I never used Jane myself back in the day, or for that matter any 128 native word processor, and even when we got a 128 I still wrote my term papers in Pocket Writer 64 or Timeworks Word Writer. However, that faulty but repairable Australian Commodore 128DCR I got last Christmas came with a big box of software and in there was a complete copy of Jane 128 along with the data disk the previous owners' family had used. They were delighted when I said I wanted to take a whack at converting their files as a thank you — and along the way we'll take a look at Jane's oddball history, the original Apple II version, the Commodore 128 version and its all-but-unknown port to the French Thomson MO5, plus those other attempts at productivity applications Commodore tried in the mid-1980s.
At least one side project would become critical to the reworked division. Engineer John Feagans had been the initial developer of the PET Kernal, its internal ROM-based operating system, which pioneered the then-novel idea of separating I/O and other system routines from the built-in Microsoft BASIC. (Prior versions of BASIC on early microcomputers often simply drove devices directly.) Feagans' innovation facilitated the use of these routines by user-written machine language programs through a standardized jump table, which was reused and greatly expanded by Bob Russell for the VIC-20 and 64.
Feagans' side project was a PET demonstration of a small picture-based file manager, using an office filing cabinet as its central metaphor and drawn with PETSCII graphic characters. In 1982 the Commodore office in Moorpark, California where Feagans worked was closed and he was reassigned to their then-executive offices in Santa Clara. With little else to do at the time, Feagans rewrote his file cabinet demo for the new 64 to get familiar with the hardware, primarily in BASIC. As an exploration of the 64's joystick and moveable sprite capabilities, he additionally implemented an animated "hand" controlled by the joystick that served as a pointer. By this time Hartmann had just weeks until Summer CES in June 1983, the deadline he had promised Tramiel. While working on licensing deals with outside firms, Hartmann also ordered Andy Finkel, promoted from the VIC team to technical manager, to find anything the software division was already working on that could get finished fast. Finkel saw Feagans' new demo and convinced Hartmann that it was both feasible and viable as a product.
Part of this cost was the expense of manufacture. Magic Desk initially came as 32K of ROM in four 8K chips, quite possibly the largest cartridge developed for the 64 at that time and requiring additional logic on the PCB for bank switching. (Its design is still used to this day for large multicarts, now supporting as much as a megabyte of ROM.) Internally, although some code had been converted to machine language using a custom compiler, there wasn't enough time to do it all and a fair portion of Magic Desk thus remained written in BASIC (moved up to start at $0a01 instead of $0801). As part of initialization this BASIC program is read out of the ROMs and copied to RAM for execution, which is the slight pause at the beginning before the title screen appears.
Despite the cool press reception, Magic Desk sold surprisingly well to new 64 users, and in numbers sufficient to solve its production problems such that its price dropped in half by late 1984. Sales were enough to propose an upgraded version called Magic Desk II for Commodore's new (and, ultimately to its detriment, incompatible) 264 series, capable of speech prompts when paired with the speech group's Magic Voice synthesizer hardware, and featuring a more sophisticated interface with "Lisa-like" pulldown menus and icons.
As Hartmann, Tomczyk and others in the software group felt the 264 would be most meaningful with built-in software, Magic Desk II became one of several possible option ROMs representing potential machine configurations, along with others featuring different programming languages or applications like a spreadsheet or terminal program. Tramiel embraced the idea, seeing it as a car with different models "just like General Motors." At least three "flavours," aligning with the internal groups in the software division (business, home and education), were envisioned — though the concept also met swift resistance from dealers and even Commodore's own sales team who protested having to handle multiple different versions of one computer. (One possible unspoken reason is that it was already known not to be 64 compatible, and dealers were undoubtedly unhappy about taking stock space away from a financial anchor.)
In January 1984 Jack Tramiel was forced out of the company he founded by chairman Irving Gould and shortly thereafter the 264 line was slashed by new CEO Marshall Smith. Three models were chosen from the various prototypes and experiments, the 16K 16 (and in Europe the 116 as well) to serve as the low-end, and the 64K 264, now renamed the Plus/4, as the high-end. Both the completed Magic Desk and Magic Desk II were cut, as was the education flavour based on a built-in Logo implementation. Only the Plus/4 would offer Trilogy, renamed to 3-Plus-1, and its ROM size was cut to 32K to reduce production costs further which in turn forced Johnson to make serious reductions in functionality. As a result 3-Plus-1 became nearly as maligned as the Plus/4 itself (mocked in the press as the "Minus/60" for its idiosyncrasies, deliberately incompatible ports and lack of 64 software compatibility), though yours truly actually used the spreadsheet for a household budget when I was a starving student, and it wasn't that bad.
The failure of the Plus/4 to succeed the 64, much less overtake it, demonstrated clearly to Commodore management that 64 compatibility was essential in their next computer. Around this same time Sig Hartmann had noted the success of Atari's own first-party software unit Atarisoft on non-Atari platforms (though AtariLab was developed externally), even for the 64, and said there were similar plans to port Commodore first-party software and licenses to the Apple II, IBM PC and PCjr. (NARRATOR: This didn't happen.) However, by mid-1984 the software division had also developed a reputation for poor quality, with Scott Mace commenting in InfoWorld that — hits like International Soccer notwithstanding — "so far, the normal standard for Commodore software is mediocrity." In the meantime, Tramiel had bought the ailing Atari from Warner Communications and lured several Commodore managers, including Hartmann (who was already in conflict with Smith over cuts to the software division), to defect. Paul Goheen, the former systems software group head, became the new software chief.
As the Commodore 128 neared completion, management constructed a new sales strategy to put it head to head against the Apple IIc and PCjr in specialty stores as well as Tramiel's favoured mass market retailers. To more plausibly bill it as a productivity machine, once again the software division had to look outside the company.
The initial release was uniquely accompanied by its own single-button mouse (Instagram link), a lower-cost unit — Kotick admitted settling for "a lesser quality product" — manufactured under contract by joystick maker Wico and incompatible with Apple's own later mouse options. This mouse interfaced to the Apple with a custom card and an 8-pin connector.
The top of the screen shows the editing tools (hand pointer, arrow, scissors [cut], camera [copy] and paste jar), icons for the three core apps, and then system-wide icons for on-line help, printing, the file manager, preferences and "STOP" to globally cancel an operation. Across the bottom, beside the window maximize gadget, are tools for adjusting line justification, font style (bold, underline, bold and underline, superscript, subscript), print settings and search. I'll have more to say about all of these when we get to Jane's Commodore port.
(Parenthetically, a frequent misconception is that the Z80 in the 128 does the boot sector check to start CP/M. In fact, the 8502 does that after the Z80 has already checked the Commodore key isn't down and entered 128 mode. In CP/M's case, the code in the CP/M disk boot sector instructs the 8502 to give control back to the Z80 so it can finish the boot sequence — if this weren't the case, the BASIC BOOT command wouldn't be able to start CP/M. Interestingly, 1581s booting CP/M have a special startup file to keep their own CP/M boot sector in a different location.)
The manual, which appears to have been written by a third party, insists on camelcasing the apps as JaneWrite, JaneCalc and JaneList even though the rear box copy doesn't distinguish them that way and the Apple II version and Arktronics' own advertising called them Janewrite, Janecalc and Janelist. I'll use the latter here.
Both Jane and the Perfect series were demonstrated with the 128 for Winter CES 1985, though Commodore didn't announce MSRPs then for any of them at the time. It eventually emerged later that year for $49.95 [$150]; by the next year some retailers were selling it for as low as $35 [$100], compared to each of the Perfect titles then going for $45 apiece [$130]. Plus, the Perfect titles, being CP/M-based, were bland and text-based and failed to show off the C128's graphics, so Commodore ended up emphasizing Jane more in its contemporary marketing.
As with the Apple II, across the top of the screen are the various icons, or what the manual calls "pictures." These were slightly altered for the 128 from the Apple II originals. The blue icons again set the mode of the pointer ("tools") from the hand pointer, the insertion arrow, scissors (cut), camera (copy) and paste jar. The current tool is highlighted. You select a new tool by clicking on it, but you can also use the function keys (CLR/HOME, F1, F3, F5 and F7).
By modern conventions Jane's tools are used backwards: instead of selecting a text range and "choosing a tool" (e.g., Command-C for copy) to operate on it, you choose a tool first and with that tool select the text range it operates on. I'll demonstrate this specific functionality in a couple instances, but any verb-object operation in Jane will work this way. (Linguistically, this makes Jane a head-initial language, while most modern GUIs are head-final.)
Next to the tools are the main apps, Janewrite (now in purple), Janecalc (in green) and Janelist (in cyan), and next to them in grey are the same standard applets built into the kernel (online help, print dialogue, disk/file manager and setup). Finally, the STOP icon, now a solid red, stops the current app, and can be used to escape some screens, though not all. Unexpectedly it doesn't serve to quit Jane entirely: you just turn the computer off. Jane remembers what app is loaded and doesn't reload the overlay if you exit and re-enter it.
However, unlike Jane for the Apple II, Jane 128 does not allow you to have multiple documents open simultaneously, a limitation that challenges the definition of "integrated software." In fact, of the three official Commodore productivity packages we've looked at so far, only poor abused 3-Plus-1 could do so. If you're working in one app and select another, Jane 128 will prompt you to save your work as if you'd clicked STOP, and the window will close. Given that the Apple II version managed to implement multiple documents in 64K of RAM, the Commodore 64 version — let alone the 128 — would seem to have little excuse, though I can think of two potential explanations. One is to increase the amount of memory available for any one document, which the Apple II version was indeed criticized for. The second is particular to the 128: its default memory configuration doesn't have a lot of free RAM, and it may have been judged too complicated to span or swap working sets across banks. (Some fiddling in the monitor shows that the documents simply occupy RAM in bank 1 and don't span elsewhere.) On the other hand, other 128 applications certainly do manage it, and it's possible development deadlines were a contributing factor.
Let's start out with the online help, one of Marks and Kotick's fundamental design goals. Assuming you got the joystick (or 1350) plugged into the right port, there's a big honking question mark. What happens when you click on it?
The window's yellow border is salient: any window that gives you choices to select from carries a yellow border. We'll see this again later. Incidentally, the manual is not in colour (just black, white and an accent magenta), and possibly as a result is all but totally ignorant of these UI choices or why they were made.
The Apple II version's windows have a corner close gadget in the typical open square style, but this would only have made visual sense to someone familiar with the Lisa or Mac — which most Commodore users in 1985 wouldn't be. Instead, here windows you can close have a big bright red EXIT button (the ESC key can also serve this role in most cases). All icons are simply drawn onto the VIC-II high-res screen; unlike Magic Desk where pretty much everything was a sprite, the only sprite here is the pointer. Windows cannot be moved or resized, but by limiting icon and window boundaries to VIC-II 8x8 cells, drawing is fairly fast.
The disk icon brings up a basic file manager allowing you to make duplicate backup files (but only to the same disk), rename or delete them, or create a new disk. The copy option is the default and is pre-selected when the window opens, which is irritating, because our modern reflexive habit of selecting the file first and then the action will cause an immediate copy to occur before you get the chance to click anything else. I don't think this was just me: there were several spurious duplicate files on the original black data disk I suspect for the same reason. If files are present on the disk, they will appear with their names and a filetype letter (W, C or L for Janewrite, Janecalc or Janelist respectively). Notice the slightly misshapen scrolling arrows, which were nice and clean on the Apple.
Along the bottom are various operators and the extent of the formula functions available (hardly competition for the Convergent WorkSlate). They can be selected with the pointer in lieu of text entry, which makes them quite discoverable, but in practice it's simply faster to type them. No other functions are implemented. Since the content window cannot be resized, the maximize gadget from the Apple II version became obviously useless and was removed.
Even in 40-column mode, and even given that the scroll doesn't animate and just snaps to the new position, scrolling is agonizingly slow and multiple clicks on the scroll controls invariably overshoot. Putting it in 80 column mode doesn't just give you more on the screen — it also means you don't need to endure as many pauses trying to move around. Again, I'll have more to say about font sizes when we get to Janewrite.
Unique to Janecalc, the scrolling actually wraps. Worksheets remain limited to 26 columns (A-Z), as in the Apple II version, but now up to 50 rows are allowed. Still, this wouldn't have been sufficient for anything larger than a modest small business, even in 1985.
The two OK buttons do different things. The top one filters cards to the query requested, but the bottom one undoes any query. You can query on top of queries for a primitive sort of logical-AND. Partial word searches are supported.
The bottom shows left, right, centre and full justification options, done per line, plus the font options, and then search (this time with a magnifying glass) and page layout. These icons are larger and more detailed than the Apple's. As with other parts of the UI, text attributes are set verb-initial, i.e., you would choose the desired style, then the insertion tool, then highlight the range of text and finally release the button (or, annoyingly, wait, which makes fine adjustments harder than they should be). Notice that there is an explicit bold-and-underline option, instead of using the separate bold and underline options in combination: that's because you can't make superscripted or subscripted text bold or underlined.
The top of the content area shows the ruler — in characters, not inches — and margin stops and the single paragraph indent stop, which you can drag to change them. Interestingly, the ruler seems to be new for Jane 128: no screenshot for the Apple II version demonstrates one.
The bottom set of options are specifically for configuring the printer. Print output is generated with control codes and text only; Jane does not generate graphic sequences, so what actually shows up on the page depends on what your printer actually supports. The default middle print size option corresponds to 80 columns (more accurately, the default or 12cpi font size), while the small font corresponds to the condensed 15cpi and/or 132 column option available on some printers, and the large font corresponds to the 60 or 64-column (e.g., 10cpi) option also only available on some printers — not the expanded double wide printing option more typical of Commodore-specific devices.
You could also select between regular fanfold tractor feed or plain paper, as well as the Near Letter Quality print option some printers supported, and finally the printer itself. There was built-in support for the MPS-801 (i.e., 1525), MPS-802 (i.e., 1526), 1101, "Oki" (generic Okidata), and generic "Epson" printers, assuming they were connected somehow to the IEC serial such as through a third-party parallel printer interface. These printer drivers varied wildly in their capabilities: the 1525/MPS-801 "driver" supported almost nothing except basic text, while the generic Okidata and Epson drivers supported nearly everything that Jane could generate.
The "OTHER" driver is customizeable, a neat idea, if a little clumsy. A BASIC program called BUILDPRT on the grey disk allows you to populate it with your printer's control codes. Control sequences up to three characters long were supported. You would enter these from the manual, it would save them to disk, and then you could use the custom driver inside Jane.
The topmost options affect the appearance of Jane. You can choose from three font sizes, all non-proportional, corresponding to 40, 64 or 80 column text. Confusingly, the 80-column option is the last when it's the middle option for printing. (The Apple II version likely didn't support 80 column text due to its smaller hi-res horizontal width.) Pointer speed and the "beep" Jane makes when you select options are also configurable.
Now that we understand what Jane can do and how it works, let's move on to converting the files. We can extract our test files we made from our .d71 with VICE's c1541 utility and hexdump it. This got a little more difficult than it should have been:
% c1541 work.d71 c1541 (VICE 3.8) Copyright 1995-2023 The VICE Development Team. C1541 is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. c1541 #8> dir 0 "data#000 " 01 2a 39 "c.TESTCALC" prg 3 "l.TESTLIST" prg 4 "w.nOTE.2" prg 4 "w.TESTCONV" prg 1278 blocks free. c1541 #8> read "w.TESTCONV" ERR = 62, FILE NOT FOUND, 00, 00 cannot read `w.TESTCONV' on unit 8 invalid filename
As you may have guessed, that's because the filenames, and it turns out the data in the files themselves, are stored as true ASCII, not PETSCII (indeed as they would have been on the Apple II, also), so c1541's filename conversion fails. Fortunately, the extract command will walk the entire directory and pull out all the files no matter what they're named.
The documents are all regular Commodore PRoGram files with a starting address and they appear to be loaded in-place at that address into bank 1. This introduces the possibility of some delightful save-file hacks but we're not here for that today.
That means we can also simply extract our previous owners' files en masse as well. I imaged their disk using my ZoomFloppy and external 1571, and since they are normal PRGs, a BAM copy of the allocated sectors is all we need to do, with one wrinkle:
% d64copy -B 9 r_jane_bam.d64 [Warning] growing image file to 683 blocks 14: --*****--*****--***** 8% 15/171[Warning] read error: 0e/00: 5 14: ?******************** 12% 21/171[Warning] giving up... 14: ?******************** 15: ********************* 16: ********************* 17: ********************* 18: ******************* 19: ******************* 20: ******************* 21: ******************* 22: ****..***...***...* 100% 171/171 170 blocks copied. % c1541 r_jane_bam.d64 c1541 (VICE 3.8) Copyright 1995-2023 The VICE Development Team. C1541 is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. c1541 #8> extract c1541 #8> quit
Fortunately it doesn't look like the bad sector was in any of the files they cared about.
Let's get an idea of the format from our test files. Janecalc files are "huge" because it looks like the entire 26x50 cell range is saved regardless of whether anything's in it or not. Looking at the directory of their data disk where they had saved a couple of sparse test spreadsheets of their own, they also were 39 blocks in length no matter how many cells were actually populated. The on-disk format is consequentially a bit wasteful:
000003a0 00 00 00 00 00 20 20 20 4a 61 6e 75 61 72 79 20 |..... January | 000003b0 20 46 65 62 72 75 61 72 79 20 20 20 20 20 4d 61 | February Ma| 000003c0 72 63 68 20 20 20 20 20 41 70 72 69 6c 20 20 20 |rch April | 000003d0 20 20 20 20 4d 61 79 20 20 20 20 20 20 4a 75 6e | May Jun| 000003e0 65 20 20 20 20 20 20 4a 75 6c 79 20 20 20 20 41 |e July A| 000003f0 75 67 75 73 74 20 53 65 70 74 65 6d 62 65 72 20 |ugust September | 00000400 20 20 4f 63 74 6f 62 65 72 20 20 4e 6f 76 65 6d | October Novem| 00000410 62 65 72 20 20 44 65 63 65 6d 62 65 72 00 00 01 |ber December...| 00000420 04 00 00 00 00 00 00 20 20 54 6f 74 61 6c 20 20 |....... Total | 00000430 20 41 76 65 72 61 67 65 20 20 20 00 00 01 04 00 | Average .....|
Cardfiles are stored much more efficiently. Here's the entirety of our two cards. The fields are in the header, padded out to the fixed 12 character length, but with technically unnecessary nulls between them. A spurious "Beatles" entry looks like a buffer that was captured as part of the save, and there is also a page-aligned (counting the starting address) pad of nulls. Otherwise, the value strings look null-delimited, in the order they appear in the field list (unpopulated fields are zero-length strings). When you run out of fields, you're on the next card.
% hexdump -C l.TESTLIST 00000000 00 7f 01 00 36 02 08 00 04 00 54 69 74 6c 65 20 |....6.....Title | 00000010 20 20 20 20 20 20 00 4c 65 61 64 20 41 72 74 69 | .Lead Arti| 00000020 73 74 20 00 4f 74 68 65 72 20 41 72 74 69 73 74 |st .Other Artist| 00000030 00 43 6f 6d 70 6f 73 65 72 20 20 20 20 00 43 6f |.Composer .Co| 00000040 6e 64 75 63 74 6f 72 20 20 20 00 57 68 65 72 65 |nductor .Where| 00000050 20 4b 65 70 74 20 20 00 42 6f 72 72 6f 77 65 64 | Kept .Borrowed| 00000060 20 62 79 20 00 44 61 74 65 20 4c 65 6e 74 20 20 | by .Date Lent | 00000070 20 00 00 00 ff ff 01 ff ff ff ff ff ff ff ff ff | ...............| 00000080 ff ff ff ff ff 00 01 02 03 04 05 06 07 00 01 02 |................| 00000090 03 04 05 06 07 ff ff ff ff ff ff ff ff ff ff 00 |................| 000000a0 00 54 68 65 20 42 65 61 74 6c 65 73 00 00 00 00 |.The Beatles....| 000000b0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| * 00000100 d7 09 80 60 60 ac 98 09 ae 04 cf a5 ea a9 d7 15 |...``...........| 00000110 80 ac 1d 21 d6 1c 80 60 60 ac 98 09 ae 04 cf 20 |...!...``...... | 00000120 75 0b 00 00 ac 0a 95 a5 f0 b0 a9 a6 b1 a9 a4 b1 |u...............| 00000130 a5 f1 b0 a9 a7 b1 a9 a5 b1 ac c0 08 cf 20 75 0b |............. u.| 00000140 fe ff a5 37 ae 8c ff 00 c0 d8 4b 80 cf a5 f0 b0 |...7......K.....| 00000150 a7 a6 b1 c0 d8 60 80 a5 f1 b0 a7 a7 b1 c0 d8 60 |.....`.........`| 00000160 80 cf a5 f0 b0 a7 a6 b1 00 00 09 00 01 5d f6 00 |.............]..| 00000170 00 00 00 09 00 09 00 01 66 f6 00 00 00 00 54 00 |........f.....T.| 00000180 50 00 01 b1 f6 69 64 65 20 12 00 42 00 01 6f f6 |P....ide ..B..o.| 00000190 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| 000001a0 00 00 00 00 00 00 00 52 65 76 6f 6c 76 65 72 00 |.......Revolver.| 000001b0 54 68 65 20 42 65 61 74 6c 65 73 00 00 4c 65 6e |The Beatles..Len| 000001c0 6e 6f 6e 2d 4d 63 43 61 72 74 6e 65 79 2d 48 61 |non-McCartney-Ha| 000001d0 72 72 69 73 6f 6e 00 47 65 6f 72 67 65 20 4d 61 |rrison.George Ma| 000001e0 72 74 69 6e 00 00 00 00 00 41 74 6f 6d 20 48 65 |rtin.....Atom He| 000001f0 61 72 74 20 4d 6f 74 68 65 72 00 50 69 6e 6b 20 |art Mother.Pink | 00000200 46 6c 6f 79 64 00 00 57 61 74 65 72 73 2d 47 69 |Floyd..Waters-Gi| 00000210 6c 6d 6f 75 72 2d 4d 61 73 6f 6e 2d 57 72 69 67 |lmour-Mason-Wrig| 00000220 68 74 2d 47 65 65 73 6f 6e 00 50 69 6e 6b 20 46 |ht-Geeson.Pink F| 00000230 6c 6f 79 64 00 00 00 00 |loyd....| 00000238
This would be very easy to convert, and I just manually did this by dumping everything with strings for their one cardlist with one single record. Other things I can pick out are the file length minus the starting address at byte offset 4 (as a little-endian unsigned 16-bit quantity), and the value at byte offset 6 might be the number of fields, though we don't need either of those to simply grab the data.
Now for the Janewrite files, starting with our test document which should contain every sort of text ornamentation that Janewrite can represent. The first section looks like this.
% hexdump -C w.TESTCONV 00000000 00 88 01 00 21 00 00 00 46 00 05 00 47 00 34 03 |....!...F...G.4.| 00000010 00 20 00 20 00 20 00 20 00 20 00 20 00 20 00 20 |. . . . . . . . | * 00000030 00 20 00 20 00 20 00 20 05 00 04 02 0a 50 01 05 |. . . . .....P..| 00000040 3d 42 01 01 6e 81 61 6e 61 20 61 62 72 83 65 83 |=B..n.ana abr.e.| 00000050 20 61 6c 82 6f 82 20 61 6c 84 65 00 01 6e 00 61 | al.o. al.e..n.a| 00000060 00 6e 00 61 00 20 00 61 00 62 00 72 03 65 00 20 |.n.a. .a.b.r.e. | 00000070 00 61 00 6c 02 6f 00 20 00 61 00 6c 04 65 00 20 |.a.l.o. .a.l.e. | 00000080 00 20 00 20 00 20 00 20 00 20 00 20 00 20 00 20 |. . . . . . . . | *
Skipping the start address (this is indeed its location in bank 1), we again see some pieces of what looks like a buffer, as well as the file length (again minus the starting address) at byte offset 14. The left margin is (probably) at byte offset 6 because the right margin is at byte offset 8 (also a 16-bit LE short), followed by the paragraph indent. After this header, we now get the text, shown here in its entirety.
000000a0 00 20 00 49 74 20 77 61 73 20 61 20 64 61 72 6b |. .It was a dark| 000000b0 20 61 6e 64 20 73 74 6f 72 6d 79 20 6e 69 67 68 | and stormy nigh| 000000c0 74 2e 00 90 49 74 90 20 90 77 61 73 90 20 90 61 |t...It. .was. .a| 000000d0 90 20 90 64 61 72 6b 90 20 90 61 6e 64 90 20 90 |. .dark. .and. .| 000000e0 75 6e 64 65 72 6c 69 6e 65 64 90 20 90 73 74 6f |underlined. .sto| 000000f0 72 6d 79 90 20 90 6e 69 67 68 74 2e 00 88 49 74 |rmy. .night...It| 00000100 88 20 88 77 61 73 88 20 88 61 88 20 88 64 61 72 |. .was. .a. .dar| 00000110 6b 88 20 88 61 6e 64 88 20 88 62 6f 6c 64 88 20 |k. .and. .bold. | 00000120 88 73 74 6f 72 6d 79 88 20 88 6e 69 67 68 74 2e |.stormy. .night.| 00000130 00 98 49 74 98 20 98 77 61 73 98 20 98 61 98 20 |..It. .was. .a. | 00000140 98 64 61 72 6b 98 20 98 61 6e 64 98 20 98 62 6f |.dark. .and. .bo| 00000150 6c 64 98 20 98 61 6e 64 98 20 98 75 6e 64 65 72 |ld. .and. .under| 00000160 6c 69 6e 65 64 98 20 98 73 74 6f 72 6d 79 98 20 |lined. .stormy. | 00000170 98 6e 69 67 68 74 2e 00 00 a0 49 74 a0 20 a0 77 |.night....It. .w| 00000180 61 73 a0 20 a0 61 6c 73 6f a0 20 a0 73 75 70 65 |as. .also. .supe| 00000190 72 73 63 72 69 70 74 65 64 a0 20 20 c0 61 6e 64 |rscripted. .and| 000001a0 c0 20 c0 73 75 62 73 63 72 69 70 74 65 64 2e c0 |. .subscripted..| 000001b0 20 20 2e 2e 2e 20 62 75 74 20 6e 6f 74 20 62 6f | ... but not bo| 000001c0 74 68 2c 00 61 6e 64 20 6e 6f 74 20 63 6f 6d 62 |th,.and not comb| 000001d0 69 6e 65 64 20 77 69 74 68 20 75 6e 64 65 72 6c |ined with underl| 000001e0 69 6e 69 6e 67 20 6f 72 20 62 6f 6c 64 2e 00 00 |ining or bold...| 000001f0 48 65 20 6c 65 66 74 20 6a 75 73 74 69 66 69 65 |He left justifie| 00000200 64 20 68 69 6d 73 65 6c 66 2e 00 ff 58 48 65 20 |d himself...XHe | 00000210 72 69 67 68 74 20 6a 75 73 74 69 66 69 65 64 20 |right justified | 00000220 68 69 6d 73 65 6c 66 2e 00 ff 24 48 65 20 70 75 |himself...$He pu| 00000230 74 20 68 69 6d 73 65 6c 66 20 72 69 67 68 74 20 |t himself right | 00000240 69 6e 20 74 68 65 20 63 65 6e 74 72 65 2e 00 ff |in the centre...| 00000250 0a 48 65 20 6d 61 64 65 20 65 76 65 72 79 74 68 |.He made everyth| 00000260 69 6e 67 20 65 78 70 61 6e 64 20 6f 75 74 2c 20 |ing expand out, | 00000270 61 20 70 61 72 61 67 72 61 70 68 20 65 78 74 65 |a paragraph exte| 00000280 6e 64 69 6e 67 20 6f 76 65 72 20 6d 75 6c 74 69 |nding over multi| 00000290 70 6c 65 00 ff 0a 6c 69 6e 65 73 2c 20 61 20 66 |ple...lines, a f| 000002a0 72 61 67 6d 65 6e 74 20 74 68 61 74 20 62 65 63 |ragment that bec| 000002b0 61 6d 65 20 61 6e 20 69 63 6f 6e 20 75 6e 74 6f |ame an icon unto| 000002c0 20 69 74 73 65 6c 66 2e 00 00 61 82 61 86 61 87 | itself...a.a.a.| 000002d0 61 82 61 81 20 61 6e 64 20 6e 65 78 74 20 41 82 |a.a. and next A.| 000002e0 41 86 41 87 41 82 41 81 20 6e 65 78 74 20 6e 82 |A.A.A.A. next n.| 000002f0 6e 86 6e 87 6e 82 6e 81 20 6e 65 78 74 20 4e 82 |n.n.n.n. next N.| 00000300 4e 86 4e 87 4e 82 4e 81 20 6d 61 81 6e 81 61 6e |N.N.N.N. ma.n.an| 00000310 61 20 61 62 72 83 65 83 20 61 6c 82 6f 82 20 61 |a abr.e. al.o. a| 00000320 6c 84 65 00 00 00 00 00 00 00 00 00 00 00 00 00 |l.e.............| 00000330 00 00 00 00 00 00 |......|
Most of the document again can be trivially understood as null-delimited strings. Notice that no single line ever exceeds 80 columns in length: as we demonstrated above, there are no Dana paragraphs, only Zuul lines.
There are also various meta-sequences for the formatting. We can pick out the easy ones right away: $90 turns on and off underlining (interesting that Jane automatically skips spaces, much as would be the convention on a typewriter), $88 bold, $98 bold and underlined, $a0 superscript and $c0 subscript.
For justification, we can infer that a line starting with $ff indicates different formatting for that line. A superficial guess would suppose the next character sets the justification type, e.g. $58 right, $24 centre, and $0a full justification (left justification is the default and implied). However, looking at some of the family's files, there were other values here, many of which were close to those numbers but many that were also rather disparate, and none that exceeded $8c (140). Eventually it became clear that all it indicates is the number of leading spaces divided by two, or more generally stated, all levels of justification are simply baked into the document and pre-calculated with spaces. This is entirely consistent with the lack of automatic reflow we saw while in Janewrite.
Accents are also a little odd. If we take the last four words with single accents, then the ornamented letter would seem surrounded by the metacharacter: $81 $63 $81 for ñ (CTRL-@), $82 $6f $82 for ô (^), $83 $65 $83 for é (CTRL-;) and $84 $65 ($84) for è (CTRL-*), though since it occurs at the end of the line it would be cancelled anyway. An experiment I did showed that this is true for multiple characters as well, as long as they're all the same diacritic. So far so good, but the strange part is alternating between multiple accented characters back to back with different accents. Our letter sequences are unaccented, circumflex, grave, acute and tilde, yet the metacharacters seem to change between the $82s, after which we drop down to $81 following the final character which sounds like the delimiter itself changed too. This is the same pattern for each of the other multibyte clusters in this example. I did a test document with "ãâáàa" (i.e., ascending metacharacters $81 $82 $83 $84) and got $81 $61 $83 $61 $81 $61 $87 $61 $84 $61, while "àáâãa" (descending $84 $83 $82 $81) is rendered $84 $61 $87 $61 $81 $61 $83 $61 $81 $61, "áàãâa" (i.e., $83 $84 $81 $82) is rendered $83 $61 $87 $61 $85 $61 $83 $61 $82 $61, and "àâáãa" (i.e., $84 $82 $83 $81) comes out $84 $61 $86 $61 $81 $61 $82 $61 $81 $61.
A couple of Pibb Xtras later, I finally realized what it was doing: it's exclusive-ORing the low nybble of adjacent metacharacters together so that it recognizes them as continuations (or, if the sequence is over, zero). Because of the specific values used, they won't generate any other valid metacharacter. To encode $81 $82 $83 $84, 1 is XORed with 2 and the high bit restored to yield $83, 2 is XORed with 3 to yield $81, 3 is XORed with 4 to yield $87 and then $84 is the expected terminal delimiter. To decode the resulting $81 $83 $81 $87 $84, $81 is used as the initial delimiter, then 1 is XORed with 3 to yield $82, then this is used as the next low nybble XORed with 1 to yield $83, then 3 is XORed with 7 to yield $84, and then 4 XOR 4 is 0, ending the sequence.
This is a little complex to model and I didn't need it to convert the family files, so I went with this small Perl script which converted them to RTF, preserving boldface, underline, superscripting and subscripting, and the leading indent. It was also an excuse to learn how to write RTF documents from scratch, though please don't consider its output to be particularly good RTF.
#!/usr/bin/perl
eval 'use bytes'; $crlf = "\015\012";
# magic read from either ARGV or STDIN, set ARGV filehandle
# splitting it up into bytes was to analyze the header
$/ = \163; @header = map { ord } split(//, <>); # get header
die("abnormally short\n") if (scalar(@header) != 163);
$/ = "\00"; # lines are null-separated
print STDOUT
"{\\rtf1\\ansi{\\fonttbl\\f0\\fmodern\\fprq1 Courier;}\\f0\\fs20\\pard${crlf}";
while(<ARGV>) { chop; # not chomp
$bold = 0; $under = 0; $bunder = 0; $sup = 0; $sub = 0;
# load indent
if (s/^\xff([\x00-\xff])//) { $_ = " " x (ord($1)/2) . $_; }
# force NBSP because Word thinks it's smarter than we are
s/ /\\u160;/g;
unless (/[\x80-\xff]/) { print STDOUT "$_\\par$crlf"; next; }
foreach (map { ord } split(//, $_)) {
unless ($_ & 128) { print STDOUT chr($_); next; }
if ($_ == 136) { $bold ^= 1; print STDOUT "\\b$bold "; next; }
if ($_ == 144) { $under ^= 1;
print STDOUT "\\ul$under "; next; }
if ($_ == 152) { $bunder ^= 1;
print STDOUT "\\ul$bunder\\b$bunder "; next; }
# XXX: sup/sub looks fine in Mac Text Edit, wrong in Word
# because it shrinks the font as well
if ($_ == 160) { $sup ^= 1;
print STDOUT (($sup) ? '{\super ' : '}'); };
if ($_ == 192) { $sub ^= 1;
print STDOUT (($sub) ? '{\sub ' : '}'); };
# XXX: accented characters not yet implemented
}
print STDOUT '\b0' if ($bold || $bunder);
print STDOUT '\ul0' if ($under || $bunder);
print STDOUT '}' if ($sub || $sup); # shouldn't be nested
print STDOUT "\\par$crlf";
}
print STDOUT "}$crlf";
I relentlessly tested the output against TextEdit, NeoOffice, LibreOffice and Microsoft Word on the Mac, and Word on my office PC, and concluded there was no good way to make them all happy. Since Janewrite is strictly non-proportional and character-oriented, the resulting document has to be monospace and spaces must be fixed-width to preserve the original formatting. TextEdit was the most flexible about this, accepted nearly anything, and rendered it as expected. Word, however, thought it was smarter than my carefully crafted output and changed the width of spaces even with an explicitly monospace font such that things failed to line up. The solution was to make it use a non-blocking space, which TextEdit also serenely accepts (though it wanted an explicit delimiter, not just a space character), but which shows up as a block character in NeoOffice and LibreOffice. I gave up at that point and declared it good enough. Change the above if you disagree. Maybe I should have done this in HTML because everything renders HTML the same, right? (Right??)
A residual problem is that superscript and subscript in Word, NeoOffice and LibreOffice also reduce the font size, which is not consistent with Jane and also messes up alignment, and there didn't seem to be a way to manually specify the vertical baseline in RTF but keep the same font size to get around that. Once again, only TextEdit's output properly resembles Janewrite's. This is another exercise left for the reader since I didn't need that to convert their documents either.
With our task complete, let's finish the story of Jane. The Atari 8-bit and IBM PC ports likewise never materialized, but Jane was ultimately ported to one more — and, outside its home country of France, rather obscure — system: the 1985 Thomson MO5, manufactured by French electronics company Thomson SA (now Vantiva SA), which had recently been nationalized out of Thomson-Brandt and Thomson-CSF by French president François Mitterand in 1982.
Although it's arguable which of the two is more powerful — the Apple II version is more flexible and more integrated, but the Commodore 128 version has a stronger interface, larger workspace and 80-column output — Jane 128 got similar reviews to Jane on the Apple II and some of the same criticism as well. RUN in March 1987 again approved of the interface, but complained about the sluggish performance — both on-screen and on-paper — and that the 80-column software output was hard to read compared to the VDC's native 80-column output, which the reviewer called "[t]he program's greatest flaw." The review also noted that printing a spreadsheet had to be stopped manually or it would merrily waste paper on empty cells until the entire sheet had been iterated over. Still, reviewer John Premack said it was "a must for beginners of all ages" and gave it a "B" grade overall.
I don't think Commodore was expecting much from Jane, and it ably occupied the low-end niche that Goheen and Smith wanted it to, so perhaps that was enough. It certainly seems to have been for my Australian 128DCR's previous owners, because the entire family used it (even some brief school papers from their kid) and some documents are dated well into 1990, long after Jane — and for that matter the 128 itself — was off the market. That probably speaks well of its interface and its suitability to basic tasks.
But this was all likely a mere distraction to Marks and particularly Kotick, who in 1987 turned around to make an audacious if unsuccessful attempt to buy Commodore, believing the Amiga could lead a comeback in computer gaming after the 1983 video game crash. Commodore managed to rebuff him and in 1991 Kotick, Marks and their partners executed a hostile takeover of another target instead for just $440,000 [about $1.15 million], an ailing software company called Mediagenic. Mediagenic had a long prior history in video games, but believed the field was unlikely to rebound, and was in the midst of an unsuccessful effort to broaden into productivity software. Kotick, who had just left that market segment, disagreed and reconstituted the company back into its prior business and under its former name: Activision. In the process Kotick didn't forget his previous benefactor and Steve Wynn ended up with a million and a half shares in the new company, which became so successful in its revitalized form that in 2012 Wynn commented in amazement to the New York Times that "[t]he kid [Kotick] was telling me I had $31 million I didn't know about."
As for Sig Hartmann, arguably the father of Commodore's software division, he passed away in 2014.
No comments:
Post a Comment
Comments are subject to moderation. Be nice.