Dear Gerhard,
> Also there is the problem of doing
> cp -p to copy a file from somewhere else back into the working copy
Or running ALL your files through a perl or sed script. Most may be
unchanged, but ALL will be updated by CVS as it currently stands.
> Given the size of our code base I'd prefer a CRC over an MD5
> (at 1 second per file overhead, an md5 would cost over an hour extra)
I think the algorithm is not the problem, since md5 can processes
many many megabytes per second.
Like all computing time problems, the real impact is only
apparent when you run it in real life and profile it. Let's try
it and see. While the algorithm is surprisingly fast, it WILL
require that every file be read into memory, something the
timestamp approach doesn't require. So I think the impact will
not be from the checksum (by whichever method) but from reading
the files.
Anyway, the proof of the pudding is in the eating, and I shan't ask
others to put up with my wonderings until I have some code to
show for it!! :-)
Regards,
Mitch.
--
| mailto:[EMAIL PROTECTED] | Not the official view of: |
| mailto:[EMAIL PROTECTED] | Australian Calculator Opn |
| Certified Linux Evangelist! | Hewlett Packard Australia |
