[Monotone-devel] the line-ending discussion

[Graydon Hoare]

Hi,

I'm really surprised this thread has carried on so long and involved so 
many harsh words. I also see a lot of over-design. I really only see two 
situations which need differentiation:

  1. Files you wish monotone to treat as a sequence of bytes.
  2. Files you wish monotone to treat as a sequence of lines.

The reason for supporting case #2 is that many tools (diff, patch, our 
diff-writer and auto-merger) work in terms of lines, and we like the 
behavior of those tools.

Monotone will address these cases by marking one sort of file with an 
attribute, and assuming a file is "the other case" when it lacks the 
attribute.  Observant readers will note that, given the mutual 
exclusivity of cases, it does not matter which gets marked. 
Implementation detail.

That leaves UI, which is where most of the debate is.

There are two debates in the UI space:

  1. Whether to assume a new file is "lines" or "bytes" by default.
  2. How to pick the line endings, when writing a "lines" file to the
     workspace, and conversely which line ending to break on when
     reading a "lines" file back from the workspace.

Different users have different, sensible preferences on these matters. 
They are "policy" for each project. Monotone will have one default wired 
into it -- my preference is "bytes" for #1 and "native" for #2 -- but 
each project will have the ability to override that default once we have 
a mechanism for storing project policy ("management branches").

Until such time as we have a mechanism for setting project policy, we 
can add an FAQ entry about setting the attribute manually, with a 
promise to move to a project-policy approach later.

Is this a clear enough position for everyone?

(Assuming we take off our "nit-picking overdesigner" hats and put on our 
"simplicity-loving compromiser" hats)

-graydon