Pivotal Suboptimal Decisions in the History of Software

Question

Throughout the history of software development, it sometimes happens that some person (usually unknown, probably unwittingly) made what, at the time, seemed a trivial, short-term decision that changed the world of programming. What events of this nature come to mind, and what have been our industry's response to mitigate the pain?

Illustration (the biggest one I can think of): When IBM designed the original PC, and decided to save a couple dollars in manufacturing costs by choosing the half-brain-dead 8088 with 8-bit-addressable memory, instead of one of the 16-bit options (8086, 680n, etc.), dooming us to 20 years of address offset address calculations.

(In response, a lot of careers in unix platform development were begun.)

Somewhere toward the other end of the scale lies the decision someone made to have a monster Shift Lock key at the left end of the keyboard, instead of a Ctrl key.

Answer 1

Allocating only 2 digits for the year field.

And the mitigation was to spend huge amounts of money and time just before the fields overflowed to extend them and fix the code.

Answer 2

Apple ousting Steve Jobs (the first time) to be led by a succession of sugar-water salemen and uninspired and uninspiring bean counters.

Answer 3

Paul Allen deciding to use the / character for command line options in MS DOS.

Answer 4

Using 4 bytes for time_t and in the internet protocols' timestamps.

This has not bitten us yet - give it a bit more time.

Answer 5

Using the qwerty keyboard on computers instead of dvorak.

Answer 6

There were a lot of suboptimal decisions in the design of C (operator precedence, the silly case statement, etc.), that are embedded in a whole lot of software in many languages (C, C++, Java, Objective-C, maybe C# - not familiar with that one).

I believe Dennis Ritchie remarked that he rethought precedence fairly soon, but wasn't going to change it. Not with a whole three installations and hundreds of thousands of lines of source code in the world.

Answer 7

Microsoft deciding to use backslash rather than forwardslash as the path delimiter. And failing to virtualize the drive letter.

Answer 8

Deciding that HTML should be used for anything other than marking up hypertext documents.

Answer 9

Thinking that a password would be a neat way to control access.

Answer 10

Gary Kildall not making a deal with IBM to license CP/M 86 to them, so they wouldn't use MS-DOS.

Answer 11

Every language designer who has made their syntax different when the only reason was "just to be different". I'm thinking of S and R, where comments start with #, and _ is an assignment operator.

Answer 12

Deciding that "network order" for multi-byte numbers in the Internet Protocol would be high order byte first.

(At the time the heterogenous nature of the net meant this was a coin toss decision. Thirty years later, Intel-derived processors so completely dominate the marketplace it seems lower-order-byte first would have been a better choice).

Answer 13

Lisp's use of the names "CAR" and "CDR" instead of something reasonable for those basic functions.

Answer 14

Netscape's decision to support Java in their browser.

Answer 15

Important web sites like banks still using "security questions" as secondary security for people who forget their passwords. Ask Sarah Palin how well that works when everybody can look up your mother's maiden name on Wikipedia. Or better yet, find the blog post that Bruce Schneier wrote about it.

Answer 16

Actually the 8088 & 8086 have same memory model and same number of address bits (20). Only difference is width of external data bus which is 8 bit for 8088 & 16 bit for 8086.

I would say that use of inconsistent line endings by different operating systems (\n - UNIX, \r\n - DOS, \r - Mac) was a bad decision. Eventually Apple relented by making \n default for OS-X but Microsoft is stubbornly sticking to \r\n. Even in Vista, Notepad can not properly display a text file using \n as line ending.

Best example of this problem is the ASCII mode of FTP which just adds /r to each /n in a file transferred from a UNIX server to Windows client even though the file originally contained /r/n.

Answer 17

HTML as a browser display language.

HTML was originally designed a content markup language, whose goal was to describe the contents of a document without making too many judgments about how that document should be displayed. Which was great except that appearance is very important for most web pages and especially important for web applications.

So, we've been patching HTML ever since with CSS, XHTML, Javascript, Flash, Silverlight and Ajax all in order to provide consistent cross-browser display rendering, dynamic content and the client-side intelligence that web applications demand.

How many times have you wished that browser control languages had been done right in the first place?

Answer 18

Re-arranging the letters on the keyboard to slow down typing productivity, back in the original mechanical typewritter days, and carrying that over to digital computers.

The History of Qwerty

Answer 19

Microsoft's decision not to add *NIX-like execute/noexecute file permissions and security in MS-DOS. I'd say that ninety percent of the windows viruses (and spyware) that we have today would be eliminated if every executable file needed to be marked as executable before it can even execute (and much less wreak havoc) on a system.

That one decision alone gave rise to the birth of the Antivirus industry.

Answer 20

Ending Alan Turing's career when he was only 42.

Answer 21

The term Translation Lookaside Buffer (which should be called something along the lines of Page Cache or Address Cache).

Answer 22

Microsoft copying the shortcut keys from the original Mac but using Ctrl instead of a Command key for Undo, Cut, Copy, Paste, etc. (Z, X, C, V, etc.), and adding a near worthless Windows key in the thumb position that does almost nothing compared to the pinky's numerous Ctrl key duties. (Modern Macs get a useful Ctrl key (for terminal commands), and a Command key in the thumb position (for program or system shortcuts) and an Alt (Option) key for typing weird characters.) (See this article.)

Answer 23

Having a key for Caps Lock instead of for Shift Lock, in effect it's a Caps Reverse key, but with Shift Lock it could have been controllable.

Answer 24

Null-terminated strings

Answer 25

Microsoft's decision to use "C:\Program Files" as the standard folder name where programs should be installed in Windows. Suddenly working from a command prompt became much more complicated because of that wordy location with an embedded space. You couldn't just type:

cd \program files\MyCompany\MyProgram

Anytime you have a space in a directory name, you have to encase the entire thing in quotes, like this:

cd "\program files\MyCompany\MyProgram"

Why couldn't they have just called it c:\programs or something like that?

Answer 26

Null References - a billion dollar mistake.

Answer 27

Microsoft's decision to base Window NT on DEC VMS instead of Unix.

Answer 28

EBCDIC, the IBM "standard" character set for mainframes. The collation sequence was "insane" (the letters of the alphabet are not contiguous).

Answer 29

7-bits for text. And then "fixing" this with code pages. Encoding issues will kill me some day.

Answer 30

We have that covered by helpful libraries these days, but the moron at Netscape that decided that cookie-expriation dates should be human readable should take measure to never meet me in person.

Answer 31

Netscape's decision to rewrite their browser from scratch. This is arguably one of the factors that contributed to Internet Explorer running away with browser market share between Netscape 4.0 and Netscape 6.0.

Answer 32

DOS's 8Dot3 file names, and Windows' adoption of using the file extension to determine what application to launch.