Last day to vote/reject on proposed EOL names

[Alexander Belchenko]

Eric Siegerman пишет:
> On Tue, 2009-03-31 at 08:43 +1000, Ian Clatworthy wrote:
>> In general, wt-format[:repo-format] where the default
>> repo-format is lf. I'm happy to change the separator to -, +, / or
>> something else but : seems as good as any.
> 
> How about splitting the two config items into separate
> properties?  E.g., what you have as "native:crlf" would instead
> be something like:
>     [...]
>     eof = native
>     repo-eof = crlf

No.

Although your proposal looks interesting to you or someone else,
it introduce new complexity to the system. And this complexity
is not needed at all, and should be avoided.

Format of the repo matters NOT to administrators or bzr guru,
but for existing windows users. Why you people don't understand this?

Please, if you're not windows bzr user, don't force new complexity
to this system.

Most people will use only simple and straightforward options:

* native
* lf
* crlf
* exact

All the point with native:crlf is to support existing repos.
All new repos should be created with lf-only to avoid all sort of undesirable effects.
Don't try to invent better flexible schemes there. They are not needed at all.
This all about support of existing repos. Please use lf-only for your new repo.

Please, think about this one more time.


> 
> and one (presumably) common configuration for Windows developers
> working on cross-platform projects would be:
>     [...]
>     eof = crlf
>     repo-eof = lf
> 
> Your defaulting of the repo-format is easily arranged by having
> the "repo-eof" config property default to "lf".  (Probably "eof"
> should default to "native", but that's a whole other discussion.)
> 
> My property names, "eof" and "repo-eof", are just off-the-cuff
> suggestions, btw; I'm sure they can be improved on.
> 
> Why is this an improvement?
> ===========================
> 
> Part of the difficulty with the current proposal is that cramming
> both config items into one property makes for confusion (witness
> the complaints that it seems "backward", that "crlf:crlf" is
> confusing, etc.)  Worse, it limits flexibility.
> 
> The wt-format and repo-format config items have different
> audiences.  As a few people have pointed out, the person who
> cares about the working-tree format is the user of that working
> tree, i.e. the individual developer.  But if anyone at all cares
> about the choice of repo format, it's the project's owner,
> administrator, source-control guru, etc.; most users don't (and
> shouldn't) give a hoot about repo formats, as long as bzr does
> the Right Thing with the working trees.
> 
> Suppose that enforceable, versionable, project-wide configuration
> were to be implemented some day.  What one would want then for
> configuring EOL treatment would be that the Big Cheese could
> enforceably specify the repo-format, while still allowing each
> individual developer to specify their own wt-format.  The
>     eol = wt-format[:repo-format]
> scheme doesn't allow for that; separate properties would do.
> 
> Even with the current config infrastructure, separate properties
> are still better.  Suppose a Big Cheese wants a repo-format of
> CRLF.  With the one-property scheme, they have to tell all their
> developers: "You can set the first part of 'eof' to whatever you
> like, but you MUST make sure the part after the ':' is 'crlf'".
> With separate properties, they could say "set 'eof' to whatever
> you like, but set 'repo-eof=crlf'".  The two are semantically
> equivalent, but IMHO, the latter is both clearer and less
> error-prone.
> 
>   - Eric
> 
> 
> 
>