[Gs-devel] refined MDRC patch (still including "relpos")

[mpsuzuki at hiroshima-u.ac.jp]

Dear Mr. Melichev,

About the "relpos" feature - today I received a comment from
Ken Lunde (the greatest authority of multibyte-handling in
PostScript, at Adobe).

Here I quote:
-----------------------------------------------------------------------
Message-Id: <200105261708.KAA10215 at blowfish>
To: mpsuzuki at hiroshima-u.ac.jp
Subject: Re: CID "regime" specification for notdefrange & cidrange
Date: Sat, 26 May 2001 10:08:01 PDT
From: "Ken Lunde" <lunde at Adobe.COM>

>Such "regime" specification is permitted in notdefrange & cidrange?
>Recent "Identity-H" uses <0000> <FFFF> for cidrange specification,
>but it can be extended in more kinky style?
>For example,
> 	1 begincidrange
> 	<0000> <7F7F> 0
>	endcidrange
>(<0000> - <007F>, <0100> - <017F>, ....) is permitted?
>
>I did several experiments on OKI PS printers, DPS on Solaris 8,
>and Acrobat Distiller. All CPSI accepted this kinky cidrange
>specification, and calculated CIDs were same.
>
>Nothing to say, I don't know any font vendors made such kinky CMap.
>But, for correct understanding of CID & CMap technology,
>I want to know this feature is "intended" design, or
>"unintended" behaviour which can be removed in future CPSI.

There are two "modes" for CID-keyed fonts. One is compatibility mode,
and this technique cannot be applied, because this mode does not
support such syntax. Conventional printers that use the CSL use this
mode.

The other mode is called native mode. Most CPSI interpreters, Acrobat
Distiller, and other devices use this mode.

Unless you can guarantee that the CMap file will never be used in
compatibility mode, I recommend that this syntax be avoided.

>Yet I could not find crucial statements in Adobe TechNotes.
>It's told that old ATM-J cannot parse such "regime", but
>does it mean "later ATM-J and other CPSI accept it"?
>If you know any documentation about this feature, please let me
>know.

ATM will never recognize this syntax (regime).
-----------------------------------------------------------------------

>From his comment, the "relpos" behaviour is intended design
on CID-native CPSI (am I misunderstanding?), but un-recommended
to use, because of the compatibility issues. Possibly this is
why official CMap does not promote this function.

Now I think including "relpos" feature is not so bad idea.
But almost nobody will write CMap using "relpos" intentionally,
so giving some message before accept (or disabling by default)
is better, I suppose.
Mr. Melichev, and other GS-developers, how do you think about?

Best Wishes,

suzuki toshiya