Re: CVS corrupts binary files ...

[Paul Sander]

>--- Forwarded mail from address@hidden

>At 1:40 PM -0700 6/17/04, Paul Sander wrote:
>>  >Why would it not work well to use a CVS Wrapper to binhex (uuencode,
>>>etc.) a binary file and then essentially have CVS to only see your
>>>file as a text file?
>>
>>The key is that there's a distinction between text files and mergeable
>>files.  Programs like binhex and uuencode produce text files, but they're
>>not mergeable.

>Well, as I stated...simply avoid the merge features on files that are 
>not mergeable. It's not terribly difficult to determine that a file 
>should not be merged after staring at a binhex'd file for <= 1 
>second, assuming one could not determine from the name of the file 
>that merge functions should not be used.

>>CVS works best on files that are both ASCII and mergeable.

>So, one is required to use the merge functions of CVS? They cannot be avoided?
>If they can be avoided, this point seems irrelevant.

Unfortunately, merging is a central aspect of the concurrent paradigm.
If you avoid merging, you abandon the concurrent model.  Such a thing
can be done, but it's not a natural usage model for CVS.

On the other hand, if CVS embraced additional merge algorithms (that could
be considered degenerate cases, like full content swapping) then CVS might
well work for unmergeable data types.

>--- End of forwarded message from address@hidden