Re: CVS and Binaries

[Guy Scharf]

Sau Dan Lee <address@hidden> wrote:
> 
> For your  case, I think  you'll be better  of saving the  binary
> files with names  containing version numbers (manually  assigned). 
> There is no space lost with this method  (since there is no generic
> way to diff two binary files to produce  a minimal diff result). 
> Moreover, one of the most useful  function of CVS is to  diff
> arbitrary versions.  With binary files, you don't have  this useful
> feature anyway.

I'm puzzled by what would be gained by saving two different versions of 
a binary file as separate entities rather than as a new version.  If you 
have file.doc version 1.1 and commit an update to 1.2, then the 
repository file.doc,v file increases in size.  If you add the new file 
as file-1.doc, you've used approximately the same amount of disk space 
in the repository, haven't you?  The disk space will be in two ,v files 
instead of 1.  At least in browsing through binaries in our repository 
suggests no space savings would result in checking in the files 
separately as opposed to updating an existing file.

It's much more convenient, from a management point of view, to use the 
update mechanism rather than to keep changing file names.  At least for 
binary files that change relatively infrequently.  I don't think that 
using CVS would be good for binary files that changed frequently (such 
as a database).

Guy