On Apr 29, 2004, at 5:13 PM, Ben Collins-Sussman wrote:
> On Thu, 2004-04-29 at 16:50, Jim Fulton wrote:
>
>>> This is bad, bad, bad. You're basically giving the SVN client
>>> permission to change bytes within binary files. Get ready for
>>> corrupted
>>> binary files!
>>
>> No, because subversion prevents setting this property for binary
>> files.
>>
>
> Ah, good. I had forgotten about that.
That's only as good as that binary detection algorithm. I've several
apps that store their files as XML (opaque to the end user).
Subversion would presumably recognize them as the text that they are
inside. However, if subversion were to successfully (conflict-free)
'merge' a change to such a file, it could easily destroy a modified
working copy. For that reason, I tend to agree with Ben's earlier
comment about how it is 'bad,bad,bad' to assume files aren't binary.
Along my line of thinking, all files should be
"svn:mime-type=application/octet-stream" by default for safety.
I do realize that I jumped to talking about the svn:mime-type property
from a conversation about the svn:eol-style property. The similar
thread is my interest in using the most conservative, safe defaults to
avoid corrupted data.
-Travis
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Fri Apr 30 01:55:39 2004