A few days ago Microsoft published on GitHub in open source mode MS-DOS 4.0 operating system source code. According to Michal Necasek, developer and manager of the OS/2 Museum blog, although the Redmond house has done an excellent deed, unfortunately it would have mutilated the code by publishing it to Git. The latter, in fact, does not conserve i timestamp and the converting files to UTF-8 It “breaks” pretty much everything.
“First, Git doesn’t preserve timestamps, which causes irreversible damage. Knowing when a source file was last modified is valuable information,” writes Necasek in a post titled “How Not To Release Historic Source Code.”
There converting source files to UTF-8 seems to be even more serious. “For practical purposes, old source files are not text files. They are binary files and should be preserved unchanged. It is no good taking an old source file and converting it to UTF-8 […] Obviously old tools can’t handle it!“.
The MS-DOS 4 Microsoft Macro Assembler (MASM) byte line length limit was 512 bytes, and UTF-8 conversion pushes certain files beyond that limit, making them unreadable. In some cases the situation is solvable, in others.
“The historical source code it should be released simply as a file archive, ZIP or tar or 7z or whateverwith all timestamps preserved and every single byte kept as it was. Git is simply not a suitable tool“, concludes Necasek.
Below the post are countless comments in which Connor Hyde, aka Starfrost, who worked with Microsoft on this release, acknowledges the problem and reveals his reasons for not including timestamps. “The reason I can’t create timestamps is because data protection law requires anonymization of source files, at least that’s the policy“, writes Starfrost, moving the discussion with Necasek via email in order to resolve the mess, if that can be done.
