Lessons learned from handling text end-of-line conventions?

[Alexander Belchenko]

Ben Finney пишет:
> Alexander Belchenko <bialix at ukr.net> writes:
> 
>> Well, I've read entire thread and I should say it worth to read it.
> 
> My intention, as stated initially, was to get a pointer to information
> to help the Mercurial folks get a better idea of the issues as they were
> encountered in Bazaar.
> 
>> Just to understand what's wrong with current eol support in bzr,
>> because it's now much more like in hg.
> 
> So, do you have good information to pass back the other direction, so
> the Mercurial guys can try to avoid the same traps already known by the
> similar work in Bazaar?

No, I don't want to provide any hints for Mercurial developers, because 
of main point of core hg developers (and this was mentioned by Mark 
Hammond in that thread):

"hg devs made very quick system that flies on Linux. If it does not work 
on Windows -- it's not our problem. If you want to fix this -- do 
anything you want, but don't touch speed and on Linux it still should fly".

If you carefully read many Mark's replies you can see the right way he 
proposed and I agree with: eol settings should be versioned and 
propagated on clone, win32text should be included in core and enabled 
*always*.