Re: [Monotone-devel] Re: Looking at the code affected in bug 9752 leaves a weird taste...

[Jon Bright]

Hi,

graydon hoare wrote:

> however, that was a conservative consensus, and it has clearly failed
> in practise. so let's work out what it ought to do, not worry too much
> about what the current (wrong) code does.

My personal opinion (doubtless shared by 100% of other people :-) is:

1. monotone should default to storing LF-terminated lines in the DB 
(having a canonical representation in the DB seems to me to be a must).

2. On Windows/DOS/similar, files should by default be checked out with 
CR+LF.  I really dislike this idea (I'd prefer LFs everywhere), but a 
fair number of Windows tools deal poorly / not at all with LF-only.

3. On other OSes, the unchanged form is checked out.

4. The DB also keeps a mode, which can have one of four characters:

a) Default.  Behaviour as in 1,2,3 above.
b) Always CR+LF.  Checked-out versions are thereafter always CR+LF. 
Whether this should store CR+LF in the DB or not, I don't know.  I think 
it probably shouldn't alter the canonical form - hence LFs in the DB.
c) Always LF.  Checked-out versions are thereafter always LF-only.
d) Binary.  Don't touch this file.  Don't attempt to convert it to a 
canonical form on the way into the DB, don't attempt to convert it to a 
platform-based form on the way out.

If this were a char(1) column in the DB, people could add their own 
customised behaviours via an LUA hook.  Values 'A'-'Z' reserved for 
Monotone's use, other values free for people's private extensions?

--
Jon Bright
Silicon Circus Ltd.
http://www.siliconcircus.com