"Mark Benedetto King" wrote
> On Wed, Oct 06, 2004 at 08:52:33PM +0100, Philip Martin wrote:
> > It's unfriendly for blame to behave this way. There is no way to
> > modify the mime-type on old revisions, so if one accidentally
commits
> > a revision with a binary mime-type there is no way to correct it to
> > enable blame to run. Perhaps blame should be changed so that only
the
> > current mime-type is used? Yes, that might allow a "real" binary
file
> > to erroneously take part in blame, but is that likely? I'd guess
that
> > when the mime-type indicates inconsistent "binaryness" the most
likely
> > explanation is an error in setting mime-type rather than a change in
> > the format of the file contents.
> >
>
> If there are truly binary revisions, the blame information is likely
> to be extremely unreliable.
>
> I think the current approach is the safest. However, we may need
> more knobs:
>
> --skip-binary-revisions
>
> Do not include revisions that are binary according to
> svn_mime_type_is_binary() in the blame calculations.
>
> --include-binary-revisions
>
> Include all revisions in the blame calculations.
How likely is it that someone will version a binary file and then re-use
that same file as a text file? If you try to economise on filenames like
that, you cannot expect historical information to be useful anyway.
Maybe there should be a switch to allow skipping of marked binary
revisions, but surely the default should be to presume that the
mime-type has been corrected rather than the content changed format?
Simon
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Oct 11 17:08:54 2004