Comment by floren
14 hours ago
Remember that back in the mists of time, computers used typewriter-esque machines for user interaction and text output. You had to send a CR followed by an LF to go to the next line on the physical device. Storing both characters in the file meant the OS didn't need to insert any additional characters when printing. Having two separate characters let you do tricks like overstriking (just send CR, no LF)
True, but I don’t think there was a common reason to ever send a linefeed without going back to the beginning. Were people printing lots of vertical pipe characters at column 70 or something?
It would’ve been far less messy to make printers process linefeed like \n acts today, and omit the redundant CR. Then you could still use CR for those overstrike purposes but have a 1-byte universal newline character, which we almost finally have today now that Windows mostly stopped resisting the inevitable.
> now that Windows mostly stopped resisting the inevitable
I've been trying to get Visual Studio to stop mucking with line endings and encodings for years. I've searched and set all the relevant settings I could find, including using a .editorconfig file, but it refuses to be consistent. Someone please tell me I'm wrong and there's a way to force LF and UTF-8 no-BOM for all files all the time. I can't believe how much time I waste on this, mainly so diffs are clean.
Ugh, I didn't realize it was still that bad.
How far can you get with setting core.autocrlf on your machine? See https://git-scm.com/book/en/v2/Customizing-Git-Git-Configura...
As I understand it (this may be apocryphal but I've seen it in multiple places) the print head on simple-minded output devices didn't move fast enough to get all the way back over to the left before it started to output the next character. Making LF a separate character to be issued after CR meant that the line feed would happen while the carriage was returning, and then it's ready to print the next character. This lets you process incoming characters at a consistent rate; otherwise you'd need some way to buffer the characters that arrived while the CR was happening.
Now, if you want to use CR by itself for fancy overstriking etc. you'd need to put something else into the character stream, like a space followed by a backspace, just to kill time.
I don't think that's right. Not saying that to argue, more to discuss this because it's fun to think about.
In any event, wouldn't you have to either buffer or use flow-control to pause receiving while a CR was being processed? You wouldn't want to start printing the next line's characters in reverse while the carriage was going back to the beginning.
My suspicion is there was a committee that was more bent on purity than practicality that day, and they were opposed to the idea of having CR for "go to column 0" and newline for "go to column 0 and also advance the paper", even though it seems extremely unlikely you'd ever want "advance the paper without going to column 0" (which you could still emulate it with newline + tab or newline + 43 spaces for those exceptional cases).
2 replies →