With the changes I mentioned in an earlier post (modifying RCS to do the
compression), the routine that opens files recognized the magic numbers and
behaved accordingly. Invoking the compression factor could also be changed
by any program that could modify the RCS file.
Our experience with this mechanism shows little performance degradation
because the CPU overhead to do the compression/decompression was outweighed
by network and disk I/O overhead.
>--- Forwarded mail from [EMAIL PROTECTED]
>On Mon, 10 Apr 2000, Noel L Yap wrote:
>> This sounds like a really good idea -- it sounds like a fairly minor change to
>> the source (at least on the surface). I'd opt for gzip/gunzip since it's
>> already distributed as part of the source.
>How would you intend to define which files were compressed? Would you
>envisage it as being repository wide (any files that are not compressed
>get compressed upon next commit) or per file basis, such as adding a 'z'
>optoin to -k such as -kz or -kbz. You would also need to define a level
>of compression would you not?
>Would the compression/decompression add much load to the system at all, I
>could imagine some large repositories it could become rather sluggish.
>--- End of forwarded message from [EMAIL PROTECTED]
