignored.
  It will result in errors with non-ascii characters from TeX fonts,
  that are usually in some TeX encoding.
  IMHO font reencoding must be added to metapost output.

          Sincerely, Michail