Hacker News Reader

Comment by edwinjm | Hacker News Reader

Heh, if you're looking for a good example of Technical Debt…

Yes, already in 1995 Unicode was an established standard (even Windows 95 started to support it). The bank should have known it would be a requirement in the future.

View on HN

Unicode's old enough that Windows NT was built to work with it natively. In fact, all the "ANSI" Windows API calls in NT were just wrapper functions around the Unicode equivalents handling Unicode/code-page conversions. And this was 1993.

Yup, in fact the biggest compatibility headaches in NT today stem from how early they adopted it: they made some assumptions about UCS-2 that turned out to be wrong and had to shoehorn in UTF-16 support that mostly works (except when it falls over a cliff). Meanwhile Linux and others waited for UTF-8 to exist and that's become the internet/web's major standard as well and there are some small papercuts interoperating between UTF-16 and UTF-8 that with today's hindsight shouldn't have been so annoying or necessary. Windows might have been better off waiting for UTF-8 itself other than Windows made the right architectural decision for the time when it made that decision and could not have suspected UTF-8 to turn up only a few years later.

> Meanwhile Linux and others waited for UTF-8

UTF-8 was already a thing in Plan9.

UTF-8 was first presented to IETF at Usenix in January 1993. NT 1.0 shipped June 1993 and had been in development for several years before that.

The famous "Plan 9 implemented UTF-8 first" thread's most specific date mentioned was September 1992 which only three months more lead time before the standardization notice in January 1993.

Are you suggesting the NT Kernel team should have somehow better paid attention to a not-yet-standard from a research laboratory Operating System? It still probably would have been a couple years too late in the design/architecture process even if they had, given the release data in June 1993.

NT was already a functional OS in October 1991: https://betawiki.net/wiki/Windows_NT_3.1_October_1991_build and http://www.os2museum.com/wp/installing-oct-91-nt-from-cd/

Those decisions were already way beyond locked back then.

UTF-8 dates back to 1992.

http://xahlee.info/comp/unicode_utf8_history.html

When speaking with hindsight, it's easy to make the mistake of conflating a point in time where an idea existed/was published (an easy fact to establish upper bounds on through records) with a point in time where that idea was popular and/or taken seriously (much harder to establish and usually localized).

UTF-8 was standardized in 1996[1][2]. As for its use in Linux, for example Debian switched to use UTF-8 as default in Etch[3], released in 2007[4].

[1] https://datatracker.ietf.org/doc/html/rfc2044

[2] https://www.iso.org/standard/26223.html

[3] https://www.debian.org/releases/etch/amd64/release-notes/ch-...

[4] https://wiki.debian.org/DebianEtch

UTf-8 in Plan9 is from 1992.

Reply on news.ycombinator.com