As what often happens, using \ for paths is for backwards compatibility.
Neither CP/M nor MS-DOS 1.0 had folders. When folders were added in MS-DOS 2.0, the syntax had to be backwards compatible. DOS already used forward slashes for command-line options (e.g. DIR /W) so using them for folders would have been ambiguous - does that DIR command have a /W option, or is it viewing the contents of the W directory at the root of the drive? Backslashes weren’t used for anything so they used them for folders.
This is the same reason why you can’t create files with device names like con, lpt1, and so on. DOS 2.0 has to retain backwards compatibility with 1.0 where you could do something like TYPE foo.txt > LPT1 to send a document to a printer. The device names are reserved globally so they can work regardless of what folder you’re in.
It makes sense why they did it, but their messed up versioning was the cause to begin with. You should always assume Devs will cut corners in inappropriate ways.
An often repeated urban legend that has no basis in reality. Software checking the version of Windows gets “6.1” for Windows 7 and “6.2” for Windows 8. The marketing name doesn’t matter and is different.
some legacy software checked if the OS name began with “Windows 9” to differentiate between 95 and future versions.
This is a myth. Windows doesn’t even have an API to give you the marketing name of the OS. Internally, Windows 95 is version 4.0 and Windows 98 is 4.1. The API to get the version returns the major and minor version separately, so to check for Windows 95 you’d check if majorVersion = 4 and minorVersion = 0.
Maybe it’s a myth, but it sure sounds plausible. The software that checks the “Windows 9” substring doesn’t even have to exist for this to be reason they chose to skip to version 10 — they just had to be concerned that it might exist.
Sure, maybe there’s no C function that returns the string, but there’s a ver command. It would be trivial to shell out to the command. https://en.wikipedia.org/wiki/Ver_(command)
And for the same reason they went straight from 2.1 3.x to 5.0 when they renamed .Net Core to just .Net. Versions 3.x and 4.x would have been too easy to confuse (either manually or programmatically) with the old .Net Framework versions that were still in use, especially for Desktop applications.
Dotnet core 4 never existed because they wanted to make it the mainline dotnet… That means framework is retired and everything is now the slimmer multiplatform runtime.
I was about to say that most apps should check the NT number but then I remembered that until XP it wasn’t common to run a NT system, but then I remembered NT 4 existed basically in the same timeframe as 95 did, and even if the argument went to “it’s a 9x application”, shouldn’t these OSes at least have some sort of build number or different identifier systems? Because as I said NT systems were around, so they would probably need a check for that
But that actually made sense! They care about backwards compatibility.
For those not in the know: some legacy software checked if the OS name began with “Windows 9” to differentiate between 95 and future versions.
deleted by creator
Say whatever you want about Microsoft, but they don’t mess around with backwards compatibility.
It’s easy to be backwards compatible when you’re backwards in general.
I once heard some YouTuber say Windows uses \ in path names instead of / like everyone else because Microsoft thinks backwards.
As what often happens, using
\for paths is for backwards compatibility.Neither CP/M nor MS-DOS 1.0 had folders. When folders were added in MS-DOS 2.0, the syntax had to be backwards compatible. DOS already used forward slashes for command-line options (e.g.
DIR /W) so using them for folders would have been ambiguous - does thatDIRcommand have a/Woption, or is it viewing the contents of theWdirectory at the root of the drive? Backslashes weren’t used for anything so they used them for folders.This is the same reason why you can’t create files with device names like
con,lpt1, and so on. DOS 2.0 has to retain backwards compatibility with 1.0 where you could do something likeTYPE foo.txt > LPT1to send a document to a printer. The device names are reserved globally so they can work regardless of what folder you’re in.Well, better to be backwards with backwards compatibility than to just be backwards.
looks at Apple
They probably search for windows n(t) somewhere too ;)
But “nine” is a word that is a number
The reason they checked that it started with “Windows 9” was because it worked for “Windows 95” and “Windows 98”
It makes sense why they did it, but their messed up versioning was the cause to begin with. You should always assume Devs will cut corners in inappropriate ways.
They’ll cut corners the more the shittier APIs and ABIs you provide
The API is fine. It returns the internal version number (which is 4.0 for Windows 95), not a string. https://learn.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-osversioninfoexa. There’s no built-in API that returns “Windows 95” as a string.
An often repeated urban legend that has no basis in reality. Software checking the version of Windows gets “6.1” for Windows 7 and “6.2” for Windows 8. The marketing name doesn’t matter and is different.
This is a myth. Windows doesn’t even have an API to give you the marketing name of the OS. Internally, Windows 95 is version 4.0 and Windows 98 is 4.1. The API to get the version returns the major and minor version separately, so to check for Windows 95 you’d check if majorVersion = 4 and minorVersion = 0.
Edit: This is the return type from the API: https://learn.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-osversioninfoexa
Maybe it’s a myth, but it sure sounds plausible. The software that checks the “Windows 9” substring doesn’t even have to exist for this to be reason they chose to skip to version 10 — they just had to be concerned that it might exist.
Sure, maybe there’s no C function that returns the string, but there’s a
vercommand. It would be trivial to shell out to the command. https://en.wikipedia.org/wiki/Ver_(command)This doesn’t prove anything, but there are a TON of examples of code that checks for the substring. It’s not hard to imagine that code written circa 2000 would not be future proof. https://sourcegraph.com/search?q=context:global+“\“windows+9\””&patternType=keyword&sm=0
oh
oh no
There’s code in the JDK that does that??
I really wish I didn’t see that.
Yup!! Never look under the hood in software, you’ll just be disappointed ☹️
I’ve been a software developer for 20 years and this comment is too real. Some days I’m amazed that any software even works at all.
Having worked in both food service and software, I encourage you not to visit the kitchen of any restaurants you enjoy either.
Please don’t show me how the sausage is made
And for the same reason they went straight from
2.13.x to 5.0 when they renamed .Net Core to just .Net. Versions3.x and4.x would have been too easy to confuse (either manually or programmatically) with the old .Net Framework versions that were still in use, especially for Desktop applications.Dotnet core 3.x exists
Dotnet core 4 never existed because they wanted to make it the mainline dotnet… That means framework is retired and everything is now the slimmer multiplatform runtime.
Oh yeah right, I totally forgot 3.x.
Strange argument… how does that prevent checks versus Windows 7, 8 and 1* all of which would be less than 9.
Because it checks if the version starts with the string “Windows 9*”, not wether the number is less than 9.
This is a myth - code that checks the version number uses the internal version number, which is 4.0 for Windows 95.
I was about to say that most apps should check the NT number but then I remembered that until XP it wasn’t common to run a NT system, but then I remembered NT 4 existed basically in the same timeframe as 95 did, and even if the argument went to “it’s a 9x application”, shouldn’t these OSes at least have some sort of build number or different identifier systems? Because as I said NT systems were around, so they would probably need a check for that
Some programs just didn’t work on NT though. A lot of installers were more OS specific back then.
Eh. I think Microsoft should have let that break so the spaghetti code finally gets fixed