8x19 Text Mode Font Origins
os2museum.com73 points by userbinator 5 days ago
73 points by userbinator 5 days ago
This article [1] has a little more information about the font. Supposedly it first appeared in "Intel's Hi-Flex BIOS".
The source for the font appears here [2] with an Intel copyright.
[1]: https://wiki.softhistory.org/wiki/PhoenixBIOS_4.0_Release_6....
[2]: https://github.com/tianocore/edk2-codereview/blob/0a5f23c06d...
Important reminder that bitmap fonts are not copyrightable in the United States of America.
It's more complicated in the UK but font-related copyrights "only" last for 25 years, so 1990s fonts are well clear now.
Something I was wondering while reading the article: why? Why would you create a 8x19 font if you already have a 8x16 don't? The answer is: so that you can use a standard 640x480 VGA resolution instead of of 640x400 to display 25 lines of text.
I'm still confused: why not stick with 8x16 and go from 25 to 30 lines of text? Is it that you need to support both 640x480 and 640x400?
There was a long legacy of using 80x25 and lots of software that would assume this layout. I think that comes from an even older legacy of dumb terminals such as the VT100 that used 80x24 plus one line for a status line. BTW 80 columns comes from an even older legacy of IBM punch cards having 80 columns. Basically, anything that wasn't 80x25 was going to break a lot of things
And the 80 columns come (losely) from a still older standard in typography, of about 70 characters per line, found empirically as a good size for a line. Even today is good design practice in UI, Web and books to stick to 60 to 80 CPL.
The reason for it in terminals is much more directly linked to the IBM punched card format.
Because 80x25 uses 4kB of RAM (one byte for character + another byte for attributes) whereas 80x30 would grow beyond 4kB so you would need 8kB. Maybe not a big deal in a VGA card, but everything was standardized on 80x25 from the olden days of MDA/CGA which had little video memory, so a lot of software expected that.
Plus, making the characters shorter would make them also a bit less legible.
And maybe cost also? Not sure, but back in the days maybe 4kB more would have a noticeable (albeit I do not think extrem) effect on final price.
Not in the days when these BIOSes were written, but in the early eighties, yes; the original IBM MDA card had only 4kB memory (enough for just one screen of text) because memory was expensive. So sticking to 80x25 was kind of important back then.
For the execution environment in which firmware SETUP ran, there were often hardwired assumptions, particularly in the parts of SETUP that provide full-screen TUIs with menus and entryfields and pop-up help boxes and whatnot, that the screen was 25 rows high.
Also remember that SETUP as a whole was often a mess of different ROM programs from different manufacturers. Far too many thought it their own ROM's business to clear the display and set text mode before they splashed their copyright strings, prompts, self-test, and hardware auto-detection information up.
Yes, they could have done things properly. The way to detect an existing screen size was there from the VGA onwards. And they could have left the display uncleared. And the whole preserve-my-graphical-splash-screen thing eventually did happen.
They did not.
The way to detect an existing screen size was there from the VGA onwards.
When you only have a few KB in total for your option ROM, that's a luxury you can't afford. You can either sacrifice some other important feature of your product to appease the tiny minority, or assume 80x25.
> Is it that you need to support both 640x480 and 640x400
Probably.
A lot of software back then was rendering UIs with box drawing characters, doing complicated layout etc - it's probably much easier to retrofit existing software to output the exact same thing in a slightly higher resolution font than to make the display logic support different resolutions properly.
Because standard.
There was also small .com utils circulating that would change to a tiny font so you could have 40, 50 rows of text with tiny fonts, or maybe even more.
Of course that asks the question of why use the graphics mode for text, which I would suspect was simply to show the seemingly obligatory Energy Star logo, but these Intel boards don’t have it.
The Energy Star logo tended to be made up of custom characters in text mode.
It's not a graphics mode. It is just a resolution from the graphics mode, repurposed for the text mode. The reason possibly being the 640x480 resolution is universally supported while 720x400 is not - for example my HDMI TV can still display 640x480.
CRTs were still commonplace though based on the bios dates in the blog. So 720x400 would still have been well supported. It’s strange that they did this!
Unlike CRT screens, LCD ones from that time had usually a fixed resolution and wouldn't resize the image on the fly.
This apparently was a text mode not a graphics mode. It would have been a tweaked text mode since the standard mode 3 text mode that you can set via int 10h is 720x400.
The energy star logo was also displayed in text modes - by using custom font glyphs!
There are many aspect ratios of fixed width fonts (see https://int10h.org/oldschool-pc-fonts/fontlist/), from square to rectangular, and this one is more than twice as tall as wide
Is there any info or research on which width/height ratio of a fixed width font is the best for readability?
Don't forget that a lot of old PC graphics modes weren't 1:1. 640x200 was a thing, for example - the pixel ratio is 3.2:1, but it gets vertically stretched to a standard 4:3. An 8x8 font wouldn't look square in that graphics mode.
Aside: Going down the "memory lane" of Intel motherboard codenames was fun. Tuscon II, Anchorage, Atlanta, Seattle, Maui, etc.
I really enjoyed Intel's desktop boards. They weren't particularly flashy, and there were certainly "duds" in there (early MU440EX revs not handling Pentium II CPUs properly, the whole RAMBUS debacle) but in general Intel made a solid (and not at all flashy) board.
With Intel's manufacturing competency you could be assured every board would be consistent. If there was a defect (I'm looking at you, MTH in the 820 chipset) every board would consistently have the same defect.
I am oddly happy that there's an OS/2 museum
0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 1,1,1,1,1,0,0,0, 0,0,0,0,0,0,0,0, 0,1,1,0,1,1,0,0, 0,0,0,0,0,0,0,0, 0,1,1,0,0,1,1,0, 1,1,0,0,0,1,1,0, 0,1,1,0,0,1,1,0, 0,1,1,0,1,1,0,0, 0,1,1,0,0,1,1,0, 0,0,1,1,1,0,0,0, 0,1,1,0,0,1,1,0, 0,0,1,1,1,0,0,0, 0,1,1,0,0,1,1,0, 0,1,1,0,1,1,0,0, 0,1,1,0,0,1,1,0, 0,1,1,0,1,1,0,0, 0,1,1,0,0,1,1,0, 1,1,0,0,0,1,1,0, 0,1,1,0,0,1,1,0, 1,1,0,0,0,1,1,0, 0,1,1,0,1,1,1,0, 0,0,0,0,0,0,0,0, 1,1,1,1,1,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,