Name the Non-Standard PC Code Page

Out of all the 8-bit character encodings commonly seen on PCs and clones, there's one which IBM has never properly standardized or bestowed with a code page designation - although it seems to have come from IBM itself, and was quite popular for a while, at least in certain quarters.  A common term for it seems to be "Denmark/Norway" or "Nordic", but there's a little more to it than just that.

At IBM

On most CGA and MDA boards as shipped by IBM, the character ROM is numbered either 5788005 or 6359300, and holds the familiar fonts for code page 437 - the PC's original character set.  But some were sold with the more unusual 4733197 ROM, which modifies the CP437 reper­toire with a dozen new chara­cters (shown below in red):

IBM 4733197 CGA/MDA character ROM

MDA font, chars 80-FF

CGA font (default), chars 80-FF

CGA font (thin), chars 80-FF

This particular chip came from an MDA card (see the first image here for a full photo); but like the more common ROM variants, it also contains the two CGA fonts, so it could have been used with either type of adapter.  Judging by the date on the chip, IBM boards with this UK-made ROM were produced at least as early as 1985, and probably quite a bit earlier.  IBM sold these adapters in some European countries - Norway and Denmark for sure, although the range of those replace­ment glyphs hints at other countries as well.

See, only ø Ø have anything much to do with Danish and Norwegian.  Ŀ ŀ are Catalan, õ Õ ã Ã are Portuguese (õ/Õ exists in Estonian too), ʼn is Afrikaans, and ℓ ³ ¤ aren't specific to any particular locale.

And still, this character set is commonly described as "Danish", "Norwegian", or "Nordic".  Some sources (hi MAME!) go so far as to identify it with code page 865, IBM's official Nordic encoding for DOS - but looking at these two side by side, they're decidedly not the same:

It's worth keeping the timeline in mind here: code page 865 came along with PC-DOS 3.3, in April 1987.  Our mystery "DK/NO" layout is definitely earlier, having been around for at least a couple of years by then.  And the connection is evident: CP865 only retained a measly three characters to set it apart from CP437, but those three maintain the exact same code points they had in the previous scheme.

As for those other characters, they turned up more or less where you'd expect them to be in the DOS code page system, except for ℓ and ʼn - they somehow wound up in the none-too-common CP853 (Latin-3 Southern European, or Turkish, depending on who you ask).  As a sensible home for these two, it's not much of an improvement over Danish/Norwegian; they're clearly the odd ones out, since the rest of that code page was chosen to match ISO 8859-3.  But there was probably a reason for shoehorning them in there together, even if they didn't get to keep their code points.

Back to our anonymous character set, the real puzzle in all this is the lack of any definite sources from IBM itself about it: no official name, designation, or IBM code page number; no information as to its actual purpose; no mention in any documentation that I'm aware of;¹ and no apparent support in any known version of PC-DOS - even though they clearly shipped MDA and CGA boards with this character set, so (in text mode) it was all you had.  It's all very weird.

Third-Party Adoption

IBM may not give us much to go on, but quite a few makers of com­patible machines (especially ones sold in Europe) felt that this character set was important enough to imple­ment.  They're far less stingy with their infor­mation, and they all stick to the 'Denmark/Norway' version of the story - despite all of those patently un-Nordic characters:

Sharp's manuals for those two portables lay out the full character table in print.  Elsewhere on the list, command and file name abbre­viations like "DAN" and "NO" are pretty unambi­guous about which languages they refer to.

The common thread here is that these compatibles all predate PC-DOS 3.3 with its standardized code pages, including 865.  Later machines - to the degree that they still used hard-wired ROM fonts - offered CP865 for 'Nordic', for instance Amstrad's PC line.  But with these early examples, the only IBM PC character sets they had to follow were CP437 and this so-called "Denmark/Norway" one; some of the above were already selling by 1984, which tells us that IBM was likely using that scheme since 1983 (when PCs began shipping in Europe).

NCR PC-series color adapters, chars 80-FF

Acer DOS 3.20 GRAFTABL, chars 80-FF

DEC VAXmate, chars 80-FF

Olivetti GO467 video BIOS, chars 80-FF

Perhaps this character set's relatively popular acceptance can explain why it kept popping up at other places, where we can continue catching glimpses of it:

• In SCO's Xenix-derived UNIX System V/386, it's available as "PC Nordic", with console fonts in three sizes.¹¹ (Incidentally, the fonts themselves appear to be exact duplicates of the above-mentioned Olivetti GO467 fonts, at least in some versions.)

• HP defined it in its PCL (Printer Command Language) as "PC-8 Danish/Norwegian"; other adopters of this standard refer to it as "PC-8 D/N" or even "Code Page 437N"(!).

Which goes to show that even outside the world of DOS-based PCs, various vendors recognized it as a PC-derived de-facto standard of sorts and continued to support it, despite IBM's attempts to kill it off.  (Shout out to Lys for helping out with extra sources here and below.)

Back at IBM: Code Page 1058?

I wrote that IBM never officially documented this character set anywhere, but there's one place where IBM nearly seems to acknowledge its existence, and on first glance it looks like a slam dunk: code page 1058.  Could this be the admission of paternity we were hoping for?

Alas, not so fast: the original IBM PC version of this code page had the 'house' character ⌂ at 0x7F, but IBM's CP1058 has ▒ (a checkered/dotted box), which is more typical of HP character sets.  As it happens, the following lists do place this IBM code page in a sequence derived from HP - 1050 through 1058:

• IBM's list of single-byte code pages identifies CP1058 as "H-P Emulation, IBM-DN".

• Aivosto's comprehensive list calls it "HP IBM-DN", and pairs it with the HP version we've seen above: "HP PC-8 D/N (437N)"; it also adds, "IBM and HP versions are somewhat different".

The (only) difference is at 0x7F: HP has the IBM-derived 'house' character ⌂ (also 'home plate' in HP parlance), while IBM code page 1058 - supposedly emulating HP - replaces it with an HP-style checkered box... which HP never had there in the first place.

Things went roughly like this: first, IBM comes up with its odd 'DK/NO' encoding for the PC, in its 4733197 CGA/MDA ROM; then, HP adopts it as-is, officially recognizing it as a character set derived from the IBM PC; finally, IBM decides to port it right back from HP, implying that it was never theirs in the first place - and adding a single pointless change, just for good measure.

Way to pass the buck back and forth, guys!  It's as if IBM simply couldn't bring itself to assume responsibility for the ill-begotten alphabet soup they had cooked up.

The Billion-U+0024 Question(s)

At this point, we're still mostly in the dark about this cryptic little (non-) standard character set.  I'd be nice to be able to clarify a few points about IBM's original version of it:

• What was IBM (UK) going for with that modified CGA/MDA ROM?  Given all those non-CP437 glyphs which don't belong in Danish or Norwegian, which markets was it intended for, and which ones did it actually end up in?

• With all those non-DK/NO characters, why do practically all non-IBM sources refer to this charset as "Denmark"/"Norway"/"Nordic"?  Did Catalonia, South Africa, and the Lusosphere join up with Denmark and Norway, in some sort of Kalmar Union on steroids?  Or does this simply tell us where IBM's modified ROM was really used in practice?

• Did IBM ever have an official designation for this character set?  Is it described in any of IBM's PC-series documentation?

• Once this character set had achieved a few years of use on the IBM PC (and had become enough of a de-facto standard to be cloned profusely), why did PC-DOS 3.3's new code page system drop it like it's hot?  And if it wasn't worth keeping around, why introduce a new Denmark/Norway variant (code page 865) which covers only three of its characters - ø, Ø, ¤?  Especially when they already had those three covered in code page 850 (DOS Latin-1)?

If anyone happens to know more, I'd love to try and make some sense of it, so any official information from the IBM side of things would be appreciated.

UPDATE (2024-10-06): Some answers to the above questions have turned up; see the next post - Missing IBM PC Localization Disks & ROMs.

This post brought to you by ʼn (U+0149 a.k.a. LATIN SMALL LETTER N PRECEDED BY APOSTROPHE) and by the spirit of the late Michael Scott Kaplan.