[Monotone-devel] Re: the line-ending discussion

[Graydon Hoare]

Yury Polyanskiy wrote:

>  My original question was very simple: how you specify what "treat by a
> sequence of lines" means? Does it mean converting all weird combinations
> of CR and LF into line-ending or just convert "native line ending" into
> line-ending? 

Let us be clear, and talk about 4 specific operations. When I say 
"sequence of bytes" I mean "std::string". When I say "sequence of lines" 
I mean "std::vector<std::string>". Here are the operations:

  1. workspace file -> std::vector<std::string>
  2. workspace file <- std::vector<std::string>
  3. workspace file -> std::string
  4. workspace file <- std::string

When a file is considered "lines" (by attribute), it is subject to 
operations 1 and 2. When a file is considered "bytes", it is subject to 
operations 3 and 4.

Operations 3 and 4 are byte-copy operations.

Operations 1 and 2 are done by splitting or joining. The split and join 
operations are governed by the following rules:

   - At first, monotone's default behavior will be to use the "native"
     line-ending forms (only LF on unix, only CR on mac, only CRLF on
     windows). If you are sharing your workspace between platforms on
     NFS or SMB, you are out of luck. For now.

   - Once we have a mechanism for specifying project policies, it will
     be a matter of project policy, and you can split and join lines
     using whatever you want.

This is not what monotone does now, but I find it an agreeable road 
forwards. Does anyone disagree?

-graydon