Comment by gregw2 | Hacker News Reader

Comment by gregw2 | Hacker News Reader

Transitioning off a long established core mainframe/AS400 app is not necessarily so easy as just changing to UTF-8 as the article author implies.

If you have no mainframe or enterprise experience to relate to that observation, consider the effort involved to transition from python 2 to (UTF-8 clean) python 3!

That said, I am not even clear from the article which diacritical markings are missing from EBCDIC and if the lawyers arguments to "not change" were legitimate in the way the article implies... you do realize there are hundreds of EBCDIC code pages covering at least all the European languages ... since these are markets which IBM has sold into for 50+ years now, right?

I only learned about EBCDIC code pages when trying to proactively properly setup character encoding handling for data extraction from one of my employer's long running AS400s... "Which EBCDIC?" is not that different a headache from "which extended ASCII code page?"... EBCDIC is not just like 7-bit (non extended) ASCII as the article implies.

> Transitioning off a long established core mainframe app is not necessarily so easy

I don't think the author implied it was easy, just that it should have been done at some point in the 25 years since the system was first implemented. The last paragraph is just an exortation to use Unicode everywhere all the time, today.

> "Which EBCDIC?" is not that different a headache from "which extended ASCII code page?"

Sure, but that's still a massive headache. You've probably never had a a headache like needing to switch ASCII or EBCDIC code pages. You generally can't just switch code pages per-record in a file, storing mixed code page data to disk is generally a bad idea, and in some operating systems you can barely switch code pages per application and sometimes need ROM hacks and entire mainframe restarts to switch code pages. (Modern z/OS supports something more like modern Linux locale switching with environment variables before running applications so should at least allow per-application code pages.)

Even if the lowest common denominator code page you choose to run your application in is a full bit or two more than the 7-bit ASCII lowest common denominator a single code page per application is still never going to cover the breadth of UTF-8 without nasty hacks. (That's of course assuming you don't have other problems such as intermediate tools that presume you are only using ASCII compatible EBCDIC subsets of code pages, which may be the case when you've got an eclectic evolution of code accreted around your mainframe apps.)

Reply on news.ycombinator.com